Acceptance Sampling in

Quality Control
Third Edition

1.00
PR
1–α

0.50
Pa

β
CR
0 p
PQL IQ CQL
 Statistics: Textbooks and Monographs
A Series Edited by
D.B. Owen

Volume 1: The Generalized Jackknife Statistic, H. L. Gray and W. R. Schucany

Volume 2: Multivariate Analysis, Anant M. Kshirsagar
Volume 3: Statistics and Society, Water T. Federer
Volume 4: 
Multivariate Analysis: A Selected and Abstracted Bibliography, 1957–1972,
Kocherlakota Subrahmaniam and Kathleen Subrahmaniam (out of print)
Volume 5: D
 esign of Experiments: A Realistic Approach, Virgil L. Anderson and
Robert A. McLean
Volume 6: Statistical and Mathematical Aspects of Pollution Problems, John W. Pratt
Volume 7: Introduction to Probability and Statistics (in two parts) Part I: Probability;
Part II: Statistics, Narayan C. Giri
Volume 8: Statistical Theory of the Analysis of Experimental Designs, J. Ogawa
Volume 9: Statistical Techniques in Simulation (in two parts), Jack P. C. Kleijnen
Volume 10: Data Quality Control and Editing, Joseph I. Naus (out of print)
Volume 11: Cost of Living Index Numbers: Practice, Precision, and Theory, Kali S. Banerjee
Volume 12: Weighing Designs: For Chemistry, Medicine, Economics, Operations Research,
Statistics, Kali S. Banerjee
Volume 13: The Search for Oil: Some Statistical Methods and Techniques, edited by D. B.
Owen
Volume 14: 
Sample Size Choice: Charts for Experiments with Linear Models, Robert
E. Odeh and Martin Fox
Volume 15: Statistical Methods for Engineers and Scientists, Robert M. Bethea, Benjamin
S. Duran, and Thomas L. Boullion
Volume 16: Statistical Quality Control Methods, Irving W. Burr
Volume 17: On the History of Statistics and Probability, edited by D. B. Owen
Volume 18: Econometrics, Peter Schmidt
Volume 19: Sufficient Statistics: Selected Contributions, Vasant S. Huzurbazar (edited by
Anant M. Kshirsagar)
Volume 20: Handbook of Statistical Distributions, Jagdish K. Patel, C. H. Kapadia, and
D. B. Owen
Volume 21: Case Studies in Sample Design, A. C. Rosander
Volume 22: 
Pocket Book of Statistical Tables, compiled by R. E. Odeh, D. B. Owen,
Z. W. Birnbaum, and L. Fisher
Volume 23: The Information in Contingency Tables, D. V. Gokhale and Solomon Kullback
Volume 24: Statistical Analysis of Reliability and Life-Testing Models: Theory and Methods,
Lee J. Bain
Volume 25: Elementary Statistical Quality Control, Irving W. Burr
Volume 26: 
An Introduction to Probability and Statistics Using BASIC, Richard
A. Groeneveld
Volume 27: Basic Applied Statistics, B. I. Raktoe and J. J. Hubert
Volume 28: A Primer in Probability, Kathleen Subrahmaniam
Volume 29: Random Processes: A First Look, R. Syski
Volume 30: Regression Methods: A Tool for Data Analysis, Rudolf J. Freund and Paul
D. Minton
Volume 31: Randomization Tests, Eugene S. Edgington
Volume 32: Tables for Normal Tolerance Limits, Sampling Plans, and Screening, Robert
E. Odeh and D. B. Owen
Volume 33: Statistical Computing, William J. Kennedy, Jr. and James E. Gentle
Volume 34: Regression Analysis and Its Application: A Data-Oriented Approach, Richard
F. Gunst and Robert L. Mason
Volume 35: Scientific Strategies to Save Your Life, I. D. J. Bross
Volume 36: Statistics in the Pharmaceutical Industry, edited by C. Ralph Buncher and
Jia-Yeong Tsay
Volume 37: Sampling from a Finite Population, J. Hájek
Volume 38: Statistical Modeling Techniques, S. S. Shapiro
Volume 39: Statistical Theory and Inference in Research, T. A. Bancroft and C.-P. Han
Volume 40: Handbook of the Normal Distribution, Jagdish K. Patel and Campbell B. Read
Volume 41: Recent Advances in Regression Methods, Hrishkesh D. Vinod and Aman Ullah
Volume 42: Acceptance Sampling in Quality Control, Edward G. Schilling
Other Volumes in Preparation
Acceptance Sampling in
Quality Control
Third Edition

Edward G. Schilling and Dean V. Neubauer

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2017 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

Version Date: 20161028

International Standard Book Number-13: 978-1-4987-3357-1 (Hardback)

This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the
validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copy-
right holders of all material reproduced in this publication and apologize to copyright holders if permission to publish
in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know
so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or
utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including pho-
tocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission
from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA
01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users.
For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been
arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Schilling, Edward G., 1931- | Neubauer, Dean V.

Title: Acceptance sampling in quality control/Edward G. Schilling,
Dean V. Neubauer.
Description: Third edition. | Boca Raton: CRC Press, 2017. | Includes
bibliographical references and indexes.
Identifiers: LCCN 2016039621 | ISBN 9781498733571 (hardback)
Subjects: LCSH: Acceptance sampling. | Sampling (Statistics)
Classification: LCC TS156.4 .S34 2017 | DDC 658.5/62–dc23 LC record
available at https://lccn.loc.gov/2016039621

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com
This text is dedicated to the memory of

Dr. Edward G. Schilling

1931–2008

and

My Parents

Virgil A. Neubauer

1931–2004

and

Fanchon K. Neubauer

1936–2014

and

To our wives Jean and Kimberly

Contents

Preface to the Third Edition.........................................................................................................xix

Note from the Series Editor for the First Edition......................................................................xxi
Foreword to the First Edition................................................................................................... xxiii
Preface to the First Edition..........................................................................................................xxv
Acknowledgments from the First Edition............................................................................. xxvii
Preface to the Second Edition....................................................................................................xxix
Authors.........................................................................................................................................xxxi
List of Tables............................................................................................................................. xxxiii

1. Introduction..............................................................................................................................1
Acceptance Quality Control....................................................................................................1
Acceptance Control and Process Control..............................................................................4
Process Quality Control...........................................................................................................4
Background of Acceptance Quality Control.........................................................................7
Top 10 Reasons for Acceptance Sampling.............................................................................9
Problems...................................................................................................................................10
References................................................................................................................................10

2. Probability and the Operating Characteristic Curve......................................................13

Probability................................................................................................................................13
Classical Definition............................................................................................................13
Empirical Definition..........................................................................................................13
Subjective Definition..........................................................................................................14
Random Samples and Random Numbers...........................................................................14
Counting Possibilities.............................................................................................................15
Probability Calculus...............................................................................................................19
Operating Characteristic Curve............................................................................................23
Software Applications............................................................................................................27
Excel .....................................................................................................................................27
Minitab.................................................................................................................................29
Problems...................................................................................................................................32
References................................................................................................................................33

3. Probability Functions............................................................................................................35
Probability Distributions.......................................................................................................36
Measures of Distribution Functions.....................................................................................39
Hypergeometric Distribution................................................................................................41
Binomial Distribution.............................................................................................................43
Poisson Distribution...............................................................................................................45
f-Binomial Distribution..........................................................................................................49
Negative Binomial Distribution............................................................................................51
Exponential and Continuous Distributions........................................................................53

ix
x Contents

Weibull Distribution...............................................................................................................55
Normal Distribution...............................................................................................................57
Summary of Distributions.....................................................................................................60
Tables of Distributions...........................................................................................................60
Hypergeometric Tables.....................................................................................................60
Binomial Tables..................................................................................................................63
Poisson Tables.....................................................................................................................64
Negative Binomial Tables.................................................................................................64
Exponential and Weibull Tables.......................................................................................65
Normal Distribution Tables..............................................................................................65
Summary..................................................................................................................................65
Useful Approximations..........................................................................................................66
Tests of Fit................................................................................................................................69
Software Applications............................................................................................................72
Excel .....................................................................................................................................72
Minitab.................................................................................................................................74
Problems...................................................................................................................................79
References................................................................................................................................80

4. Concepts and Terminology..................................................................................................83

Average Run Length of Type B Plans...................................................................................86
Sample Size and Lot Size.......................................................................................................87
Effect of Inspection Error.......................................................................................................90
Rectification.............................................................................................................................92
Curtailment..............................................................................................................................94
Tolerance and Confidence Intervals.....................................................................................97
Levels and Risks......................................................................................................................99
Choosing Quality Levels......................................................................................................101
Classification of Defects.......................................................................................................102
Measures of Sampling Plans: Terminology.......................................................................102
Graphs of Measures..............................................................................................................104
Specifying a Plan...................................................................................................................107
Software Applications..........................................................................................................108
Minitab...............................................................................................................................108
Problems................................................................................................................................. 110
References.............................................................................................................................. 111

5. Single Sampling by Attributes......................................................................................... 113

Operation............................................................................................................................... 113
Selection................................................................................................................................. 113
Tables of Poisson Unity Values...................................................................................... 114
Binomial Nomograph...................................................................................................... 116
f-Binomial Nomograph................................................................................................... 116
Thorndyke Chart.............................................................................................................. 118
Hypergeometric Tables................................................................................................... 119
Measures................................................................................................................................121
Software Applications..........................................................................................................123
Contents xi

Excel...................................................................................................................................123
Minitab...............................................................................................................................125
Statgraphics.......................................................................................................................127
Problems.................................................................................................................................130
References..............................................................................................................................131

6. Double and Multiple Sampling by Attributes..............................................................133

Operation...............................................................................................................................134
Double Sampling..............................................................................................................134
Multiple Sampling...........................................................................................................134
Selection.................................................................................................................................135
Measures................................................................................................................................141
Double Sampling..............................................................................................................141
Multiple Sampling...........................................................................................................143
Further Considerations........................................................................................................151
Software Applications..........................................................................................................151
Excel...................................................................................................................................151
Minitab...............................................................................................................................154
Statgraphics.......................................................................................................................154
Problems.................................................................................................................................156
References..............................................................................................................................157

7. Sequential Sampling by Attributes.................................................................................159

Operation...............................................................................................................................159
Selection.................................................................................................................................161
Measures................................................................................................................................163
Sequential Sampling for Defects per Unit.........................................................................165
Software Applications..........................................................................................................166
Excel...................................................................................................................................166
Problems.................................................................................................................................170
References..............................................................................................................................171

8. Variables Sampling for Process Parameter.....................................................................173

Single Sampling for Process Parameter.............................................................................173
Acceptance Control Charts..................................................................................................175
Sequential Plans for Process Parameter (σ Known).........................................................179
Sequential Plans for Process Parameter (σ Unknown)....................................................187
Cumulative Sum Charts.......................................................................................................190
Problems.................................................................................................................................193
References..............................................................................................................................194

9. Bulk Sampling......................................................................................................................195
Construction of the Sample.................................................................................................196
Estimation..............................................................................................................................198
Sampling Plans......................................................................................................................207
Simple Random Sampling of a Unique Lot (Components of Variance Unknown)....209
Sampling from Stream of Lots............................................................................................212
xii Contents

Estimation of Testing and Reduction Variances..........................................................212

Estimation of Segment and Increment Variances........................................................213
Application of Plan to Stream of Lots...........................................................................214
ISO 10725................................................................................................................................216
Problems.................................................................................................................................218
References..............................................................................................................................219

10. Sampling by Variables for Proportion Nonconforming..............................................221

Specification Limits..............................................................................................................222
Assumptions and Theory....................................................................................................223
Operation...............................................................................................................................225
X Method..........................................................................................................................225
k Method............................................................................................................................225
Double Specification Limits............................................................................................226
Selection.................................................................................................................................231
Tables.................................................................................................................................231
Formulas............................................................................................................................232
Jacobson Nomograph for Plan Selection......................................................................233
Measures................................................................................................................................234
Jacobson Nomograph for Operating Characteristics..................................................234
Calculation: σ Known......................................................................................................235
Calculation: σ Unknown.................................................................................................236
Double Specification Limits............................................................................................240
Measures of Performance...............................................................................................242
M Method...............................................................................................................................243
Plans Based on Sample Range............................................................................................248
Example.............................................................................................................................250
Double Sampling by Variables............................................................................................252
Tolerance Intervals and Variables Plans for Percent Nonconforming..........................255
Sequential Plans for Proportion Nonconforming............................................................259
Further Considerations........................................................................................................260
Derivation of n, k Formulas............................................................................................260
Need for Normality.........................................................................................................262
Software Applications..........................................................................................................262
Minitab...............................................................................................................................262
Statgraphics.......................................................................................................................269
Problems.................................................................................................................................273
References..............................................................................................................................274

11. Attributes Sampling Schemes...........................................................................................277

Sampling Schemes................................................................................................................277
Quick Switching Systems....................................................................................................277
Tightened–Normal–Tightened Plans.................................................................................279
MIL-STD-105E and Derivative Standards.........................................................................279
Operation...............................................................................................................................281
Selection.................................................................................................................................283
Measures................................................................................................................................286
Scheme Properties.................................................................................................................287
Implementation of MIL-STD-105E.....................................................................................289
Contents xiii

Matching Individual Sampling Plans to MIL-STD-105E System Performance...........294

MIL-STD-105 Derivatives....................................................................................................296
ANSI/ASQ Standard Z1.4..............................................................................................297
ASTM International Standard E2234.............................................................................297
ISO Standard 2859............................................................................................................297
Further Considerations........................................................................................................299
Software Applications..........................................................................................................300
Minitab...............................................................................................................................300
Snap Sampling Plans!......................................................................................................300
Statgraphics.......................................................................................................................301
Problems.................................................................................................................................306
References..............................................................................................................................307

12. Variables Sampling Schemes............................................................................................309

MIL-STD-414.........................................................................................................................309
Operation............................................................................................................................... 311
Selection.................................................................................................................................313
Measures................................................................................................................................316
Implementation of Form 2...................................................................................................317
Implementation of Form 1...................................................................................................319
Implementation of Plans for Range and Variability Known..........................................320
Match between MIL-STD-414 and MIL-STD-105E..........................................................320
Conversion of MIL-STD-414 to ANSI/ASQ Z1.9.............................................................321
MIL-STD-414 Derivatives....................................................................................................322
ANSI/ASQ Z1.9...............................................................................................................322
ASTM International Standard E2762.............................................................................322
ISO 3951-1..........................................................................................................................323
Further Considerations........................................................................................................326
Software Applications..........................................................................................................326
Excel...................................................................................................................................326
Minitab...............................................................................................................................327
Statgraphics.......................................................................................................................327
Problems.................................................................................................................................334
References..............................................................................................................................335

13. Special Plans and Procedures............................................................................................337

No-Calc Plans........................................................................................................................337
Lot Plot Plans.........................................................................................................................339
Narrow-Limit Gauging........................................................................................................344
Mixed Variables: Attributes Plan........................................................................................354
Advantages and Disadvantages of Mixed Plans.........................................................356
Generalized Mixed Dependent Procedure...................................................................357
Measures: Independent Mixed Plan..............................................................................358
Measures: Dependent Mixed Plan.................................................................................360
MIL-STD-414 Dependent Mixed Plans.........................................................................365
Comparison of Independent and Dependent Mixed Plans.......................................366
Comparison of Mixed with Other Type Plans.............................................................367
Problems.................................................................................................................................368
References..............................................................................................................................369
xiv Contents

14. Series of Lots: Rectification Schemes...............................................................................371

Single-Sampling AOQL Plan...............................................................................................372
Dodge–Romig Sampling Scheme.......................................................................................373
Operation..........................................................................................................................374
Selection.............................................................................................................................374
Measures............................................................................................................................381
Further Considerations...................................................................................................381
Constructing LTPD Plan with Minimum ATI.........................................................386
Constructing AOQL Plan with Minimum ATI........................................................388
Anscombe Rectifying Inspection Procedure.....................................................................391
Operation..........................................................................................................................391
Selection.............................................................................................................................392
Measures............................................................................................................................394
Credit-Based Schemes..........................................................................................................395
Dodge–Romig Plan Derivatives.........................................................................................398
Software Applications..........................................................................................................398
Excel...................................................................................................................................398
Snap Sampling Plans!......................................................................................................405
Problems.................................................................................................................................407
References..............................................................................................................................407

15. Continuous Sampling Plans..............................................................................................409

Dodge Continuous Plans.....................................................................................................409
Dodge CSP-1.....................................................................................................................409
Dodge–Torrey CSP-2 and CSP-3....................................................................................412
Measures of CSP-1, 2, and 3............................................................................................415
Stopping Rules and Selection of CSP-1 Plans..............................................................420
Multilevel Plans....................................................................................................................423
Tightened Multilevel Plans..................................................................................................427
Block Continuous Plans.......................................................................................................428
Wald–Wolfowitz Plan......................................................................................................428
Girshick Plan.....................................................................................................................430
MIL-STD-1235B.....................................................................................................................431
MIL-STD-1235B Derivatives................................................................................................439
ASTM International Standard E2819.............................................................................439
Software Applications..........................................................................................................439
Excel...................................................................................................................................439
Snap Sampling Plans!......................................................................................................442
Problems.................................................................................................................................443
References..............................................................................................................................444

16. Cumulative Results Plans..................................................................................................445

Skip-Lot Sampling Plans......................................................................................................445
SkSP-1................................................................................................................................445
SkSP-2................................................................................................................................447
Chain Sampling Plans..........................................................................................................453
ChSP-1...............................................................................................................................453
Two-Stage Plans...............................................................................................................457
Contents xv

Deferred Sentencing Schemes.............................................................................................458

Demerit Rating Plan.............................................................................................................462
CRC Plan................................................................................................................................466
Software Applications..........................................................................................................468
Excel...................................................................................................................................468
Problems.................................................................................................................................474
References..............................................................................................................................475

17. Compliance Sampling.........................................................................................................477

LSP..........................................................................................................................................478
Procedure..........................................................................................................................478
Protection..........................................................................................................................479
Producer’s Risk................................................................................................................481
Examples of LSP Applications.......................................................................................482
Further Considerations...................................................................................................484
TNT Scheme...........................................................................................................................484
Procedure..........................................................................................................................485
Protection..........................................................................................................................485
Selection.............................................................................................................................487
Quick Switching System......................................................................................................489
MIL-STD-1916.......................................................................................................................494
Structure............................................................................................................................494
Operation..........................................................................................................................494
Implementation................................................................................................................496
Measures............................................................................................................................498
Further Considerations...................................................................................................498
Simplified Grand Lot Procedure.........................................................................................499
Simon’s Approach............................................................................................................500
Simplified Procedure: Attributes...................................................................................502
Example: Attributes.........................................................................................................503
Simplified Procedure: Variables.....................................................................................504
Continuing Series of Lots................................................................................................510
Further Considerations...................................................................................................513
Nomograph for Samples Having Zero Defectives...........................................................515
Accept on Zero Plans............................................................................................................515
U.S. Department of Defense Approach.........................................................................515
Squeglia Plans...................................................................................................................516
AoZ and AQL Plans.........................................................................................................517
Chain Sampling Alternative...........................................................................................519
The Zero Acceptance Number Chained Quick Switching System...........................521
Summary...........................................................................................................................524
Software Applications..........................................................................................................525
Excel...................................................................................................................................525
LSP Plans......................................................................................................................525
TNT Plans.....................................................................................................................525
QSS Plans......................................................................................................................527
MIL-STD-1916 Plans...................................................................................................537
Chain Sampling Alternative Plans............................................................................538
Zero Acceptance Number Chained Quick Switching System Plans...................538
xvi Contents

Minitab...............................................................................................................................543
Snap Sampling Plans!......................................................................................................544
LSP Plans......................................................................................................................544
Squeglia Plans..............................................................................................................545
Statgraphics.......................................................................................................................545
MIL-STD-1916 Plans...................................................................................................545
Problems.................................................................................................................................547
References..............................................................................................................................548

18. Reliability Sampling...........................................................................................................551

Censored Sampling...............................................................................................................552
Variables Plans for Life Testing and Reliability (Juran 1999)..........................................554
Handbook H-108...................................................................................................................555
Operation..........................................................................................................................556
Failure Terminated......................................................................................................558
Time Terminated..........................................................................................................558
Sequential.....................................................................................................................558
Proportion Failing by Specified Time.......................................................................558
Selection........................................................................................................................558
Example of H-108 Application.......................................................................................561
Failure Terminated......................................................................................................561
Time Terminated..........................................................................................................561
Sequential.....................................................................................................................563
Proportion of Lot Failing by Specified Time...........................................................563
Measures.......................................................................................................................564
Further Considerations...................................................................................................564
Technical Report TR7............................................................................................................565
Mean Life Criterion..........................................................................................................566
Hazard Rate Criterion.....................................................................................................568
Reliable Life Criterion.....................................................................................................568
TR7 Tables.........................................................................................................................568
Operation..........................................................................................................................569
TR7 with the MIL-STD-105E System.............................................................................576
Further Considerations...................................................................................................576
Problems.................................................................................................................................578
References..............................................................................................................................579

19. Administration of Acceptance Sampling........................................................................581

Selection and Implementation of a Sampling Procedure................................................583
Determining Quality Levels................................................................................................586
Setting AQL.......................................................................................................................587
Setting AOQL....................................................................................................................590
Setting IQ...........................................................................................................................590
Setting LTPD (or LQ).......................................................................................................590
Relation of Levels.............................................................................................................591
Setting PQL and CQL......................................................................................................593
Contents xvii

Economic Considerations....................................................................................................593
Mandatory Standards...........................................................................................................595
Basic Principle of Administration.......................................................................................596
Problems.................................................................................................................................596
References..............................................................................................................................597

Answers to Problems..................................................................................................................599
Appendix......................................................................................................................................615
Index..............................................................................................................................................829
Preface to the Third Edition

As I prepared to work on this edition, I wondered what has changed since the last edition,
what topics may have been omitted from that edition due to time, and which topics would be
of most interest to readers. I spent several years converting many of the old military accep-
tance sampling standards to American Society for Testing and Materials (ASTM) ­standards,
which means that they are now regularly reviewed and supported. Companies that have
traditionally used military standards can now refer to these new ASTM ­standards as they
contain the same tables. The reader will find out more about these ASTM s­ tandards in the
chapters covering these types of sampling plans.
As for omitted topics, I considered the addition of multivariate sampling and sampling
plans based on Cpk. In the end, I decided that readers would be particularly interested in
two topics—the implementation of computer spreadsheets and software in the design and
the evaluation of sampling plans and additional material on compliance sampling plans.
The last time I saw Ed Schilling, we met in my RIT office and he expressed his interest
in having me include a CD with the 3rd edition of this book that would include the Excel
templates I had developed for my acceptance sampling students at Rochester Institute of
Technology. Fortunately, since there is now a new book website put up by CRC Press at
www.crcpress.com/product/isbn/9781498733571, there is no need to include a CD with
the book that cannot be periodically updated. On the book website, the reader will find
not only a variety of Excel spreadsheets but also other files that are discussed at the end of
the appropriate chapters in this book. Ed believed, as I do, that readers are more likely to
understand and utilize the methods covered in this book when provided with a computer
approach. When I taught acceptance sampling, I provided my students with a number of
Excel spreadsheets for several types of plans. I believed that it was more likely that the stu-
dent could “hit the ground running” so to speak when they had to implement such plans
back on the job. This proved to be the case.
For this edition, I created several more Excel* templates to address sampling plans that
I didn’t use in my teaching, and which even existing commercial software does not handle.
At the end of many of the chapters in this text, I have included a section called “Software
Applications.” In this section, I discuss any Excel templates that I have available for the
reader on the book website as well as whether any commercial software will handle such
problems.
I got some help. I approached a few software companies for assistance in doing this
­edition. Minitab,† Snap Sampling Plans!,‡ and Statgraphics§ have all been very generous in
providing me with either materials relating to how their program can design and analyze
acceptance sampling plans or a complimentary license for their product so I can showcase

* Excel is a registered trademark of Microsoft Inc. You must purchase Microsoft Office to install Excel as no free
demos are available.
† Minitab is a registered trademark of Minitab, Inc. A 30-day free demo of the software is available at

www.minitab.com.
‡ Snap Sampling Plans! is a registered trademark of Quality Assurance Solutions. The software home page can be

found at http://www.quality-assurance-solutions.com/Snap-Sampling-Plans.html. No demo is available.

§ Statgraphics is a registered trademark of Statpoint Technologies, Inc. A 30-day free demo of the software is

available at http://www.statgraphics.com/download_trial.

xix
xx Preface to the Third Edition

its features for acceptance sampling. At Minitab, I specifically thank Lou Johnson and his
associates, especially Yanling Zuo with whom I worked with to develop the initial accep-
tance sampling procedures included in Version 15 and that still exist in later versions.
Minitab also supplied some training materials that I could adapt to the examples in this
book. I also thank Dr. Neil Polhemus, CTO and director of Development for Statgraphics,
for his willingness to work with me to incorporate their Statgraphics Centurion XVII prod-
uct into this book. Finally, I thank Robert Broughton of Quality Assurance Solutions who
was also very supportive of my efforts to include his program, Snap Sampling Plans!,
among the suite of programs illustrated in this book to demonstrate the capabilities of his
software. Minitab, Snap Sampling Plans!, and Statgraphics represent excellent examples of
commercial software that can handle most of the acceptance sampling situations covered
in this book. There are a number of other lesser-known acceptance sampling programs for
sale on the Internet and a simple Internet search can find many of them.
One topic that continues to be important in the field of acceptance sampling is that of
compliance plans. The most common forms of these plans are accept on zero (AoZ) sam-
pling plans popularized by Nick Squeglia and his book Zero Acceptance Number Sampling
Plans (now in its 5th edition). As in the 2nd edition, Chapter 17 discusses lot sampling
plans, tightened-normal-tightened (TNT) plans, quick switching system (QSS) plans, MIL-
STD-1916, the simplified grand lot procedure, AoZ plans, and a chain sampling alterna-
tive. I expanded the section on QSSs as there has been considerable activity in this area
over the past more than 20 years. Much of this work comes from universities in India and
New Zealand, and the reader will find many references for these authors in Chapter 17.
Furthermore, recent work in 2011 by Govindaraju combining the ideas of many of these
types of plans into a zero acceptance number chained QSS has been added to this chapter.
The reader will find that many of the plans discussed in this chapter are supported by
Excel templates, Minitab, Snap Sampling Plans!, and Statgraphics.
Once again I thank my editor David Grubbs at Taylor & Francis Group for his patience
and understanding as I put in more time than expected in preparing this new edition due
to some health issues. Finally, and certainly not the least, I thank my wife, Kimberly, for all
her support as I worked on this edition. Without her having my back, this edition would
not have been possible.

Dean V. Neubauer
Horseheads, New York
Note from the Series Editor for the First Edition

The use of acceptance sampling has grown tremendously since the Dodge and Romig
Sampling Inspection Tables were first widely distributed in 1944. Throughout this period,
many people have contributed methods and insight to the subject. One of these contribu-
tors is the author of this book, which might better be identified as a compendium of accep-
tance sampling methods. The American Society for Quality Control has recognized
Dr. Schilling’s contributions by awarding him the Brumbaugh Award four times, first in
1973 and again in 1978, 1979, and 1981. This award is given each year to the author of
that paper published in either the Journal of Quality Technology or Quality Progress that the
American Society for Quality Control committee judges has made the largest single contri-
bution to the development of industrial applications of quality control.
Dr. Schilling has been employed both as an educator and as an industrial statistician.
This broad experience qualifies him to write this treatise as few others are qualified. The
beginner will find much interest in this work, while the experienced person will also find
many interesting items because of its encyclopedic coverage.
I am very pleased with the completeness and clarity exhibited in this book, and it is with
great pleasure that I recommend it to others for their use.

D. B. Owen
Department of Statistics
Southern Methodist University
Dallas, Texas

xxi
Foreword to the First Edition

As the field of quality control enters the 1980s, it is having new responsibilities thrust upon
it. The public is demanding products free from defects and often making these demands in
costly court cases. Management is demanding that all departments contribute to technical
innovation and cost reduction while still continuing to justify its own costs. The quality
control specialist is caught like others in this squeeze between perfect performance and
minimum cost. He or she needs all the help that fellow professionals can give, and Edward
Schilling’s book is a worthy contribution. Written by one of the foremost professionals in
the field, it is comprehensive and lucid. It will take its place as a valuable reference source
in the quality control specialist’s library.
My own first contact with a draft of the book came when I was teaching a quality con-
trol course to industrial engineers. Over the semester, I found myself turning to this new
source for examples, for better explanations of standard concepts, and for the many charts,
graphs, and tables, which are often difficult to track down from reference. Acceptance sam-
pling is not the whole of statistical quality control, much less the whole of quality control.
But Dr. Schilling has stuck to his title and produced a text of second-level depth in this one
area, resisting the temptation to include the other parts of quality control to make a “self-
contained work.” The added depth in this approach makes the text a pleasure for a teacher
to own and will make it a pleasure for students to use. This is one book that any student
should take into the world where knowledge is applied to the solution of problems.

Colin G. Drury
Department of Industrial Engineering
SUNY at Buffalo
Buffalo, New York

xxiii
Preface to the First Edition

The methods of statistical acceptance sampling in business and industry are many and
varied. They range from simple to profound, from practical to infeasible and naive. This
book is intended to present some of the techniques of acceptance quality control that are
best known and most practical—in a style that provides sufficient detail for the novice,
while including enough theoretical background and reference material to satisfy the more
discriminating and knowledgeable reader. The demands of such a goal have made it neces-
sary to omit many worthwhile approaches; however, it is hoped the student of acceptance
sampling will find sufficient material herein to form a basis for further explorations of the
literature and methods of the field.
While the prime goal is the straightforward presentation of methods for practical appli-
cation in industry, sufficient theoretical material is included to allow the book to be used as
a college-level text for courses in acceptance sampling at a junior, senior, or graduate level.
Proofs of the material presented for classroom use will be found in the references cited. It is
assumed, however, that the reader has some familiarity with statistical quality control pro-
cedures at least at the level of Irving W. Burr’s Statistical Quality Control Methods (Marcel
Dekker, Inc., New York, 1976). Thus, an acceptance sampling course is a natural sequel to
a survey course at the level suggested.
The text begins with a fundamental discussion of the probability theory necessary for
an understanding of the procedures of acceptance sampling. Individual sampling plans
are then presented in increasing complexity for use in the inspection of single lots. There
follows a discussion of schemes that may be applied to the more common situation of a
stream of lots from a steady supplier. Finally, specific applications are treated in the areas
of compliance sampling and reliability. The last chapter is concerned with the administra-
tion of acceptance control and, as such, is intended as a guide to the user of what sampling
plan to use (and when). Readers having some familiarity with acceptance sampling may
wish to read the last chapter first to put into context the methods presented.
This book views acceptance quality control as an integral and necessary part of a total
quality control system. As such, it stands with statistical process quality control as a bul-
wark against poor-quality products, whose foundations are rooted deep in mathematics
but whose ramparts are held only by the integrity and competence of its champions in the
heat of confrontation.
It is fitting that this book on acceptance sampling should begin with the name of
Harold F. Dodge. His contributions have been chronicled and are represented in the Dodge
Memorial Issue of the Journal of Quality Technology (Vol. 9, No. 3, July 1977). Professor
Dodge, as a member of that small band of quality control pioneers at the Bell Telephone
Laboratories of the Western Electric Company, is considered by some to be the father of
acceptance sampling as a statistical science. Certainly, he nurtured it, lived with it, and fol-
lowed its development from infancy, through adolescence, and on into maturity. In no small
way he did the same for the author’s interest in the field, as his professor and his friend.

Edward G. Schilling

xxv
Acknowledgments from the First Edition

Books are not made—they grow. It is impossible to acknowledge all the help and support
that has come from friends and associates in the development and construction of the pres-
ent volume. A few may be singled out not only for their individual contributions but also
as a sample of those yet unnamed. In particular, I thank Carl Mentch for suggesting the
possibility of such an undertaking in September of 1965 and for his unflagging encourage-
ment and help since that time. My thanks also go to Lucille I. Johnson whose technical and
editorial assistance helped to bring concept into reality. I also mention Dr. Lloyd S. Nelson
for his continued interest and suggestions and Dan J. Sommers and Professor Emil Jebe
for their constructive comments and theoretical insights. Certainly, my appreciation goes
to Dr. Donald P. Petarra, Dr. James R. Donnalley, and Dr. Pieter J. von Herrmann of the
General Electric Lighting Research and Technical Services Operation for their encourage-
ment and support throughout.
I am indebted to the American Society for Quality Control, the American Society
for Testing and Materials, the American Statistical Association, the Institute of
Mathematical Statistics, the Philips Research Laboratories, the Royal Statistical Society,
and Bell Laboratories for the permission to reprint a variety of materials taken from
their publications. I am also indebted to Addison-Wesley Publishing, Inc., for the per-
mission to reprint material from D. B. Owen, Handbook of Statistical Tables; to Cambridge
University Press for the permission to reprint material from E. S. Pearson, Tables of the
Incomplete Beta-Function; to McGraw-Hill Book Company for the permission to reprint
material from A. H. Bowker and H. P. Goode, Sampling Inspection by Variables, I. W. Burr,
Engineering Statistics and Quality Control, and J. M. Juran, Quality Control Handbook; to
John Wiley and Sons, Inc., and Bell Laboratories for the permission to reprint mate-
rial from H. F. Dodge and H. G. Romig, Sampling Inspection Tables; to Prentice Hall,
Inc., for the permission to reprint material from A. H. Bowker and G. J. Lieberman,
Engineering Statistics; to Stanford University Press for the permission to reprint material
from G. J. Lieberman and D. B. Owen, Tables of the Hypergeometric Probability Distribution,
and G. J. Resnikoff and G. J. Lieberman, Tables of the Non-Central t-Distribution; to the
European American Music Distributors Corporation for the permission to use the
English translation of “O Fortuna” from Carl Orff’s scenic cantata Carmina Burana; and
to my associates K. S. Stephens, H. A. Lasater, L. D. Romboski, R. L. Perry, and J. R.
Troxel for the material from their PhD dissertations taken at Rutgers University under
Professor Harold F. Dodge in a unique intellectual environment that was created and
sustained at the Statistics Center under the inspired direction of Dr. Ellis R. Ott and with
the ­dedicated administrative support of Dr. Mason E. Wescott.
Finally, these debts of gratitude are in terms of time and talent, how much more the debt
to my wife, Jean, and to my daughters, Elizabeth and Kathryn, who are as much a part of
this book as the author himself.

xxvii
Preface to the Second Edition

So why does another edition of this acceptance sampling text make sense? For one reason,
it is important to mention the paucity of research and publication in acceptance control
that took place in the last two decades of the twentieth century—the life of the first edition.
This is the result of a serious misunderstanding of the role of acceptance sampling in qual-
ity and process control. Thus, this period of time accounts for the fewest references in this
edition. Nonetheless, the twenty-first century exhibits a resurgence of interest in this field.
In fact, the inevitable passage of time and the events associated with it have made a sec-
ond edition of this book desirable. International trade has become the hallmark of a global
economy. However, in many cases, the producer has become increasingly removed from
the consumer, not only by distance but also by language, culture, and governmental differ-
ences. This has accentuated the need for economic appraisal of material as it passes from
border to border through the global maze. State-of-the-art knowledge of the methodology
of sampling and its advantages and limitations is essential in an environment such as this.
At the same time, corporate culture has changed in response to an intensely competitive
business environment. Manufacturers must protect the quality of their product in the most
efficient and economic way possible. Judicious use of acceptance control can supplement
and support applications of statistical process control. Used alone, it provides a proven
resource for the evaluation of product.
From a global perspective, the International Organization for Standardization (ISO) has
contributed greatly in recent years. As derivatives of MIL-STD-105E and MIL-STD-414,
their ISO 2859 and ISO 3951 standards are part of a series of standards that have been cre-
ated to address the role that acceptance sampling plays when dealing with a flow of prod-
uct, with emphasis on the producer’s process. Both attributes and variables plans comprise
the series. This edition covers these standards while tracing their origins to 105 and 414.
This includes conversion to the new definition of AQL and changes in the switching rules.
Credit-based sampling plans were originated in Europe and are the subject of a new sec-
tion in this text. These developments are intended to enhance the use of acceptance control
in international trade.
In a highly competitive environment, acceptance sampling plans must be appropriately
applied. This edition stresses the role of sampling schemes with switching rules in making
sampling more efficient when dealing with a flow of lots. An increased number of deriva-
tive plans are dealt with in a historical perspective to simplify comparison and under-
standing. MIL-STD-105E and MIL-STD-414 are taken as the principal standards in that
regard since they will remain constant and, while discontinued by the military, are still
used extensively for domestic applications (e.g., ASTM E2234 conforms to MIL-STD-105E).
The conversion of MIL-STD-414 to ANSI/ASQ Z1.9 is highlighted, and scheme properties
of Z1.9 emphasized. Variables plans matching the Dodge–Romig attributes plans are also
addressed. The ASTM version of 105E (E2234) is now included, along with their version of
TR7 (E2555), which has expanded tables for application of TR-7.
One of the changes in corporate culture is the demand for c = 0, so-called “accept zero”
plans. The legal implications of higher acceptance numbers is said to require this.
While such plans have been shown to have serious disadvantages (R = 44.9), the
demand for such plans is considerable. This edition provides a more extensive discussion
in Chapter 17 of accept zero plans, especially TNT, credit-based, the Nelson monograph

xxix
xxx Preface to the Second Edition

for c = 0, and MIL-STD-1916, along with the QSS and the simplified grand lot plans.
MIL-STD-1916 receives extensive coverage since it is new to the text and provides a
­variety of c = 0 plans. Chapter 19 includes a discussion of how to set quality levels.
These are a few of the changes that were made in the text. Another, and more important,
is the addition of Dean V. Neubauer as coauthor. Dr. Schilling is extremely pleased by this
addition and by the level of expertise that he brings to the text and to the field.
We thank our editor David Grubbs at Taylor & Francis Group for his patience and under-
standing as we put in more time than we expected in preparing this new edition. Also,
we would be remiss not to thank our project coordinator Jessica Vakili at Taylor & Francis
Group for her efforts in typing the manuscript from the first edition so that we could work
with electronic files in our editing process.
Finally, and certainly not the least, we thank our wives, Jean and Kimberly, for all their
love and patience as we toiled on this edition. Without their understanding, this edition
would not have been possible.

Edward G. Schilling
Dean V. Neubauer
Authors

Dr. Edward G. Schilling was a professor emeritus of statistics at the Center for Quality
and Applied Statistics, at Rochester Institute of Technology (RIT), where he had been the
director of the center and the chair of the Graduate Statistics Department.
Before joining RIT, he was the manager of the lighting quality operation for the Lighting
Business Group of General Electric Company. He received his BA and MBA from the State
University of New York (SUNY) at Buffalo and his MS and PhD in statistics from Rutgers
University. He had been on the faculties of SUNY at Buffalo, Rutgers University, and Case
Western Reserve University. He had extensive industrial experience in quality engineering
at Radio Corporation of America and the Carborundum Co. and in statistical consulting
and quality management at General Electric Company. Dr. Schilling was a fellow of the
American Society for Quality (ASQ), the American Statistical Association (ASA), and the
American Society for Testing and Materials (ASTM). He was also a member of the Institute
of Mathematical Statistics and the American Economic Association. He was registered as a
professional engineer in California and certified by ASQ as a quality and a reliability engi-
neer. He served as the founding series editor for the Marcel Dekker series of books on qual-
ity and reliability and published extensively in the field of quality control and statistics.
Dr. Schilling was the Shewhart Medalist in 1983 and the recipient of the E.L. Grant Award
in 1999, the Freund–Marquardt Medal in 2005, and the Distinguished Service Medal in
2002; he was the first person to receive the Brumbaugh Award four times from the ASQ.
He was also the recipient of the Ellis R. Ott Award in 1984 for his contributions to quality
management from the Metropolitan New York Section of that society and was honored by
being invited to present the 1986 W.J. Youden Memorial Address at the Joint ASQ/ASA
Annual Fall Technical Conference. He was the recipient of the H.F. Dodge Award by the
ASTM in 1993 and the Award of Merit in 2002.
Dr. Schilling was an associate editor of the fifth edition of Juran’s Quality Control Handbook.
His book, Process Quality Control (with E.R. Ott and D.V. Neubauer), is among the leading
texts in the field.
Dean V. Neubauer is an engineering fellow in the Process Engineering Directorate in
the Manufacturing, Technology and Engineering Division at Corning Incorporated. He
has worked for Corning Incorporated since 1981 and has been involved in projects in
every major business over this time. He has 15 U.S. patents for compositional work
on a variety of materials and devices for the manufacture of display glass. His back-
ground includes degrees in statistics from Iowa State University and Rochester Institute
of Technology (RIT).
Outside of Corning Incorporated, Dean has taught as an adjunct professor in the MS
program for the Center for Quality and Applied Statistics at RIT since 1992. He also par-
ticipates on the American Society for Testing and Materials (ASTM) E-11 Committee
on Quality and Statistics where he is a former chairman and is a member of the ASTM
International Committee on Standards. Dean is a book reviewer for the Technometrics
and Journal of Quality Technology journals, serves on the Management Committee of the
Technometrics journal, serves on the Editorial Board of the Quality Engineering journal, and
is the coauthor of Process Quality Control, 4th Edition, with the late Dr. Edward G. Schilling.
He is also a former U.S. lead delegate to the International Organization for Standardization

xxxi
xxxii Authors

(ISO) Technical Committee 69 on Statistical Methods Subcommittee SC5 on Acceptance

Sampling. In this role, he was responsible for the review of ISO, as well as ANSI/ASQ Z,
acceptance sampling standards.
Dean was elected a fellow of the American Society for Quality (ASQ) in 2000 and is a
certified quality engineer. Dean is also a former chair of the ASQ’s Chemical and Process
Industries Division and a former chair of the Corning-Elmira ASQ Section. He is a recipient
of the 2003 Shewell Award from ASQ and the 2011 Harold F. Dodge Award from ASTM.
He has also been a member of the American Statistical Association since 1979 and is an
accredited professional statistician (PStat™). In 1993, he was elected a fellow of the Royal
Statistical Society in London, England, and holds the certification of chartered statistician.
In 2010, he was given the Award of Merit carrying the title of fellow from the ASTM.
List of Tables

Table 3.1   Distributions Useful in Acceptance Sampling.....................................................61

Table 4.1   Probability Distributions for Operating Characteristic Curves........................84
Table 4.2   P
 robabilities of Acceptance for Hypergeometric, f-Binomial, Binomial,
and Poisson (N = 20, n = 10, c = 1).........................................................................84
Table 4.3   Conversion of Pa to Average Run Length................................................................87
Table 4.4   T
 ype A Probability of Acceptance Using the Hypergeometric
Distribution (n = 10, c = 1).......................................................................................87
Table 4.5   T
 ype A Probability of Acceptance Using the Hypergeometric
Distribution (N = 20, c = 1)......................................................................................88
Table 4.6   Type B Probability of Acceptance Using the Binomial Distribution (c = 1)...... 88
Table 4.7   T
 ype A Probability of Acceptance Using the Hypergeometric
Distribution (N = 20, n = 10)...................................................................................89
Table 4.8   Type B Probability of Acceptance Using the Binomial Distribution (c/n = .1)...... 89
Table 4.9   T
 ype A Probability of Acceptance Using the Hypergeometric
Distribution when Sample Size is Proportionate to Lot Size.............................90
Table 4.10  Effect of Lot Size on Acceptance (Order of 1000)................................................90
Table 4.11  Proportions Defective Misclassified......................................................................91
Table 4.12  C
 alculation of Average Outgoing Quality and Average Total Inspection
(N = 20, n = 10, c = 1)................................................................................................93
Table 4.13  Determination of Average Outgoing Quality Limit................................................94
Table 4.14  Producer and Consumer Interests.........................................................................99
Table 4.15  Specifications and Defects....................................................................................103
Table 5.1   Procedures for Determining Single-Sampling Plans........................................ 114
Table 5.2   Measures of Single-Sampling Attributes Plans.................................................122
Table 6.1   Measures of Double-Sampling Attributes Plans...............................................141
Table 6.2   Measures of Multiple-Sampling Attributes Plans.............................................144
Table 6.3   Explicit Formulas for Measures of Multiple-Sampling Plans..........................149
Table 7.1   S
 equential Sampling by Attributes for Proportion Defective Points
on the Operating Characteristic, Average Sample Number, and
Average Outgoing Quality Curves......................................................................164
Table 7.2   S
 equential Sampling by Attributes for Defects per Unit Points on the
Operating Characteristic and Average Sample Number Curves....................165

xxxiii
xxxiv List of Tables

Table 8.1   Statistical Tests of Hypotheses.............................................................................174

Table 8.2   Minimum Δ Values before Correction Terms Need Be Used..........................177
Table 8.3   F
 ormulas for Single Upper Limit Sequential Plans for Process
Parameter (For Common Logs, L = 2.3026; for Natural Logs, L = 1)..............180
Table 9.1   Developing a Standard Sampling Method.........................................................196
Table 9.2   Possible Combinations of n1 and n2.....................................................................200
Table 9.3   Analysis of Variance Table for Nested Sampling..............................................202
Table 9.4   Percent Potassium Bitartrate in Shipment of Argol..........................................203
Table 10.1  X Method................................................................................................................225
Table 10.2  k Method (Given n, k)............................................................................................226
Table 10.3  Polygon Derived by the Wallis Method..............................................................231
Table 10.4  Calculation of Probability of Acceptance: n = 7 and k = 1.44...........................237
Table 10.5  C
 alculation of Probability of Acceptance: Double Specification Limits
(Known Standard Deviation σ = 2; n = 7, k = 1.44)............................................241
Table 10.6  M Method Standard Deviation Known (Given n, M).......................................244
Table 10.7  Tensile Strength of Stud Bolts...............................................................................254
Table 10.8  Content of Odeh–Owen Tables of Tolerance Limits.........................................257
Table 10.9  Calculation of k and n for Lower and Upper Specification Limits.................258
Table 11.1  Scheme Pa and ASN Compared to the Normal Plan Only for Single,
Double, and Multiple Sampling (Code F, 2.5% AQL).......................................293
Table 12.1  Matching the Letters..............................................................................................321
Table 12.2  Procedure and Application of ANSI/ASQ Z1.9 and ISO 3951-1....................324
Table 13.1  3σ Values for Lot Plot.............................................................................................342
Table 13.2  Calculation of Probability of Acceptance for n = 15, c = 2, t = 1......................346
Table 13.3  Iterative Use of Ladany Nomograph..................................................................351
Table 13.4  N
 arrow-Limit Plans Substituted for Attributes Plans in MIL-STD-105E,
Code F, 2.5 AQL......................................................................................................354
Table 13.5  Formulas for Measures of Dependent Mixed Plans..........................................360
Table 13.6  C
 omparison of Pa and Average Sample Number for a Specified Mixed
Plan Applied in Dependent and Independent Forms: n1 = 5, n2 = 20,
k = 2, c1 = c2 = 0........................................................................................................367
Table 13.7  C
 omparison of Various Plans to Match p1 = .008, Pa = .953; p2 = .107,
Pa = .098....................................................................................................................368
Table 14.1  D
 odge–Romig Single-Sampling Table for Average Outgoing Quality
Limit = 4.0%............................................................................................................377
List of Tables xxxv

Table 14.2  D
 odge–Romig Double-Sampling Table for Average Outgoing
Quality Limit = 4.0%..............................................................................................378
Table 14.3  D
 odge–Romig Single-Sampling Table for Lot Tolerance Percent
Defective = 4.0%.....................................................................................................379
Table 14.4  D
 odge–Romig Double-Sampling Table for Lot Tolerance Percent
Defective = 4.0%.....................................................................................................380
Table 14.5  D
 odge–Romig Variables Plans for Average Outgoing Quality
Limit = 4.0% Variability Known...........................................................................382
Table 14.6  D
 odge–Romig Variables Plans for Average Outgoing Quality
Limit = 4.0% Variability Unknown......................................................................383
Table 14.7  A
 nscombe Rectifying Inspection Schemes for Lot Tolerance Zt = ptN
with Risk e = ß = .10...............................................................................................393
Table 15.1  Stopping Rules for CSP-1 Plans...........................................................................420
Table 15.2  Type and Purpose of MIL-STD-1235B Plans......................................................432
Table 15.3  Sampling Frequency Code Letters.......................................................................435
Table 15.4  Values of i for CSP-T Plans...................................................................................436
Table 15.5  Values of S for CSP-T Plans..................................................................................436
Table 16.1  Comparison of CSP-1 and SkSP-1.......................................................................446
Table 16.2  Values of Lot AOQL2 for Given f and i................................................................449
Table 16.3  Unity Values for Evaluation of ChSP-1 OC Curves..........................................455
Table 16.4  ChSP-1 Plans Having Given AOQL....................................................................456
Table 16.5  Values of Dp Having Specified Probability of Acceptance..............................460
Table 16.6  Sample Size for Check Inspection and Demerit Rating....................................462
Table 16.7  Important Aspects of Classification of Defects..................................................463
Table 16.8  Nonconformance Criteria.....................................................................................464
Table 17.1  Values of D = Npt Corresponding to f.................................................................480
Table 17.2  Factors for Constructing the OC Curve..............................................................481
Table 17.3  c = 0 Sampling Plan Table.....................................................................................518
Table 17.4  Small Lot Size Supplement Table to Table 17.3..................................................519
Table 17.5  Summary of Input Parameters to Total Cost of Acceptance Sampling..........519
Table 18.1  Probability Positions for Motorette Data............................................................553
Table 18.2  Life Characteristics for Two Failure Distributions............................................555
Table 18.3  Values of γ = [Γ(1 + (1/β))]β for the Weibull Distribution.................................555
Table 18.4  Operation of H-108................................................................................................559
xxxvi List of Tables

Table 18.5   Operation of TR7..................................................................................................571

Table 18.6   Value of Q = t2Z(t2)/t1Z(t1) Corresponding to the Ratio t2/t1........................573
Table 19.1   Life Cycle of Acceptance Control Application.................................................582
Table 19.2   Selection of Plan...................................................................................................584
Table 19.3   C
 omparison of Administrative Aspects of Single, Double, Multiple,
and Sequential Sampling....................................................................................586
Table T1.1   Control Chart Limits for Samples of n.............................................................616
Table T2.1   Random Numbers...............................................................................................617
Table T3.1   Values of e−x........................................................................................................................................................................... 620
Table T3.2   Cumulative Normal Probability, F(z)...............................................................621
Table T3.3   L
 ieberman-Owen Table of Hypergeometric Distribution, F(x)
(N = 2, n = 1 through N = 11, n = 9)...................................................................623
Table T3.4   Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)......629
Table T3.5   Molina Table of the Poisson Distribution, 1 – F(c – 1),
¥ axe - a
Pr ( x ³ c|a ) =
x =c x ! å
.....................................................................................637

Table T5.1   Cameron Table of Unity Values for Constructing Single-Sampling Plans..... 649
Table T5.2   Cameron Table of Unity Values to Determine Probability of Acceptance..... 651
Table T6.1   U
 nity Values for Construction and Evaluation of Single-, Double-,
and Multiple-Sampling Plans............................................................................653
Table T7.1   Statistical Research Group: Table of Sequential Sampling Plans..................662
Table T7.2   Statistical Research Group: Table of Values of a and b for Sequential
1-b 1- a
Sampling a = log b = log α for computing a, β for computing b........671
a b
Table T7.3   g1 and g2 in Terms of p1 and p2 Using Common Logarithms for
æp ö æ 1 - p1 ö
Sequential Sampling g1 = log ç 2 ÷ g 2 = log ç ÷....................................672
è p1 ø è 1 - p2 ø
Table T8.6   Factors for Acceptance Control Limits.............................................................676
Table T8.7   Correction Terms for Acceptance Control Factors..........................................678
Table T8.8   Boundary Values for Barnard’s Sequential t-Test...........................................679
*
Table T10.1  d 2 Factors and Degrees of Freedom ν for Estimating the Standard
Deviation for the Average Range of k Samples of n.......................................680
Table T10.2  M
 atched Single and Double, Known (σ) and Unknown (s)
Standard Deviation, Variables Sampling Plans for Values of p1 and p2
with α = .05, β = .10 (n1 = n2, kt = kr)...................................................................682
Table T10.3  C
 omparison of Approximate and Exact Values of N and k for
Variables Sampling Plans...................................................................................686
List of Tables xxxvii

Table T10.4   Odeh-Owen Table 5: Two-Sided Sampling Plan Factors to Control

Equal Tails..........................................................................................................687
Table T10.5   O
 deh-Owen Table 6: Two-Sided Sampling Plan Factors to Control
Tails Separately..................................................................................................689
Table T11.1   MIL-STD-105E Table VIII—Limit Numbers for Reduced Inspection........691
Table T11.2   MIL-STD-105E Table I—Sample Size Code Letters......................................692
Table T11.3   M
 IL-STD-105E Table II-A—Single-Sampling Plans for Normal
Inspection (Master Table).................................................................................693
Table T11.4   M
 IL-STD-105E Table II-B—Single-Sampling Plans for Tightened
Inspection (Master Table).................................................................................694
Table T11.5   M
 IL-STD-105E Table II-C—Single-Sampling Plans for Reduced
Inspection (Master Table).................................................................................695
Table T11.6   M
 IL-STD-105E Table III-A—Double-Sampling Plans for Normal
Inspection (Master Table).................................................................................696
Table T11.7   M
 IL-STD-105E Table III-B—Double-Sampling Plans for Tightened
Inspection (Master Table).................................................................................697
Table T11.8   M
 IL-STD-105E Table III-C—Double-Sampling Plans for Reduced
Inspection (Master Table).................................................................................698
Table T11.9   MIL-STD-105E Table IV-A—Multiple-Sampling Plans for Normal
Inspection (Master Table).................................................................................699
Table T11.10  M
 IL-STD-105E Table IV-B—Multiple-Sampling Plans for Tightened
Inspection (Master Table).................................................................................703
Table T11.11  M
 IL-STD-105E Table IV-C—Multiple-Sampling Plans for Reduced
Inspection (Master Table).................................................................................707
Table T11.12  M
 IL-STD-105E Table V-A—Average Outgoing Quality Limit Factors
for Normal Inspection (Single Sampling)......................................................710
Table T11.13  M
 IL-STD-105E Table V-B—Average Outgoing Quality Limit Factors
for Tightened Inspection (Single Sampling).................................................. 711
Table T11.14  M
 IL-STD-105E Table VI-A—Limiting Quality (in Percent Defective)
for Which Pa = 10% (for Normal Inspection, Single Sampling)..................712
Table T11.15  M
 IL-STD-105E Table VI-B—Limiting Quality (in Defects per Hundred
Units) for Which Pa = 10% (for Normal Inspection, Single Sampling)........ 713
Table T11.16  M
 IL-STD-105E Table VII-A—Limiting Quality (in Percent Defective)
for Which Pa = 5% (for Normal Inspection, Single Sampling)....................714
Table T11.17  M
 IL-STD-105E Table VII-B—Limiting Quality (in Defects per Hundred
Units) for Which Pa = 5% (for Normal Inspection, Single Sampling)............715
Table T11.19  MIL-STD-105E X-F Tables for Sample Size Code Letter: F..........................717
Table T11.20  M
 IL-STD-105E Scheme Average Outgoing Quality Limit Factors
(in Defects per Hundred Units).......................................................................719
xxxviii List of Tables

Table T11.21  MIL-STD-105E Scheme Limiting Quality (in Defects per Hundred
Units) for Which Pa = 10%................................................................................720
Table T11.22  M
 IL-STD-105E Scheme Limiting Quality (in Defects per Hundred
Units) for Which Pa = 5%..................................................................................721
Table T11.23  Scheme Measures of Performance for MIL-STD-105E, Code F..................722
Table T11.24  O
 perating Ratios for the MIL-STD-105E Scheme
(R = p.10/p.95, Calculated Using Poisson Distribution)..................................726
Table T12.1   M
 IL-STD-414 Table B-6—Values of T for Tightened Inspection:
Standard Deviation Method............................................................................727
Table T12.2   M
 IL-STD-414 Table B-7—Limits of Estimated Lot Percent Defective
for Reduced Inspection: Standard Deviation Method.................................729
Table T12.3   MIL-STD-414 Table A-1—AQL Conversion Table........................................731
Table T12.4   MIL-STD-414 Table A-2—Sample Size Code Letters...................................732
Table T12.5   M
 IL-STD-414 Table B-3—Master Table for Normal and Tightened
Inspection for Plans Based on Variability Unknown: Standard
Deviation Method (Double Specification Limit and Form 2,
Single-Specification Limit)...............................................................................733
Table T12.6   M
 IL-STD-414 Table B-4—Master Table for Reduced Inspection for
Plans Based on Variability Unknown: Standard Deviation Method
(Double Specification Limit and Form 2, Single-Specification Limit).......734
Table T12.7   M
 IL-STD-414 Table B-5—Table for Estimating the Lot Percent
Defective Using Standard Deviation Method...............................................735
Table T12.8   M
 ID-STD-414 Table B-1—Master Table for Normal and Tightened
Inspection for Plans Based on Variability Unknown: Standard
Deviation Method (Single-Specification Limit, Form 1)..............................749
Table T12.9   M
 IL-STD-414 Table B-2—Master Table for Reduced Inspection for
Plans Based on Variability Unknown: Standard Deviation Method
(Single-Specification Limit, Form 1)...............................................................750
Table T12.10  M
 IL-STD-414 TABLE B-8—Values of F for Maximum Standard
Deviation (MSD)................................................................................................751
Table T13.1   V
 alues of Plotting Positions (pi) to be Used in Plotting on Normal
Probability Paper for the No-Calc Procedure................................................752
Table T13.2   V
 alues of Maximum Estimated Percentage Defective Allowing
Acceptance of the Lot (p*)................................................................................753
Table T13.3   M
 atched Attributes Narrow Limit, Known (σ) and Unknown (s)
Standard Deviation Variables Plans for values of p1 and p2 with
α = .05, β = .10.....................................................................................................754
Table T13.4   Tightened Inspection Optimal Narrow Limit Plans for MIL-STD-105E...758
Table T13.5   Normal Inspection Optimal Narrow Limit Plans for MIL-STD-105E....... 761
List of Tables xxxix

Table T13.6   Reduced Inspection Optimal Narrow Limit Plans for MIL-STD-105E..... 764
Table T13.7   M
 IL-STD-105E Scheme Probability of Acceptance (Pa) and Average
Sample Number (ASN) at AQL Using Narrow Limit Plans (Limit
Numbers for Switching to Reduced Inspection Not Used).........................767
Table T13.8   Joint Probabilities for Mixed Plans.................................................................769
Table T14.1   Values of x and y for Determining AOQL......................................................776
Table T16.1   Values of Y for Determining AOQL, for SkSP-2 Plans.................................777
Table T16.2   Unity Values for SkSP-2 and Matched Single-Sampling Plans...................779
Table T16.3   Poisson Unity Values for Constructing ChSP-1 Plans..................................780
Table T16.4   ChSP-1 Plans Indexed by AQL (p.95) and LTPD (p.10)....................................781
Table T16.5   ChSP-1 Plans Indexed by AQL (p.95) and AOQL...........................................782
Table T17.1   Parametric Values of Some TNT Plans...........................................................785
Table T17.2   Unity Values for the QSS System....................................................................786
Table T17.3   Unity Values for the QSS-2(n; cN, cT) System.................................................792
Table T17.4   Unity Values for the QSS-3(n; cN, cT) System.................................................793
Table T17.5   Unity Values for the QSS-4(n; cN, cT) System.................................................794
Table T17.6   Unity Values for the QSS-2(n; kn, c0) System.................................................795
Table T17.7   Unity Values for the QSS-3(n; kn, c0) System.................................................797
Table T17.8   Unity Values for the QSS-4(n; kn, c0) System.................................................799
Table T17.9   Unity Values for the QSS-1(n; kn, 0) System..................................................801
Table T17.10  Unity Values for the QSS-2(n; kn, 0) System..................................................802
Table T17.11  Unity Values for the QSS-3(n; kn, 0) System..................................................803
Table T17.12  Unity Values for the QSS-4(n; kn, 0) System..................................................804
Table T17.13  M
 IL-STD-1916 Table I—Code Letters (CL) for Entry
into the Sampling Tables..................................................................................805
Table T17.14  MIL-STD-1916 Table II—Attributes Sampling Plans...................................806
Table T17.15  MIL-STD-1916 Table III—Variables Sampling Plans....................................807
Table T17.16  MIL-STD-1916 Table IV—Continuous Sampling Plans...............................808
Table T17.17  Hα Values for Simplified Grand lot Sampling...............................................809
Table T18.1   H
 azard Values Corresponding to Probability Plotting Positions for
Censored Data.................................................................................................... 811
Table T18.2   H108 Table 2A–1—Life Test Sampling Plan Code Designation..................815
Table T18.3   H
 108 Table 2B–1—Master Table for Life Tests Terminated upon
Occurrence of Preassigned Number of Failures...........................................816
xl List of Tables

Table T18.4   H
 108 Table 2C–1 (b)—Master Table for Life Tests Terminated at
Preassigned Time: Testing Without Replacement (Values of T/α0
for α = 0.05)......................................................................................................... 817
Table T18.5     H
 108 Table 2C–2 (b)—Master Table for Life Tests Terminated at
Preassigned Time: Testing with Replacement (Values of T/α0 for α = 0.05).... 818
Table T18.6   H108 Table 2D–1 (b)—Master Table for Sequential Life Tests (α = 0.05).... 819
Table T18.7   H
 108 Table 2C–5—Master Table for Proportion Failing before
Specified Time: Life Test Sampling Plans for Specified α, β, and p1/p0..... 820
Table T18.8   TR3 Table 1—Table of Values for Percent Truncation, (t/μ) × 100............... 821
Table T18.9   T
 R7 Table 1A—100t/μ Ratios at the Acceptable Quality Level
(Normal Inspection) for the MIL–STD–105E Plans...................................... 822
Table T18.10  T
 R7 Table 1B—100t/μ Ratios at the Limiting Quality Level for the
MIL–STD–105E Plans: Consumer’s Risk = 0.10............................................. 823
Table T18.11   T
 R7 Table 1C—100t/μ Ratios at the Limiting Quality Level for the
MIL–STD–105E Plans: Consumer’s Risk = 0.05............................................ 826
1
Introduction

Dodge (1969b, p. 156) has indicated that in the early days of the development of m ­ ilitary
standards during World War II, a distinction became apparent between acceptance
­sampling plans, on one hand, and acceptance quality control, on the other. The former are
merely specific sampling plans, which, when instituted, prescribe conditions for accep-
tance or rejection of the immediate lot inspected. The latter may be compared to process
quality control, which utilizes various indicators (such as control charts) and strategies
(such as process capability studies) to maintain and improve existing levels of quality in a
production process. In like manner, acceptance quality control exploits various acceptance
sampling plans as tactical elements in overall strategies designed to achieve desired ends.
Such strategies utilize the elements of systems engineering, industrial psychology, and, of
course, statistics and probability theory, together with other diverse disciplines, to bring
pressures to bear to maintain and improve the quality levels of the submitted product. For
example, in the development of the Army Ordnance sampling tables in 1942, Dodge (1969b,
p. 156) points out that

basically the “acceptance quality control” system that was developed encompassed the
concept of protecting the consumer from getting unacceptably defective material and
encouraging the producer in the use of process quality control by varying the quan-
tity and severity of acceptance inspections in direct relation to the importance of the
characteristics inspected and in inverse relation to the goodness of the quality level as
indicated by those inspections.

The resulting tables utilize not just one sampling plan, but many in a scheme for quality
improvement.
This book stresses acceptance quality control in recognition of the importance of such
systems as a vital element in the control of quality. There is little control of quality in the
act of lot acceptance or rejection. While the utilization of sampling plans in assessing lot
quality is an important aspect of acceptance sampling, it is essentially short run in effect.
The long-run consequences of a well-designed system for lot acceptance can be more effec-
tive where applicable. Thus, an individual sampling plan has much effect of a long sniper,
while the sampling scheme can provide a fusillade in the battle for quality improvement.

Acceptance Quality Control

Individual sampling plans are used to protect against irregular degradation of levels of
quality in submitted lots below that considered permissible by the consumer. A good
­sampling plan will also protect the producer in the sense that lots produced at permissible
levels of quality will have a good chance to be accepted by the plan. In no sense, however, is
it possible to “inspect quality into the product.” In fact, it can be shown (Mood 1943) that if

1
2 Acceptance Sampling in Quality Control

a producer continues to submit to the consumer the product from a process with a ­constant
proportion defective, lot after lot, simple acceptance or rejection of the lots ­submitted will
not change the proportion defective the consumer will eventually receive. The consumer
will receive the same proportion defective as was in the original process.
This idea may be simply illustrated as follows. Suppose you are in the business of
­repackaging playing cards. You have an abundance of face cards (kings, queens, and
jacks) and so submit an order to the printer for 5000 cards having an equal selection of
­nonface cards. Any face cards, then, can be considered as defectives if they are found
in the shipment. The cards are supposed to come to you in packages of 50 resembling
­standard 52-card decks. Unknown to you, the printer has mixed up your order and is
­simply ­sending standard decks. Your sampling plan is to accept the deck if a sample of one
card is acceptable. The lot size is actually, of course, 52.
What will be the consequences? Nearly 12 of the 52 cards in a standard deck are face
cards, so the probability of finding a face card on one draw is 12/52 = 0.23, or 23%. This
means that in 100 decks examined there should be roughly 23 rejections. Suppose these
rejected decks are thrown into the fire, what will be the proportion of face cards in the
accepted material? Why 23%, of course, since all the decks were the same. Thus, the
­sampling plan had no effect on the quality of the material accepted while the process
­proportion defective remained constant. The proportion defective accepted is the same as
if no inspection had ever been performed.
Suppose, instead, the printer had become even more mixed up. The printer fills half
the order with ordinary playing cards and the other half with cards from pinochle decks.
Pinochle decks are composed of 48 cards, half of which (or 24) are face cards. The printer
ships 50 ordinary decks (2600 cards) and 50 pinochle decks (2400 cards). Inspection of
the 50 ordinary decks by the same plan will reject about 23%, or about 12 of them. The
remaining 38 will pass and be put into stock. Of the 50 pinochle decks, however, half will
be rejected and so 25 will go into stock.
Some calculation will show that, with no sampling (i.e., 100% lot acceptance), the stock
would consist of

(12 ´ 50 ) + ( 24 ´ 50 ) = 1800
face cards out of a total stock of 5000 cards or
1800
´ 100 = 36.0%
5000
face cards.
Using the sampling plan, simple and ineffective as it was, the stock would consist of

(12 ´ 38 ) + ( 24 ´ 25 ) = 1056
face cards out of a total stock of

( 52 ´ 38 ) + ( 48 ´ 25 ) = 3176
or
1056
´ 100 = 33.2%
3176

face cards.
Introduction 3

Thus, quality can be improved by the imposition of a sampling plan in the face
of ­f luctuation in proportion defective since it will tend to selectively screen out the
highly defective material relative to the better lots. Clearly, a larger improvement
could have been made in the previous example if a more discriminating sampling
plan had been used.
Now, consider the imposition of some rules beyond the single sampling plan itself.
Suppose the rejected decks are 100% inspected with any face cards found being replaced
with nonface cards. Then, in the last part of the example, the number of face cards in
stock would be 1056 out of 3176 as before, since they came from the accepted lots. But,
since the 36 rejected lots would have been replaced with perfect products, the stock
would be increased by 50 × 36 = 1800 cards to a level of 4976 cards. The stock would
now consist of

1056
´ 100 = 21.2%
4976

face cards. Here, we have a substantial improvement in the level of quality even when
using an extremely loose plan in the context of a sampling strategy—in this case what is
called a “rectification scheme.”
Finally, suppose if 100% complete inspection was instituted. It is generally conceded
that no screening operation is 100% effective and, in the real world, 100% inspection of
a large number of units may be only about 80% effective according to Juran (1999). If
this is the case, about 20% of the defective cards will be missed and the final stock will
contain

1800 ´ 0.20 = 360

defective cards, or a percent defective of

360
´ 100 = 7.2%
5000

at a cost of examining all 5000 cards. Even 100% inspection will not necessarily eliminate
defective items once they are produced.
Thus, it is that sampling strategies can be developed to attain far more protection than
the imposition of a simple sampling plan alone. What is required, of course, is a continu-
ing supply of lots from the same producer to allow the strategy to be effective. It is for
this reason that there are two basic approaches to acceptance quality control, depending
upon the nature of the lots to be inspected. A continuing supply of lots from the same
producer is most effectively treated by a sampling scheme. A single lot, unique in itself,
is treated by sampling plans designed for use with an “isolated lot.” This distinction is
fundamental to acceptance sampling, and even the basic probability distributions used
in the two cases are not the same. We speak of Type A sampling plans and probabilities
when they are to be used with a single lot and Type B when used in the context of a con-
tinuing series of lots produced from the same supplier’s process. Effective acceptance
quality control will utilize the schemes and plans of acceptance sampling to advantage
in either case.
4 Acceptance Sampling in Quality Control

Acceptance Control and Process Control

Acceptance sampling procedures are necessarily defensive measures, instituted as protec-
tive devices against the threat of deterioration in quality. As such, they should be set up
with the aim of discontinuance in favor of process control procedures as soon as possible.
Process quality control is that aspect of the quality system concerned with monitor-
ing and improving the production process by analysis of trends and signals of quality
problems or opportunities for the enhancement of quality. Its methods include various
types of control charts, experiment designs, response surface methodologies, evolutionary
operations, and other procedures including, on occasion, those of acceptance sampling.
Considering the following, these methods are an essential adjunct for effective acceptance
control:

1. Quality levels for selecting an appropriate sampling procedure should be deter-

mined from the control chart analysis to ascertain what minimum levels the pro-
ducer can reasonably and economically guarantee and what maximum levels can
be tolerated by the consumer’s process or will fulfill the consumer’s wants and
needs.
2. Acceptance sampling procedures should be set up to “self-destruct” after a reason-
able period in favor of the process controls on the quality characteristic in ques-
tion. Simultaneous use of acceptance quality control and process quality should
eventually lead to improvement in quality levels to the point that regular applica-
tion of acceptance sampling is no longer needed.

Thus, at the beginning and at the end of an acceptance sampling procedure, process qual-
ity control plays an important part.

Process Quality Control

With the invention of the control chart by Shewhart in 1924, process quality control had
gained its most valuable tool as well as its genesis. When samples are taken periodically
on a process, the average of the samples will tend to cluster about some overall average,
or process level, as long as the process is not affected or changed to some new average,
or level. Such changes in the process level may be intentional or completely inadvertent
and unexpected. The control chart is essentially a means for determining and signaling
when the process level has actually shifted to a new level in the face of chance variation in
sample results. Observations are collected in what are called “rational subgroups.” These
are taken to maximize the opportunity to show the source of a change in the process.
The Shewhart chart consists simply of three parallel lines, two outside lines, called
upper and lower “control limits,” and a “center line.” An example of such a chart is given in
Figure 1.1. It shows Shewhart’s first control chart and the memorandum that accompanied
it (Olmstead 1967, p. 72). In practice, sample results are plotted on the chart in sequence.
The center line reflects the average of the data, while the control limits are calculated to
have a high probability (usually 331:1 odds) of the sample data being contained between
them if the process is stable. It is, then, very unlikely (3 chances in 1000) that a point will
Introduction 5

Case 18013

WAS-724-5-16-24-FQ

MR. R. I. JONES:-

A few days ago, you mentioned some of the problems

connected with the development of an acceptable form of
inspection report which might be modified from time to time,
in order to give at a glance the greatest amount of accurate
information
The attached form of report is designed to indicate
whether or not the observed variations in the present of
defective apparatus of a given type are significant; that is,
to indicate whether or not the product is satisfactory. The
theory underlying the method of determining the significance
of the variations in the value of p is somewhat involved when
considered in such a form as to cover practically all types
of problems. I have already started the preparation of a
series of memoranda converting these points in detail. Should
it be found desirable, however, to make use of this form of
chart in any of the studies now being conducted within the
Inspection Department, it will be possible to indicate the
method to be followed in the particular examples.

W. A. SHEWHART.
Type apparatus
Enc.: Inspected for
Form of Report. Tolerance

Sept.
Aug.

Nov.
Mar.

Dec.
June

Oct.
May
Apr.
Feb.

July
Jan.

Number
manufactured
Number
inspected

Observed points This point indicates trouble

Upper
limit
Percent defective

60% P

Lower
limit

FIGURE 1.1 
The first Shewhart control chart. (Reprinted from Olmstead, P.S., Ind. Qual. Control, 24, 72, 1967. With permission.)

plot outside the limits when the process is running well. In this event, it can safely be left
alone. If the process level shifts, however, points will plot outside the limits, indicating the
need for corrective action on the process. In some cases, the points may plot outside the
limits in a favorable direction. This is an indication of the possibility for process improve-
ment when the source of the process change is detected. The control chart, then, provides
control of a process in the face of measurement and other sources of variation in the sense
that it shows when the process has significantly degraded or improved. This provides a
timely opportunity for assessment of the reasons for the change and hence for positive
action on the process.
6 Acceptance Sampling in Quality Control

A control chart, in control for 20 or 30 samples, that is, with all the points plotted within
the limits, is usually considered evidence of a stable process. The center line of such a chart
may be taken as a measure of the process average and used as an input to an acceptance
sampling plan. The charts out of control, that is, with points outside the limits, are an indi-
cation of lack of stability. Such charts can be interpreted to mean that the overall average
will not give a true representation of the data plotted on the chart.
Process control engineers and inspectors have at their disposal many auxiliary meth-
ods for the analysis of control charts, which are used for early detection of an out-of-
control condition before a point plots outside the limits. They are also used to isolate
evidence of the nature of the “assignable cause” of an out-of-control point. These meth-
ods, covered in the literature of statistical quality control, should be utilized by quali-
fied individuals to determine the fundamental causes of process shifts before evidence
from the control charts, which are out of control, is used in setting up an acceptance
sampling plan.
Instructions for constructing a control chart will be found in any basic text on quality
control (Burr 2005). An excellent discussion will be found in Wescott (1959) and Knowler
(1946), while factors for determining limits are given in the American Society for Testing
and Materials (2010). A few factors of the control chart are given in Appendix Table T1.1
for the convenience of the reader who is familiar with control chart construction and
interpretation.
Still another procedure in process quality control is the process optimization study. As
defined by Mentch (1980), such studies include

1.
Process performance check. A grab sample to estimate the characteristics of the pro-
cess at a given time.
2.
Process performance evaluation. An analysis of past data by control chart methods,
which is used to estimate process capability, limits the process performance if the
process were to remain in a constant state of control.
Process capability study. An ongoing, real-time study of the process including cor-
3.
rection of assignable causes to bring the process into a state of control so that esti-
mates of process capability can actually be realized or surpassed.
Process improvement study. Basic modification of the process through designed
4.
experiments and other means when existing process capability is not sufficient to
meet required specifications.

Process capability has been defined by Ekvall and Juran (1974, pp. 9–16) as follows:
“Process capability is the measured, inherent reproducibility of the product turned out by
a process.” It is of utmost importance in acceptance sampling, since in no event should the
requirements of a sampling procedure exceed the producer’s process capability. When this
happens, either a new supplier should be selected or the specifications should be changed.
In a similar manner, it is sometimes the case that a consumer’s process can tolerate varia-
tion in raw material beyond design requirements imposed by engineering. This provides
the opportunity for widening the specifications, with associated economic advantages. It
is important to determine what variables must be controlled through either acceptance
control or process control to achieve a desirable steady-state process level in the consum-
er’s process. Process optimization studies (on the real process as installed) can do this,
frequently with large cost savings in the process.
Introduction 7

The two important aspects of process quality control, control charts and process capa-
bility studies, have been discussed. It should be recognized that successful application
of the principles of acceptance control requires intimate knowledge of process control.
The reader is well advised to consult basic texts on the subject to take full advantage of
the synergism that can be achieved by simultaneous application of both forms of quality
control.

Background of Acceptance Quality Control

The development of the statistical science of acceptance sampling can be traced back
to the formation of the Inspection Engineering Department of Western Electric’s Bell
Telephone Laboratories in 1924. The department comprised H. F. Dodge, R. B. Miller, E.
G. D. Paterson, D. A. Quarles, and W. A. Shewhart. Later, H. G. Romig, P. S. Olmstead,
and M. N. Torrey became members of the group. It was directed initially by R. L. Jones
with G. D. Edwards becoming its second and long-term director. Applications to shop
operations at the Western Electric Hawthorne plant in Chicago were later formalized
in 1926 by the formation of the Western Electric Committee on Rating of Quality of
Telephone Products and a special committee on Inspection Statistics and Economy.
Early members of these committees included J. A. Davidson, A. B. Hazard, M. E. Berry,
E. D. Hall, J. M. Juran, C. A. Melsheimer, S. M. Osborne, C. W. Robbins, W. L. Robertson,
and W. Bartkey (consultant).
Out of this group and its lineage came the following early developments among others:

1924 The first control chart

1925–1926 Terminology of acceptance sampling (consumer’s risk, producer’s risk, probability of
acceptance, operating characteristic curves, lot tolerance percent defective [LTPD], average
total inspection, double sampling, Type A and Type B risks); LTPD sampling tables
1927 Average outgoing quality limit (AOQL) sampling tables; multiple sampling
1928 Demerit rating system

The 1930s saw applications of acceptance sampling within Western Electric and else-
where. A Joint Committee for the Development of Statistical Applications in Development
and Manufacturing was formed in 1930 by the American Society of Mechanical Engineers,
ASTM, American Institute of Electrical Engineers, American Statistical Association, and
American Mathematical Society with W. A. Shewhart as the chairperson. By the mid-
1930s, Pearson (1935) had developed British Standards Institution Standard Number 600,
Application of Statistical Methods to Industrial Standardization and Quality Control, which
helped incite interest in England. Moreover, in England, Jennett and Welch (1939) pub-
lished their paper on variables plans entitled, The Control of Proportion Defective as Judged by
a Single Quality Characteristic Varying on a Continuous Scale. Meanwhile, in the same year in
the United States, Romig (1939) submitted his doctoral dissertation to Columbia University
on “Allowable Average in Sampling Inspection,” presenting variables sampling plans
along the lines of the Dodge–Romig tables, which had been in use in Western Electric for
some time.
8 Acceptance Sampling in Quality Control

The early 1940s saw the publication of the Dodge and Romig (1941) Sampling Inspection
Tables, which provided plans based on fixed consumer risk (LTPD protection) and also
plans for rectification (AOQL protection), which guaranteed stated protection after 100%
inspection of rejected lots.
With the war, quality control and, particularly, acceptance sampling came of age. This
included the development by the Army’s Office of the Chief of Ordnance (1942) of Standard
Inspection Procedures of which the Ordnance sampling tables, using a sampling system
based on a designated acceptable quality level (AQL), were a part. The development of the
system was largely due to G. D. Edwards, H. F. Dodge, and G. R. Gause, with the assistance
of H. G. Romig and M. N. Torrey. This work later developed into the Army Service Forces
(ASF) tables of 1944 (U.S. Department of the Army, 1944).
In this period, Dodge (1943) developed an acceptance sampling plan that would perform
rectification inspection on a continuous sequence of product guaranteeing the consumer
protection in terms of the maximum average quality the consumer would receive (AOQL
protection). Furthermore, Wald (1943) put forward his new theory of sequential sampling
as a member of the Statistical Research Group, Columbia University (1945), which later
published applications of Wald’s work. This group was responsible for some outstand-
ing contributions during the war. Its senior scientific staff consisted of K. J. Arnold, R.
F. Bennett, J. H. Bigelow, A. H. Bowker, C. Eisenhart, H. A. Freeman, M. Friedman, M.
A. Girshick, M. W. Hastay, H. Hotelling, E. Paulson, L. J. Savage, J. Stigler, A. Wald, W.
A. Wallis, and J. Wolfowitz. Their output consisted of advancements in variables and
attributes sampling in addition to sequential analysis. Some of these are documented in
the Statistical Research Group, Columbia University (1947) under the title Techniques of
Statistical Analysis. They were active in theoretical developments in process quality con-
trol, design of experiments, and other areas of industrial and applied statistics as well.
Out of the work of the Statistical Research Group came a manual on sampling inspection
prepared for the U.S. Navy Office of Procurement and Material. Like the Army Ordnance
tables, it was a sampling system also based on the specification of an AQL and later
published by the Statistical Research Group, Columbia University (1948) under the title
Sampling Inspection. In 1949, the manual became the basis for the Defense Department’s
nonmandatory Joint Army–Navy Standard JAN-105; however, a committee of military
quality control specialists had to be formed to reach a compromise between JAN-105
and the ASF tables. This resulted in MIL-STD-105A issued in 1950 and was subsequently
revised as 105B, 105C, 105D, and 105E. The Statistical Research Group had considered the
development of a set of variables plans to match the AQL attributes system it had set forth
in the Navy manual. However, the group was disbanded on September 30, 1945, before it
was possible to construct such tables. Fortunately, the Office of Naval Research supported
preparation of such a work at Stanford University resulting in the book by Bowker and
Goode (1952), which was a milestone in the development of variables sampling plans. The
work of the Statistical Research Group has been documented by Wallis (1980).
Work in the area of acceptance sampling did not end with World War II. Many, if not
most, of the procedures presented in this book were developed later. This brief history,
however, has been presented to place the rest of the book in context so that each method
discussed can, in some sense, be traced to its natural origins. More detailed accounts of the
history and development of acceptance sampling will be found in Dodge (1969a–c, 1970a)
and in a series of papers published by the American Statistical Association (1950) under
the title Acceptance Sampling.
Introduction 9

Top 10 Reasons for Acceptance Sampling

While much has been written about the need for process control as a means of reducing
the dependency on acceptance sampling, the reality is that sampling will never go away.
Here are the top 10 reasons why acceptance sampling is still necessary:

1.
Tests are destructive, necessitating sampling. It is obviously counterproductive to
use 100% sampling with a destructive test. While all the defective material might
be eliminated, all the good material would be eliminated as well, leaving noth-
ing to sell.
2.
Process not in control, necessitating sampling to evaluate the product. An out-of-control
condition implies erratic behavior that cannot be predicted. Therefore, to evalu-
ate the product it is necessary to take a random sample of total production after
the fact.
3.
100% sampling is inefficient; 0% is risky. The efficiency of 100% inspection has been
estimated at around 80% in screening the product. No inspection provides any
assurance. Under sampling, rejected lots rather than individual defective pieces
are returned, getting management’s attention.
4.
Special causes may occur after process inspection. Process control ends when
the control chart is plotted, but the product moves on and is affected by sub-
sequent causes on its way to the customer. Sampling the final or incoming
product provides assurance against problems occurring after the process is
completed.
5.
Need for assurance while instituting process control. The process must operate for
some time to implement control charts and achieve control. The product pro-
duced in this period of unknown control must be evaluated. Sampling is a way
to evaluate this product and provide information useful in the start-up of pro-
cess control.
6.
Rational subgroups for process control may not reflect outgoing quality. Rational sub-
groups are set up to indicate stability of the process (or lack thereof), not for
evaluating the totality of the product produced. Random sampling of the product
provides an accurate representation of the population sampled.
7.
Deliberate submission of defective material. A real-world experience has shown that
pressures of production or profit may lead to fraud. Sampling can help prevent
and detect this.
8.
Process control may be impractical because of cost or lack of sophistication of personnel. It
is sometimes not cost-effective to institute process control, yet the product needs
to be evaluated. Sampling is also easier to implement.
9.
100% inspection does not promote process/product improvement. Often 100% inspection
is used as an excuse for not evaluating and controlling the underlying process.
Sampling with feedback of information often leads to process improvement.
10.
Customer mandates sampling plan. Customers may insist on mandatory sampling
procedures, which must be met.
10 Acceptance Sampling in Quality Control

Problems
1. Distinguish acceptance sampling from acceptance control.
2. Explain why installation of a sampling plan is futile if the level of quality is poor but
stable and cannot be improved.
3. Distinguish between Type A and Type B sampling plans.
4. Distinguish process quality control from acceptance quality control. How is process
quality control used in acceptance sampling?
5. What are the odds of an incorrect signal of a process change on a conventional
Shewhart chart?
6. What are the four constituents of a process optimization study?
7. Define process capability.
8. What was one of G.D. Edward’s principal contributions to quality control?
9. Which came first, the control chart or AOQL sampling plans? Where were they
developed?
10. Who invented continuous sampling plans? When?

References
American Society for Testing and Materials, 2010, Manual on Presentation of Data and Control Chart
Analysis, 8th ed., ASTM, Special Technical Publication (STP 15D), West Conshohocken, PA.
American Statistical Association, 1950, Acceptance Sampling—A Symposium, American Statistical
Association, Washington, DC.
Bowker, A. H. and H. P. Goode, 1952, Sampling Inspection by Variables, McGraw-Hill, New York.
Burr, J. T., 2005, Elementary Statistical Quality Control, 2nd ed., Marcel Dekker, New York.
Dodge, H. F., 1943, A sampling plan for continuous production, Annals of Mathematical Statistics,
14(3): 264–279.
Dodge, H. F., 1969a–c; 1970a, Notes on the evolution of acceptance sampling plans, Journal of Quality
Technology, Part I, 1(2): 77–88; Part II, 1(3): 155–162; Part III, 1(4): 225–232; Part IV, 2(1): 1–8.
Dodge, H. F. and H. G. Romig, 1941, Single sampling and double sampling inspection tables, The Bell
System Technical Journal, 20(1): 1–61.
Ekvall, D. N. and J. M. Juran, 1974, Manufacturing Planning, Quality Control Handbook, 3rd ed.,
McGraw-Hill, New York.
Jennett, W. J. and B. L. Welch, 1939, The control of proportion defective as judged by a single qual-
ity characteristic varying on a continuous scale, Supplement to the Journal of the Royal Statistical
Society, 6: 80–88.
Juran, J. M., Ed., 1999, Quality Control Handbook, 5th ed., McGraw-Hill, New York.
Knowler, L. A., 1946, Fundamentals of quality control, Industrial Quality Control, 3(1): 7–18.
Mentch, C. C., 1980, Manufacturing process quality optimization studies, Journal of Quality
Technology, 12(3): 119–129.
Mood, A. M., 1943, On the dependence of sampling inspection plans upon population distributions,
Annals of Mathematical Statistics, 14: 415–425.
Olmstead, P. S., 1967, Our debt to Walter Shewhart, Industrial Quality Control, 24(2): 72–73.
Introduction 11

Pearson, E. S., 1935, The Application of Statistical Methods to Industrial Standardization and Quality
Control, British Standards Institution, London, U.K., British Standard 600.
Romig, H. G., 1939, Allowable average in sampling inspection, PhD dissertation, Columbia
University, New York.
Statistical Research Group, Columbia University, 1945, Sequential Analysis of Statistical Data:
Applications, Columbia University Press, New York.
Statistical Research Group, Columbia University, 1947, Techniques of Statistical Analysis, McGraw-Hill,
New York.
Statistical Research Group, Columbia University, 1948, Sampling Inspection, McGraw-Hill, New York.
United States Department of the Army, 1944, Standard Inspection Procedures, Quality Control, Army
Service Forces, Office of the Chief of Ordnance, Washington, DC.
Wald, A., 1943, Sequential analysis of statistical data: Theory, report submitted by the Statistical
Research Group, Columbia University, to the Applied Mathematics Panel, National Defense
Research Committee, New York.
Wallis, W. A., 1980, The statistical research group, 1942–1945, Journal of the American Statistical
Association, 75(370): 320–330.
Wescott, M. E., 1959, Fundamental control techniques, Rubber World, Part I: 252–262; Part IIa: 717–722;
Part IIb: 869–872.
2
Probability and the Operating Characteristic Curve

Undoubtedly the most important single working tool in acceptance quality control is
probability theory itself. This does not mean that good quality engineers have to be accom-
plished probabilists or erudite mathematical statisticians. They must be aware, however,
of the practical aspects of probability and how to apply its principles to the problem at
hand. This is because most information in quality control is generated in the form of sam-
ples from larger, sometimes essentially infinite, populations. It is vital that the quality
­engineers have some background in probability theory. Only the most basic elements are
presented here.

Probability
It is important to note that the term probability has come to mean different things to different
people. In fact, these differences are recognized in defining probability, for there is not just
one but at least three important definitions of the term. Each definition gives insight into the
nature of probability itself. Two of them are objectivistic in the sense that they are subject to
verification, while the third is personalistic and refers to the degree of belief of an individual.

Classical Definition
“If there be a number of events of which one must happen and all are equally likely, and
if any one of a (smaller) number of these events will produce a certain result which cannot
otherwise happen, the probability of this result is expressed by the ratio of this smaller
number to the whole number of events” (Whitworth 1965, rule IV). Here probability is
defined as the ratio of favorable to total possible equally likely and mutually exclusive cases.
Example: There are 52 cards in a deck of which 4 are aces. If the cards are shuffled so that
they are equally likely to be drawn, the probability of obtaining an ace is 4/52 = 1/13.
This is the definition of probability, which is familiar from high school mathematics.

Empirical Definition
“The limiting value of the relative frequency of a given attribute, assumed to be indepen-
dent of any place selection, will be called ‘the probability of that attribute…’.” (von Mises
1957, p. 29). Thus, probability is regarded as the ratio of successes to the total number of
trials in the long run.

Example: In determining if a penny was in fact a true coin, it was flipped 2000 times
resulting in 1010 heads. An estimate of the probability of heads for this coin is .505.

13
14 Acceptance Sampling in Quality Control

It would be expected that this probability would approach 1/2 as the sequence of
tosses was lengthened if the coin were true.
This is the sort of probability that is involved in saying that Casey has a .333 batting
average. It implies that the probability of a hit the next time at bat is approximately 1/3.

Subjective Definition
“Probability measures the confidence that a particular individual has in the truth of a par-
ticular proposition, for example, the proposition that it will rain tomorrow” (Savage 1972).
Thus, probability may be thought of as a degree of belief on the part of an individual, not
necessarily the same from one person to another.
Example: There is a high probability of intelligent life elsewhere in the universe. Here
we have neither counted the occurrences and nonoccurrences of life in a number
of universes nor sampled universes to build up a ratio of trials. This statement
implies a degree of belief on the part of an individual, which may differ consider-
ably from one individual to another.
These definitions have immediate applications in acceptance quality control. Classical
probability calculations are involved in the determination of the probability of acceptance of
a lot of finite size, where all the possibilities can be enumerated and samples taken therefrom.
Empirical probabilities are used when sampling from a process running in a state of statistical
control. Here, the process could conceivably produce an uncountable number of units so that
the only way to get at the probability of a defective unit is in the empirical sense. Subjective
probabilities have been used in the evaluation of sampling plans, particularly under cost con-
straints. They reflect the judgment of an individual or a group as to the probabilities involved.
While sampling plans have been derived, which incorporate subjective probabilities, they
appear to be difficult to apply in an adversary relationship unless the producer and the con-
sumer can be expected to agree on the specific subjective elements involved.
There are many sources for information on probability and its definition. Some interest-
ing references of historic value are Whitworth (1965) on classical probability, von Mises
(1957) on empirical probability, and the Savage (1972) on subjective probability. Since the
classical and empirical definitions of probability are objectivistic and can be shown to
agree in the long run, and since the empirical definition is more general, the empirical defi-
nition of probability will be used here unless otherwise stated or implied. When subjective
probabilities are employed, their nature will be specifically pointed out.

Random Samples and Random Numbers

Random samples are those in which every item in the lot or population sampled has an
equal chance to be drawn. Such samples may be taken with or without replacement. That
is, items may be returned to the population once drawn, or they may be withheld. If they
are withheld, the probability of drawing a particular item from a finite population changes
from trial to trial. Whereas, if the items are replaced, or if the population is uncountably
large, the probability of drawing a particular item will not change from trial to trial. In any
event, every item should have an equal opportunity for selection on a given trial, whether
the probabilities change from trial to trial or not.
Probability and the Operating Characteristic Curve 15

This may be illustrated with a deck of cards. There are 52 cards, one of which is the ace
of spades. Sampling without replacement, the probability of drawing the ace of spades on
the first draw is 1 out of 52, while on the second draw it is 1 out of the 51 cards that remain,
assuming it was not drawn on the first trial. If the cards were replaced as drawn, the prob-
ability would be 1 out of 52 on any draw since there would always be 52 cards in the deck.
Note that if the population is very large, the change in probability when samples are
not replaced will be very small and will remain essentially the same from trial to trial.
In a raffle of 100,000 tickets, the chances of being drawn on the first trial are 1 in 100,000
and on the second trial are 1 in 99,999. Essentially, it is 0.00001 in each case. Few raffles are
conducted in which a winning ticket is replaced for subsequent draws.
At the core of random sampling is the concept of equal opportunity for each item in the
population sampled to be drawn on any trial. Sometimes special sampling structures are
used such as stratified sampling in which the population is segmented and samples are
taken from the segments. Formulas exist for the estimation of population characteristics
from such samples. In any event, equal opportunity should be provided within a segment
for items to be selected.
To guarantee randomness of selection, tables of random numbers have been prepared.
These numbers have been set up to mimic the output of a truly random process. They are
intended to occur with equal frequency but in a random order. Appendix Table T2.1 is one
such table. To use the random number tables, consider the following:

1. Number the items in the population.

2. Specify a fixed pattern for the selection of the random numbers (e.g., right to left,
bottom to top, every third on a diagonal).
3. Choose an arbitrary starting place and select as many random numbers as needed
for the sample.
4. Choose as a sample those items with numbers corresponding to the random num-
bers selected.

The resulting sample will be truly representative in the sense that every item in the popu-
lation will have had an essentially equal chance to be selected.
Sometimes it is impractical or impossible to number all the items in a population. In
such cases, the sample should be taken with the principle of random sampling in mind to
obtain as good a sample as possible. Avoid bias, and avoid examining the samples before
they are selected. Avoid sampling only from the most convenient location (the top of the
container, the spigot at the bottom, etc.). In one sampling situation, an inspector was sent
to the producer’s plant to sample the product as a boxcar was being loaded, since it was
impossible to obtain a random sample thereafter. Such strategies as these can help provide
randomness as much as the random sampling tables themselves.

Counting Possibilities
Evaluation of the probability of an event under the classical definition involves count-
ing the number of possibilities favorable to the event and forming the ratio of that num-
ber to the total of equally likely possibilities. The possibilities must be such that they
16 Acceptance Sampling in Quality Control

cannot occur together on a single draw, that is, they must be mutually exclusive. There are
three important aids in making counts of this type: permutations, combinations, and tree
diagrams.
Suppose a lot of three items, each identified by a serial number, is received, two of
which are good. The sampling plan to be employed is to sample two items and accept
the lot if no defectives are obtained. Reject if one or more defectives are found. Thus, the
sampling plan is n = 2 and c = 0, where n is the sample size and c represents the accep-
tance number or maximum number of defectives allowed in the sample for acceptance
of the lot.
If the items are removed from the shipping container one at a time, we may ask in
how many different orders (permutations) the three items can be removed from the box.
Suppose the serial numbers are the same except for the last digit, which is 5, 7, and 8,
respectively. Enumerating the orders, we have

578 875
758 857
785 587

The formula for the number of permutations of n things taken n at a time is

Pnn = n ! = n ( n - 1) ( n - 2 )1

where n!, or n factorial, is the symbol for multiplications of the number n by all the succes-
sively smaller integers down to one. Thus,

1! = 1
2 ! = 2 ( 1) = 2
3 ! = 3 ( 2 )( 1) = 6
4 ! = 4 ( 3 )( 2 )( 1) = 24

and so on. It is important to note that we define

0 ! = 1

In the example, we want the number of permutations of 3 things taken 3 at a time or

P33 = 3 ! = 3 ( 2 )( 1) = 6

which agrees with the enumeration.

In how many orders can we select the two items for our sample? Enumerating again:

57 87
75 85
78 58
Probability and the Operating Characteristic Curve 17

The formula for the number of permutations of n things taken r at a time is

n!
Prn =
( n - r )!

Clearly, the previous formula for Pnn is a special case of this formula. To determine the
number of permutations of three objects taken two at a time, we have

3! 3 ! 3 ( 2 )( 1)
P23 = = = =6
( 3 - 2 ) 1!
! 1

This makes sense and agrees with the previous result since the last item drawn is com-
pletely determined by the previous two items drawn and so does not contribute to the
number of possible orders (permutations).
Now, let us ask how many possible orders are there if some of the items are indistin-
guishable one from the other. For example, disregarding the serial numbers, we have one
defective item and two good ones. The good items are indistinguishable from each other
and we may ask in how many orders we can draw one defective and two good items. The
answer may be found in the formula for the number of permutations of n things, r of which
are alike (good) and (n − r) are alike (bad).

n!
Prn,( n - r ) =
r ! ( n - r ) !

and for example, the answer is

3! 3 ( 2 )( 1)
P23,( 3 - 2) = P23,1 = = =3
2 ! 1! 2 ( 1)( 1)

Enumerating them, we have

B G G
G B G
G G B

The reader may notice the similarity of the following formula

n!
Prn,( n - r ) =
r ! ( n - r ) !

and the classic formula for the number of combinations (groups) that can be made from n
things taken r at a time. The formula is

n!
Crn =
r !( n - r )!
18 Acceptance Sampling in Quality Control

and shows how many distinct groups of size r can be formed from n distinguishable
objects. If we phrase the question, “in how many ways can we select two objects (to be the
good ones) out of three,” we have

Good Bad

Group 1 57 8
75 8
Group 2 78 5
87 5
Group 3 85 7
58 7

or

3! 3!
C23 = = =3
2 ! ( 3 - 2 ) ! 2 ! 1!

Thus, we see

Prn,( n - r ) = Crn

In general, the combinatorial formula may be used to determine the number of groupings
of various kinds. For example, the number of ways (groups) to select 4 cards from a deck
of 52 to form hands of 4 cards (where the order is not important) is

13 17 25 1
52 ! 52 × 51 × 50 × 49 × 48 !
C452 = = 270, 725
4 ! 48 ! 4 × 3 × 2 × 1× 48 !

Using the classical definition of probability, then the probability of getting a hand contain-
ing all four aces is

number of four ace hands 1

P ( 4 aces ) = =
number of four card hands 270, 725

Here we have counted groups where the order in the group is not important.
In the same way, probabilities can be calculated for use in evaluating acceptance sam-
pling plans. The plan given in the earlier example was sample size 2; accept when there are
no defectives in the sample. That is, n = 2 and c = 0. To evaluate the probability of accep-
tance when there is one defective in the lot of N = 3, we would proceed as follows:

1. To obtain probability of acceptance, we must count the number of samples in

which we would obtain 0 defectives in a sample of 2.
2. The probability is the quantity obtained in step 1 divided by the total number of
samples of 2 that could possibly be obtained.
Probability and the Operating Characteristic Curve 19

Start

First draw B G G

Second draw G G B G B G

Acceptance decision R R R A R A

FIGURE 2.1
Tree diagram.

Then,

1. To obtain samples of 2 having no defectives, we would have to select both items

from the two items, which are good. The number of such samples is C22 = 1.
2. There are C23 = 3 different unordered samples. So the probability of accepting Pa
with this sampling plan is

C22 1
Pa = =
C23 3

The third tool in counting possibilities in simple cases such as this is the tree diagram.
Figure 2.1 shows such a diagram for this example, for the acceptance (A) and rejection (R)
of samples of good (G) and bad (B) pieces. Each branch of the tree going downward shows
a given sample permutation. We see that 1/3 of these permutations lead to lot acceptance.
Counting the permutations, we have

3!
P23 = =6
1!

possible samples, of which

2!
P22 = =2
0!

lead to acceptance. Then, the probability of acceptance is

P22 2 1
Pa = = =
P23 6 3

which shows that the probability of acceptance can be obtained by using either permuta-
tions or combinations.

Probability Calculus
There are certain rules for manipulating probabilities that suffice for many of the elemen-
tary calculations needed in acceptance control theory. These are based on the recognition
of two kinds of events.
20 Acceptance Sampling in Quality Control

Mutually exclusive events: Two events are mutually exclusive if, on a single trial, the
occurrence of one of the events precludes the occurrence of the other.
Independent events: Two events are stochastically independent if their occurrence on a
trial does not change the probability of occurrence of the other on that trial.

Thus, the events head and tail are mutually exclusive in a single trial of flipping a coin.
They are also not independent events since the occurrence of either on a trial drives the
probability of occurrence of the other on that trial to zero.
In contrast, the events ace and heart are not mutually exclusive in drawing cards since
they can occur together in the ace of hearts. Further, they are also independent since the
probability of drawing an ace is 4/52 = 1/13. If you know that a heart was drawn, the prob-
ability of the card being also an ace is still 1/13. Note that the events face card and queen
are not independent. The probability of drawing a queen is 4/52 = 1/13; however, if you
know a face card was drawn, the probability of that card being a queen is now 4/12 = 1/3.
Trials are sometimes spoken of as being independent. This means the sampling situa-
tion is such that the probabilities of the events being investigated do not change from trial
to trial. Flips of a coin are such as this in that the odds remain 50:50 from trial to trial.
However, if cards are drawn from a deck and not replaced, the trials are dependent. Thus,
the probability of a queen of hearts is 1/52 on the first draw from a deck, but it increases to
1/51 on the second draw assuming it was not drawn on the first.
Kolmogorov (1956) has developed the entire calculus of probabilities from a few simple
axioms. Crudely stated and somewhat condensed, they are as follows:

1. The probability of an event, E, is always positive or zero, never negative: P(E) ≥ 0.

2. The sum of the probabilities of events in the universe U, or population to which
E belongs, is one: P(U) = 1.
3. If events A and B are mutually exclusive, the probability of A or B occurring is
P(A or B) = P(A) + P(B).
From the axioms, the following consequences can be obtained:
4. The probability of an event must be less than or equal to one, never greater than
one: P(E) ≤ 1.
5. The probability of the null set (no event occurring) is zero: P(no event) = 0.
6. The probability of an event not occurring is the complement of the probability of
the event: P(not E) = 1 – P(E).

The most useful rules in dealing with probabilities are the so-called
General rule of addition. This shows the probability of A or B occurring on a single trial:

P ( A or B ) = P ( A ) + P ( B ) - P ( A and B )

Clearly, if A and B are mutually exclusive, the term P(A and B) = 0 and we have the so-called
special rule of addition

P ( A or B ) = P ( A ) + P ( B )

for A and B mutually exclusive.

Probability and the Operating Characteristic Curve 21

General rule of multiplication. This shows the probability of A and B both occurring on
a single trial where P(B|A) is the conditional probability of B given A is known to
have occurred:

P ( A and B ) = P ( A ) P ( B A )

= P ( B )P ( A B )

Clearly, if A and B are independent, the factor P(B|A) = P(B) since the probability of B is
unchanged even if we know A has occurred (similar to P(A|B)). We then have the so-called
special rule of multiplication:

P ( A and B ) = P ( A ) P ( B ) , A and B independent

This is sometimes used as a test for the independence of A and B since if the relationship
holds, A and B are independent.
These rules can be generalized to any number of events. The special rules become

P ( A or B or C or D ) = P ( A ) + P ( B ) + P ( C ) + P ( D ) , A , B , C, D mutually exclusive

P ( A and B and C and D ) = P ( A ) P ( B ) P ( C ) P ( D ) , A , B , C, D independent

and so on. These are especially useful since they can be employed to calculate probabilities
over several independent trials. The general rule for addition is

P ( A or B or C or D ) = P ( A ) + P ( B ) + P ( C ) + P ( D )
- P ( AB ) - P ( AC ) - P ( AD ) - P ( BC ) - P ( BD ) - P ( CD )
+ P ( ABC ) + P ( ABD ) + P ( ACD ) + P ( BCD ) - P ( ABCD )

alternating additions and subtractions of each higher level of joint probability, while that
for multiplication becomes

P ( A and B and C and D ) = P ( A ) P ( B A ) P ( C AB ) P ( D ABC )

when there are four events. Each probability multiplied is conditional on those that went
before.
These rules may be illustrated using the example given earlier involving the computa-
tion of the probability of acceptance Pa of a lot consisting of 3 units when one of them is
defective and the sampling plan is n = 2, c = 0. Acceptance will occur only when both items
in the sample are good. If we assume random samples are drawn without replacement, the
events will be dependent from trial to trial. We need the probability of a good item on the
first draw and a good item on the second draw.
Let A = {event good on first draw} and B = (event good on second draw}; then

2 1
P (A) = P (B A) =
3 2
22 Acceptance Sampling in Quality Control

since there are only two pieces left on the second draw. Applying the general rule of
multiplication,
2æ1ö 1
Pa = P ( A ) P ( B A ) = ç ÷=
3 è 2 ø 3

which agrees with the result of the previous section:

C22 1
Pa = =
C23 3

Now, what if the items were put back into the lot after inspection and the next sample
drawn? This is a highly unusual procedure in practice but serves as a model for some of
the probability distributions developed later. It simulates an infinite lot 1/3 defective, since
using this method of inspection the lot would never be depleted. Under these conditions,
the special rule of multiplication could be employed since the events would be indepen-
dent of each other from trial to trial. We obtain
2æ2ö 4
Pa = P ( A ) P ( B ) = ç ÷=
3 è 3 ø 9

This makes sense since the previous method depleted the lot and made it more likely to
obtain the defective unit on the second draw.
Further, suppose two such lots are inspected using the procedure of sampling without
replacement. What is the probability that at least one will be accepted? That is, what is the
probability that one or the other will be passed? Here, let C = {event first lot is passed} and
D = {event second lot is passed}, then the probability both lots are passed is

1æ1ö 1
P ( C and D ) = ç ÷=
3 è 3 ø 9

using the special rule of multiplication since they are inspected independently. Then, the
probability of at least one passing is

P ( C or D ) = P ( C ) + P ( D ) - P ( C and D )
1 1 1 5
= + - =
3 3 9 9
The probability of not having at least one lot pass is
5 4
P ( both fail ) = 1 - P ( C or D ) = 1 - =
9 9

which could have been calculated using the special rule of multiplication as

P ( both fail ) = éë1 - P ( C ) ùû éë1 - P ( D ) ùû

2æ2ö
= ç ÷
3è3ø
4
=
9
Probability and the Operating Characteristic Curve 23

Finally, suppose there are five inspectors, V, W, X, Y, Z, each with the same probability of
selection. The lot is to be inspected. What is the probability that the inspector chosen is X,
Y, or Z? Since in this case the use of the inspectors is mutually exclusive, the special rule
of addition may be used:

P ( X or Y or Z ) = P ( X ) + P ( Y ) + P ( Z )
1 1 1
= + +
5 5 5
3
=
5

These are a few of the tools of probability theory. Fortunately, they have been put to use
by theorists in the design of the methods of acceptance quality control to develop proce-
dures that do not require extensive knowledge of the subject for application. These meth-
ods are presented here in subsequent chapters. Nevertheless, to gain a true appreciation
for the subtleties of acceptance sampling, a sound background in probability theory is
invaluable.

Operating Characteristic Curve

A fundamental use of probability with regard to acceptance sampling comes in describing
the chances of a lot passing sampling inspection if it is composed of a given proportion
defective. The very simplest sampling plan is, of course, as follows:

1. Sample one piece from the lot.

2. If the sampled piece is good, accept the lot.
3. If the sampled piece is defective, reject the lot.

This plan is said to have a sample size n of one and an acceptance number of zero since
the sample must contain zero defectives for lot acceptance to occur; otherwise, the
lot will be rejected. That is, n = 1, c = 0. Now, if the lot were perfect, it would have no
chance of rejection since the sample would never contain a defective piece. Similarly,
if the lot were completely bad, there would be no acceptances since the sample piece
would always be defective. But what if the lot were mixed defective and good? This is
where probability enters in. Suppose one-half of the lot was defective, then the chance
of drawing out a defective piece from the lot would be 50:50 and we would have 50%
probability of acceptance. But it might be one-quarter defective leading to a 75% chance
for acceptance, since there are three-quarters good pieces in the lot. Or again, the lot
might be three-quarters defective leading to a 25% chance of finding a good piece.
Since the lot might be any of a multitude of possible proportions defective from 0 to 1,
how can we describe the behavior of this simple sampling plan? The answer lies in
the operating characteristic (OC) curve that plots the probability of acceptance against
possible values of proportion defective. The curve for this particular plan is shown in
Figure 2.2.
24 Acceptance Sampling in Quality Control

1.0

0.9

0.8

0.7

0.6

0.5
Pa

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1.0
p

FIGURE 2.2
OC curve, n = 1, c = 0.

We see that for any proportion defective p, the probability of acceptance Pa is just the
complement of p, that is,

Pa = 1 - p

This is only true of the plan n = 1, c = 0. Thus, the OC curve stands as a unique represen-
tation of the performance of the plan against possible alternative proportions defective.
A given lot can have only one proportion defective associated with it. But we see from
the curve that lots that have a proportion defective greater than 0.75 have less than a 25%
chance to be accepted and those lots with less than 0.25 defective pieces will have greater
than a 75% chance of pass. The OC curve gives at a glance a characterization of the poten-
tial performance of the plan, telling how the plan will perform for any submitted fraction
defective.
Now consider the plan n = 5, c = 0. The OC curve can be easily constructed using the
rules for manipulation of probabilities given earlier. First, however, let us assume we are
sampling from a very large lot or, better yet, from the producer’s process so the prob-
abilities will remain essentially independent from trial to trial. Note that the probability of
acceptance Pa for any proportion defective p can be computed as

Pa = ( 1 - p ) ( 1 - p ) ( 1 - p ) ( 1 - p ) ( 1 - p )
= (1 - p )
5

Probability and the Operating Characteristic Curve 25

since all the pieces must be good in the sample of 5 for lot acceptance. To plot the OC curve,
we compute Pa for various values of p
p (1 − p) Pa
.005 .995 .975
.01 .99 .951
.05 .95 .774
.10 .90 .590
.20 .80 .328
.30 .70 .168
.40 .60 .078
.50 .50 .031

and graph the result as in Figure 2.3.

We see from Figure 2.3 that if the producer can maintain a fraction defective less than .01
the product will be accepted 95% of the time or more by the plan. If the product submitted
is 13% defective, it will have a 50:50 chance of acceptance, while the product that is 37%
defective has only a 10% chance of acceptance by this plan. It is conventional to designate
proportions defective having a given probability of acceptance as probability points. Thus,
a fraction defective having a probability of γ is shown as pγ. Particular probability points
may be designated as follows:

Pa Term Abbreviation Probability Point

.95 Acceptable quality level AQL p.95

.50 Indifference quality IQ p.50
.10 Lot tolerance percent defective (10% limiting quality) LTPD (LQ [.10]) p.10

1.0

0.9

0.8

0.7

0.6

0.5
Pa

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5
p

FIGURE 2.3
OC curve, n = 5, c = 0.
26 Acceptance Sampling in Quality Control

Designation of these points gives a quick summary of plan performance. The term
acceptable quality level (AQL) is commonly used as the 95% point of probability of accep-
tance, although most definitions do not tie the term to a specific point on the OC curve
and simply associate it with a “high” probability of acceptance. The term is used here
as it was used by the Columbia Statistical Research Group in preparing the Navy (1946)
input to the JAN-STD-105 standard. Lot tolerance percent defective (LTPD) refers to the
10% probability point of the OC curve and is generally associated with percent defec-
tive. The advent of plans controlling other parameters of the distribution led to the term
limiting quality (LQ), usually preceded by the percentage point controlled. Thus, “10%
LQ” is the LTPD.
The OC curve is often viewed in the sense of an adversary relationship between the pro-
ducer and the consumer. The producer is primarily interested in insuring that good lots
are accepted, while the consumer wants to be reasonably sure that bad lots will be rejected.
In this sense, we may think of a producer’s quality level (PQL) and associated producer’s
risk α and a consumer’s quality level (CQL) with associated consumer’s risk β. Viewed
against the OC curve, the PQL and CQL appear as in Figure 2.4.
Plans are often designated and constructed in terms of these two points and the associ-
ated risks. As indicated earlier, the risks are often taken as α = .05 for the producer’s risk
and β = .10 for the consumer’s risk.
The OC curve sketches the performance of a plan for various possible proportions
defective. It is plotted using appropriate probability functions for the sampling situation
involved. The probability functions are simply formulas for the direct calculation of prob-
abilities, which have been developed using the appropriate probability theory.

1.0
α

0.9

0.8

0.7

0.6

0.5
Pa

0.4

0.3

0.2

0.1
β
0
PQL CQL
p

FIGURE 2.4
PQL and CQL.
Probability and the Operating Characteristic Curve 27

Software Applications
It becomes quickly apparent to the user of acceptance sampling plans that the tasks of
deriving and evaluating them involve some mathematics. Fortunately, the computer can
make an easy chore of such tasks. The development of published sampling plans, as will
be discussed in subsequent chapters, negated the need to deal with mathematics as each
set of plans contained a wide variety of choices for the user.
Today, acceptance sampling software is available for anyone to construct a plan meet-
ing the criteria of the user. Commercial programs, such as Minitab and Statgraphics, will
be discussed in this text with examples on their use pertaining to the chapter material
they follow. There are other lesser known programs that can be purchased from various
vendors on the Internet. A simple Google or Bing search through an Internet browser for
“acceptance sampling software” will bring up many of these vendors. A few of these pro-
grams will be mentioned along the way in this text where appropriate.
However, it should also be noted that a widely available spreadsheet program, such
as Microsoft Excel, can be used to create sampling plans and generate needed sampling
plans. One of the authors developed a set of Excel sampling plan templates for his accep-
tance sampling classes that he taught over many years. These templates enabled the stu-
dents to more effectively learn sampling plans and play “what if” and enabled them to
more quickly implement plans when back at work.

Excel
It is easy to demonstrate the effect of varying n or c while holding the other constant with
a simple Excel template (Comparison of Single Sampling Plans.xlsm from the CRC book web-
site). Note that this template has a .xlsm extension implying that you will need to enable its
macro to open the file (this file does not contain one). The first tab in the worksheet is called
OC Curve Calcs (Figure 2.5). The light gray–shaded cells in the template are input cells for
n and c. A darker gray–shaded cell is used to indicate which probability distribution is desired
(1 = Poisson, 2 = binomial). This feature is useful so that the user can see how the distributions
agree with each other. The user is urged not to change any of the formulas in any cells in the
template so as to maintain its integrity. If any formulas are changed, the results will be incor-
rect and a new copy of the template would need to be downloaded from the CRC book website.
Suppose we wish to illustrate the effect of changing the acceptance number c for a con-
stant sample size n. Assume that the sample size is n = 20 and we want to see the effect
of changing c from 0 to 1 to 2. We would enter 20 into cells B8, B14, and B20, and 0 into
B9, 1 into B15, and 2 into B21. Notice that the rejection number r (cells B10, B16, and B22) is
automatically set equal to one more than c. Set the probability distribution to the binomial
by entering a 2 into cell F4.
There are three other tabs in this worksheet. The second tab is the OC curve of all three
plans. The third and fourth tabs are the average outgoing quality (AOQ) curve (with aver-
age outgoing quality limit [AOQL]) and average total inspection curves, which will be
discussed in Chapter 4. For now, we will only discuss the OC curve (Figure 2.6).
As c gets larger, the producer risk gets smaller so that lots are more readily accepted at
a given process nonconforming level in the lot (x-axis). Note that the c = 0 plan, aka accept
on zero (AoZ) plan, carries a heavy price for the producer who will be rejecting many good
lots. AoZ plans are discussed in detail in Chapter 17. Also, as c gets larger, the consumer
risk gets larger because lots are more readily accepted at a given process nonconforming
 28

FIGURE 2.5
Excel template Comparison of Single Sampling Plans.xlsm.
Acceptance Sampling in Quality Control
Probability and the Operating Characteristic Curve 29

OC curve for comparing single sampling plans

100%

90%
n = 20, c = 0
80% n = 20, c = 1
Probability of accepting lot, Pa

70% n = 20, c = 2

60%

50%

40%

30%

20%

10%

0%
0% 5% 10% 15% 20% 25%
Percent nonconforming in lot, p

FIGURE 2.6
OC curve for comparison of sampling plans (n, c): (20,0), (20,1), and (20,2).

level in the lot. Note here that the c = 0 plan benefits the consumer who will be receiving
fewer bad lots.
Now, to illustrate the effect of varying n on the OC curve, we will set c = 1 and vary n
from 10 to 20 to 40. We will enter 1 into cells B9, B15, and B21, and 10 into B8, 20 into B14,
and 40 into B20. The rejection number should be 2 in cells B10, B16, and B22 based on
its formula of being one more than the acceptance number. Again, let’s use the binomial
­distribution and leave a 2 in cell F4.
Based on the OC curve on the second tab (Figure 2.7), we find that as n gets larger,
the producer risk gets larger so that lots are more readily rejected at a given process
­nonconforming level (x-axis). Furthermore, as n gets larger, the consumer risk gets smaller
because lots are more readily accepted at a given process nonconforming level.

Minitab
In 2006, the authors worked with Minitab to incorporate acceptance sampling into Version
15. While the capabilities produced in that version are virtually unchanged in the current
version (17), Minitab does a credible job in providing basic acceptance sampling function-
ality to design and evaluate attributes and variables single sampling plans. One of these
features is to compare user-defined sampling plans. This can be used to demonstrate the
effect of varying c with a constant n or varying n with a constant c as was done in the pre-
vious section with Excel.
Choose Stat > Quality Tools > Acceptance Sampling by Attributes from the menu. In
order to illustrate the effect of varying c with a constant n, complete the dialog box as
shown in Figure 2.8. The OC curves are shown in Figure 2.9 that agree with what Excel
generated in Figure 2.6. Similarly, we can show the effect of varying n with a constant c by
completing the dialog box as shown in Figure 2.10. The OC curves are shown in Figure 2.11
that agree with what Excel generated in Figure 2.7.
30 Acceptance Sampling in Quality Control

OC curve for comparing single sampling plans

100%

90%
n = 10, c = 1
80% n = 20, c = 1
Probability of accepting lot, Pa

70% n = 40, c = 1

60%

50%

40%

30%

20%

10%

0%
0% 5% 10% 15% 20% 25%
Percent nonconforming in lot, p

FIGURE 2.7
OC curve for comparison of sampling plans (n, c): (10,1), (20,1), and (40,1).

FIGURE 2.8
Minitab dialog for comparison of sampling plans (n, c): (20,0), (20,1), and (20,2).
Probability and the Operating Characteristic Curve 31

Operating characteristic (OC) curve

1.0
n c
20 0
20 1
0.8 20 2
Probability of acceptance

n sample size
c acceptance number
0.6

0.4

0.2

0.0

0 5 10 15 20
Lot percent defective

FIGURE 2.9
Minitab OC curve for comparison of sampling plans (n, c): (20,0), (20,1), and (20,2).

FIGURE 2.10
Minitab dialog for comparison of sampling plans (n, c): (10,1), (20,1), and (40,1).
32 Acceptance Sampling in Quality Control

Operating characteristic (OC) curve

1.0
n c
10 1
20 1
0.8 40 1
Probability of acceptance

n sample size
c acceptance number
0.6

0.4

0.2

0.0

0 5 10 15 20 25
Lot percent defective

FIGURE 2.11
Minitab OC curve for comparison of sampling plans (n, c): (10,1), (20,1), and (40,1).

Problems
1. A lot of 50 items contains 1 defective unit. If one unit is drawn at random from the lot,
what is the probability that the lot will be accepted if c = 0?
2. A bottle of 500 aspirin tablets is to be randomly sampled. The tablets are allowed to
drop out one at a time to form a string, those coming out first at one end and those
coming out last at the other. A random number from 1 to 1000 is selected and divided
by 2, rounding up. The tablet in the corresponding numerical position is selected.
Is this procedure truly random?
3. Two out of six machines producing bottles are bad. The bottles feed in a successive
order into groups of six, which are scrambled during further processing and packed
in six packs. In how many different orders can the two defective bottles appear
among the six?
4. Six castings await inspection. Two of them have not been properly finished. The
inspector will pick two and look at them. How many groups of two can be formed
from the six castings? How many groups of two can be formed from the two defec-
tive castings? What is the probability that the inspector will find both castings looked
at are bad?
5. Form a probability tree to obtain the probability that the inspector will find both
castings bad in Problem 4.
6. Use the probability calculus to find the probability that the inspector will find two
bad castings in selecting two. Why is it not 2/6 × 2/6 = 4/36 = 1/9? What is the
Probability and the Operating Characteristic Curve 33

probability that they are both good? What is the probability that they are both the
same? What types of events allow these probabilities to be added?
7. At a given quality level, the probability of acceptance under a certain sampling
plan is .95. If the lot is rejected the sampling plan is applied again, “just to be
sure,” and a final decision is made. What is the probability of acceptance under this
procedure?
8. Draw the OC curve for the plan n = 3, c = 0. What are the approximate AQL, indiffer-
ence quality (IQ), and LTPD values for this plan?
9. In a mixed acceptance sampling procedure, two types of plans are used. The first plan
is used only to accept. If the lot is not accepted, the second plan is used. If both types
of plans have PQL = .03, CQL = .09 with α = .05 and β = .10. What is the probability of
acceptance of the mixed procedure when the fraction defective is .09?
10. At the IQ level, the probability of acceptance is .5. In five successive independent lots,
what is the probability that all fail when quality is at the IQ level? What is the prob-
ability that all pass? What is the probability of at least one failure?

References
Kolmogorov, A. N., 1956, Foundations of the Theory of Probability, 2nd ed., Chelsea, New York.
Savage, L. J., 1972, Foundations of Statistics, 2nd ed., John Wiley & Sons, New York.
United States Department of the Navy, 1946, General Specifications for Inspection of Material,
Superintendent of Documents, Washington, DC. Appendix X, April 1, 1946; see also U.S.
Navy Material Inspection Service, Standard Sampling Inspection Procedures, Administration
Manual, Part D, Chapter 4.
von Mises, R., 1957, Probability, Statistics and Truth, 2nd ed., Macmillan, New York.
Whitworth, W. A., 1965, Choice and Chance, Hafner, New York.
3
Probability Functions

Many sampling situations can be generalized to the extent that specific functions have
proved useful in computing the probabilities associated with the operating characteristic
curve and other sampling characteristics.
These are functions of a random variable X that take on specific values x at random with
a probability evaluated by the function. Such functions are of two types:

Frequency function: It gives the relative frequency (or density) for a specific value of
the random variable X. It is represented by the function f(x).
Distribution function: It gives the cumulative probability of the random variable X up to
and including a specific value of the random variable. It can be used to obtain prob-
ability over a specified range by appropriate manipulation. It is represented by F(x).

In the case of a discrete, go/no-go, random variable

f ( x ) = P ( X = x )

and the distribution function is simply the sum of the values of the frequency function up
to and including x:
X

F (x) = å f (x)
i =0
X discrete

When X is continuous, that is, a measurement variable, it is the integral from the lowest
possible value of X, taken here to be –∞, up to x:
x

F (x) =
ò f (t )dt
-¥
X continuous

where the notation

b

ò f (t ) dt
a

may be thought of as representing the cumulative probability of f(t) from a lower limit
of a to an upper limit of b. In either case, these functions provide a tool for assessment of
sampling plans and usually have been sufficiently well tabulated to avoid extensive math-
ematical calculation.
The probability functions can be simply illustrated by a single toss of a six-sided die.
Here, the random variable X is discrete and represents the number of spots showing
on the upward face of the die. It takes on the values 1, 2, 3, 4, 5, and 6. This is called the

35
36 Acceptance Sampling in Quality Control

sample space. Since the probability of any of these values is constant, namely, 1/6, the
frequency function is

1
f (x) = x = 1, 2, 3, 4, 5, 6
6

and the distribution function is

x

å6 = 6
1 x
F (x) =
i =1

With these, it is possible to determine the probability of rolling a 1

1
f ( 1) =
6

or of getting a result of 3 or less

3
F (3) =
6

Values of the random variable over a range may be found by subtraction. Thus, the prob-
ability of throwing a 4 or a 5 is

P ( 4 or 5 ) = P ( X £ 5 ) - P ( X £ 3 )
5 3 2
= F ( 5) - F ( 3) = - =
6 6 6

Probability Distributions
Using the frequency function, it is possible to find the distribution of probabilities over
all possible values of the random variable X. The frequency function and the distribution
function may then be displayed in tabular form as follows:

X f(x) F(x)

1 1/6 1/6
2 1/6 2/6
3 1/6 3/6
4 1/6 4/6
5 1/6 5/6
6 1/6 6/6

When plotted, the probability distribution is shown in terms of its frequency function in
Figure 3.1 and in terms of its distribution function in Figure 3.2.
Probability Functions 37

f(x)

1/6

1 2 3 4 5 6 x

FIGURE 3.1 
Frequency function for die.

6/6

5/6

4/6
F(x)

3/6

2/6

1/6

0 1 2 3 4 5 6 x

FIGURE 3.2 
Distribution function for die.

Now consider a continuous distribution. An example might be the position of the second
hand of watches when they stop. The distribution of these values could be assumed to be
rectangular in the interval from 0 to 60. The frequency function of such a distribution is
1
f (x) = 0 £ x < 60
60

and its distribution function is

x
F (x) = 0 £ x < 60
60

If measured close enough, there is an infinity of possible positions at which the second
hand might stop (e.g., 47.2186327 … s). The probability of stopping exactly at any given
position, specified to an infinity of possible decimal places, is infinitesimally small. This is
why the frequency function is often referred to as a probability density function in the
38 Acceptance Sampling in Quality Control

continuous case. It shows density, not probability. This is true for all continuous distribu-
tions. The distribution function cannot be obtained by summing the values of the frequency
function in the same sense as with discrete data but requires use of the calculus. Thus,
x

ò
F ( x ) = f ( t ) dt
0
x
1
F (x) =
ò 60 dt
0
x
F (x) = 0 £ x < 60
60

A plot of the probability distribution is given in Figure 3.3 and a graph of the distribution
function is given in Figure 3.4. Such graphs are useful in visualizing the shape, nature, and
properties of distribution functions.

f(x)

1/60

0 30 60 x

FIGURE 3.3 
Frequency function for watch stoppage.

1.0
F(x)

0.5

0
0 30 60 x

FIGURE 3.4 
Distribution function for watch stoppage.
Probability Functions 39

Measures of Distribution Functions

There are several important measures of distribution functions that show the location and
spread of the distribution. The most important measure of location, or central tendency, is
the first moment of the distribution or its mean (center of gravity). It gives the arithmetic
average of the values in the distribution and is its expected value. The mean of a distribu-
tion is calculated as follows:

m= åxf ( x )
all x
discrete distribution

or
¥

m=
ò xf ( x ) dx
-¥
continuous distribution

where the limits for the continuous distribution are taken at the extreme values of X.
For the discrete distribution of the results of a toss of the die, we have

æ1ö æ1ö æ1ö æ1ö æ1ö æ1ö

m = 1 ç ÷ + 2 ç ÷ + 3 ç ÷ + 4 ç ÷ + 5 ç ÷ + 6 ç ÷ = 3.5
è6ø è6ø è6ø è6ø è6ø è6ø

while for the distribution of the second hand

60 60
1 x2 3600

ò
m = x dx =
60
0
2 ( 60 )
0
=
120
= 30

Note that for a finite population of size N

Sx
m=
N

Other measures of location of a distribution are the median (middle value) and the mode
(most frequently occurring value).
The standard deviation stands as the primary measure of the spread of a distribution.
It is the square root of the second central moment about the mean (moment of inertia). For
discrete data, it is calculated as

s = S ( x - m) f ( x)
2

For continuous data, the variance or square of the standard deviation is calculated as
¥

ò ( x - m) f ( x ) dx
2
s2 =
-¥
40 Acceptance Sampling in Quality Control

so

s = s2

The standard deviation is the root-mean-square average deviation of an observation from

the mean. In this sense, it can be considered to measure the average distance of an observa-
tion from the mean.
For the results of the die, we have

æ1ö 2æ1ö 2æ1ö 2æ1ö 2æ1ö 2æ1ö

(1 - 3.5 ) ç ÷ + ( 2 - 3.5 ) ç ÷ + ( 3 - 3.5 ) ç ÷ + ( 4 - 3.5 ) ç ÷ + ( 5 - 3.5 ) ç ÷ + ( 6 - 3.5 ) ç ÷
2
s=
6
è ø 6
è ø 6
è ø 6
è ø 6
è ø è6ø
6.25 + 2.25 + .225 + .25 + 2.25 + 6.25
=
6
17.5
= = 1.71
6

Note that for a finite population of size N

S ( x - m)
2

s=
N

The standard deviation of the continuous distribution of stopping times of the second
hand is
60
1
ò ( x - 30 )
2 2
s = dx
60
0

60

=
ò (x 2
- 60 x + 900 ) 601 dx
0

60
x 3 60 x 2 900 x
= - +
180 120 60 0

60 3 60 3 900 ( 60 )
= - +
180 120 60
= 1200 - 1800 + 900

= 300

so

s = 300 = 17.3

The other principal measure of spread used in acceptance sampling is the range
(the  ­difference between the highest and the lowest observed values). This is not usually
Probability Functions 41

applied to populations, but rather to measure the spread in sample data. The primary mea-
sures of the sample location and spread are the sample mean (X) and the sample standard
deviation (s). The appropriate formulas for a sample of size n are
Sx
X=
n

and

S(x - x)
2

s=
n -1

The n − 1 denominator in the standard deviation formula can be shown to be necessary

to make the expected (mean) value of s2 equal to σ2 when the sample variances are aver-
aged over all possible samples of size n from the population. Often sample estimates
are denoted by a carat over the symbol for the parameter. Thus, for example, we have
ŝ = s with measurement data and use the symbol p̂ to represent an estimate of p from
attribute data.

Hypergeometric Distribution
The hypergeometric distribution is fundamental to much of acceptance sampling.
It is applicable when sampling an attribute characteristic from a finite lot without
­replacement. Here
N is the lot size, N > 0
p is the proportion defective in the lot, p = 0, 1/N, 2/N, …, 1
q is the proportion effective in the lot, q = 1 – p
n is the sample size, n = 1, 2, …, N
x is the number of occurrences, x = 0, 1, 2, …, n
Its frequency function is

CxNpCnNq- x
f (x) = ,
CnN

where, because of discreteness in the lot, the proportion defective is restricted to one of
the values p = 0, 1/N, 2/N, 3/N, …, 1. A recursion formula to obtain successive values of the
hypergeometric distribution is

f ( x + 1) =
( n - x ) ( Np - x ) f x
( )
( x + 1) ( Nq + x - n + 1)

The hypergeometric was, in fact, the distribution used in Chapter 2 to obtain the probabil-
ity of a four-card hand of aces; there

4 48
N = 52 p = q= n=4
52 52
42 Acceptance Sampling in Quality Control

Now

æ 4 ö
Np = 52 ç ÷ = 4
è 52 ø
and

æ 48 ö
Nq = 52 ç ÷ = 48,
è 52 ø
where
Np is the number of defective units in the lot
Nq is the number of effective units in the lot

The formula gives

C44C048
f ( 4) =
C452
4 ! 48 !
4 ! 48 ! 4 ×3 ×2×1 1
= ! 0 ! 0 ! 48 ! =
4 = =
52 52 ! 52 × 51 × 50 × 49 270, 725
4 ! 48 !

and we see the usefulness of a ready-made probability function in solving a problem.

Again, in the acceptance sampling problem of Chapter 2, the plan n = 2, c = 0 was applied
to a lot of size 3 containing 1 defective. Since sampling was without replacement, the
hypergeometric distribution is applicable. Here

N = 3 Np = 1 Nq = 2 n = 2

and
1! 2!
×
C01C22 0 ! 1! 2 ! 0 ! 1 × 1 1
f (0) = = = =
C23 3! 3 3
2 ! 1!
as before.
In this simple problem, it may be perfect to completely specify the distribution for 0 or 1
defectives in a sample of 2. Using the recursion formula, we can list out the distribution as

x p(x)

0 1/3
1 2/3

since f(1) may be obtained from f(0) by

f ( 1) = f ( 0 + 1) =
( 2 - 0 ) (1 - 0 ) 1 = 2 æ 1 ö = 2
( 0 + 1) ( 2 + 0 - 2 + 1) 3 çè 3 ÷ø 3
Probability Functions 43

For the hypergeometric distribution, the mean is

æ1ö 2
m = np = 2 ç ÷ =
è 3 ø 3

Thus, we would expect to get an average of two defective units in every three draws.
The standard deviation is

N -n æ 1 öæ 2 ö 3 - 2 2 1
s = npq = 2ç ÷ç ÷ = = .471
N -1 è 3 øè 3 ø 3 -1 3 2

which represents the average distance of an observation from the mean using the root-
mean-square average. Since the mean is 2/3, we see that one-third of the observations
(which are zeros) deviate from the mean by 2/3 and two-thirds of the observations, which
are ones, deviate from the mean by 1/3. Taking the arithmetic average, we obtain the mean
deviation (MD):

MD =
(1/3 )( 2/3 ) + ( 2/3 )(1/3 ) = 4 = .444
1/3 + 2/3 9

Moreover, if we had used the mode as an average, the modal deviation (MOD) is

1
MOD = = .333
3

The median deviation is also .333. And so the standard deviation can be seen to be just one
method of computing the average distance of an observation from the mean.

Binomial Distribution
Undoubtedly, the most used distribution in acceptance sampling is the binomial.
It  ­complements the hypergeometric in the sense that it is employed when sampling an
attribute characteristic from an infinite lot (or process) or a finite lot when sampling with
replacement. Here
n is the sample size, n > 0
p is the proportion defective, 0 ≤ p ≤ 1
q is the proportion effective, q = 1 – p
x is the number of occurrences x = 0, 1, 2, …, n
Its frequency function is

f ( x ) = Cxn p x ( 1 - p )
n- x
= Cxn p x qn - x

The mean of the binomial distribution is

m = np
44 Acceptance Sampling in Quality Control

and its standard deviation is

s = npq

Values of the frequency function can be calculated recursively using the following formula:

f ( x + 1) =
(n - x) p f x
( )
( x + 1) q

As an illustration of the binomial distribution, consider the sampling plan n = 5, c = 0 pre-

sented in Chapter 2. In setting up such a plan, it may be desirable to obtain the distribution
of the number of defectives in a sample of 5 when p is at the producer’s process average
proportion defective, say, p = .01. We have

5!
f ( 0 ) = C05 ( .01) ( 1 - .01) = (1)(.99 ) = .951
0 5 5

0! 5!

Similarly

5!
f ( 1) = C15 ( .01) ( 1 - .01) (.01)(.99 ) = 5 (.01)(.961) = .048
1 5 -1 4
=
1! 4 !

and using the recursion formula for f(2)

f ( 2 ) = f ( 1 + 1) =
( 5 - 1) æ .01 ö .048 = .001
( )
(1 + 1) çè .99 ÷ø

so that the distribution of the number defective is

x f (x)

0 .951
1 .048
2 .001
3 .000
4 .000
5 .000

It is apparent that it is highly unlikely to obtain 3, 4, or 5 defectives in application of this

plan. Such a result would be a clear indication that the process proportion defective is
higher than .01.
For this distribution, the mean is

m = 5 ( .01) = .05

and the standard deviation is

s = 5 ( .01)( .99 ) = .0495 = .22

Probability Functions 45

It should be noted that in using tables for this distribution, it is possible to use 1 − p as an
argument instead of p and vice versa. This is done using the following relationship:

B( x n, p) = 1 - B(n - x - 1 n, 1 - p),

where B(x|n, p) is read as the binomial distribution function evaluated at x given ­parameters
n and p. Thus, for example, B(1|5,  . 01) =  . 999, which may be obtained as

B(1 5,.01) = 1 - B(3 5,.99)

3

åC (.99) (.01)
5 x n- x
= 1- x
x =0

= 1 - éC05 ( .99 ) ( .01) + C15 ( .99 ) ( .01) + C25 ( .99 ) ( .01) + C35 ( .99 ) ( .01) ù
0 5 1 4 2 3 3 2
ë û
= .999

The Larson (1966) nomograph for the binomial distribution is extremely useful in accep-
tance sampling applications (see Figure 3.5). The probability of c or fewer successes in a
sample of n for a specific proportion defective p is characterized by a single line on the
chart. The point representing p is set on the left scale, the pair of values n and c determine a
point in the grid, and the cumulative probability P(x ≤ c) is read from the right scale. Thus,
when p = .25, the plan n = 20, c = 6 has P(x ≤ 6) = .79, which is, of course, the probability of
acceptance. A straight line connecting any two of the points representing p, (n, c) or P(X ≤ c)
will give the third. Thus, the Larson nomograph is a very versatile tool for use in evaluat-
ing acceptance sampling plans.

Poisson Distribution
The Poisson distribution is used in calculating the characteristics of sampling plans,
which specify a given number of defects per unit such as the number of defective rivets
in an a­ ircraft wing or the number of stones allowed in a piece of glass of a given size. The
parameter in the Poisson distribution is simply μ. Here
μ is the mean number of defects, μ > 0
x is the number of occurrences, x = 0, 1, 2, …
and its frequency function

m x e -m
f (x) =
x!

where e = 2.71828…. Values of e−x are shown in Appendix Table T3.1. The mean and ­standard
deviation are simply

m = m s = m
 46 Acceptance Sampling in Quality Control

0.01 c
n!
P {m ≤ c} = p m (1 – p)n–m
0 m!(n – m)!
0 100 m=0

0 Example
70 Required: x sampling plan having
0.02 0
5 50 pt = .95 of pv = .02
0
10 40 pt = .10 of px = .08
0.03 0
30 Solution: make alignments and read

Nu
sample size [η] and acceptances

mb
0
0.04 20 20 number (c) as in diagram below:

e ro
ences (c)

0 p
14

f tr
0.05 30 0.001

ial
0 p
10 0.02

so
0.06 40

rs
of occurr

0.10 0.005
70

am
0.07 50 0x
0.01

ple
0.08 50
A 19B 0.95

siz
40
Probability of occurrence in a single trial (p)

70 0.02
0.09

e(
Number

c 14
30

n)

Probability of c or fewer occurrence in n trials (P)

0.10 0.05
100 20
0.10

10 0.20
0.15 140
0.30
0.40
5 0.50
0.20 200 0.60
2 0 0.70
0.80
0.25
0.90
1
0.30 0.95
2
0.98
s (c)

0.35 3 0.99
nce

4 0.995
urre

5
Occ

0.40
7 0.999

0.45 9

0.50

FIGURE 3.5 
Larson binomial nomograph. If p is less than 0.01, set kp on the p-scale and n/k on the n-scale, where
k = 0.01/9, rounded upward conveniently. (Reprinted from Larson, H.R., Ind. Qual. Control, 23, 273, 1996.
With permission.)

Successive values of the Poisson distribution can be calculated using the recursion formula:

m
f ( x + 1) = f (x)
x +1

Suppose an importer of glassware wishes to ensure that the process average of his sup-
plier is no more than the specified two bubbles per piece. The number of bubbles would
be expected to vary from piece to piece. The Poisson distribution can be used to determine
Probability Functions 47

how the number of bubbles per piece would vary if the producer maintained the agreed
upon average. Evaluating the Poisson distribution in this case, we obtain

20 e -2
f (0) = = e -2 = .1353
0!
21 e -2
f ( 1) = = 2e -2 = .2707
1!
22 e -2
f ( 2) = = .2707
2!

and so on. Using the recursion relationship, subsequent values can be obtained. For example,

2
f ( 3 ) = f ( 2 + 1) = (.2707 ) = .1805
3
2
f ( 4 ) = f ( 3 + 1) = ( .1805 ) = .0902
4
2
f ( 5 ) = f ( 4 + 1) = ( .0902 ) = .0361
5
2
f ( 6 ) = f ( 5 + 1) = ( .0361) = .0120
6
P ( X > 6 ) = 1 - .1353 - .2707 - .2707 - .1805 - .0902 - .0361 - .0120 = .0045

Note that there is no upper limit on the number of bubbles that could be obtained so that
the probability distribution is

x f (x)

0 .1353
1 .2707
2 .2707
3 .1805
4 .0902
5 .3610
6 .0120
>6 .0045

We see that pieces with more than six bubbles would be very rare, occurring less than
half a percent of the time. On the average, we would expect two bubbles per piece with a
standard deviation of

s = 2 = 1.41

A very useful tool in determining Poisson probabilities is the Thorndyke chart (Figure 3.6).
This chart shows the probability of x or fewer defects on the vertical axis and gives values
of μ on the horizontal axis. To use the chart, a vertical line is drawn at μ for the Poisson
distribution to be evaluated. Its intersection with the curves for x = 0, 1, 2, …, determines
the cumulative probability of x or fewer defects when read on the probability axis hori-
zontally from the intersection. This chart was developed by Thorndyke (1926) and was
subsequently modified by Dodge and Romig (1941).
 48

c=6 7 8 9 10 15 20 30 40 c = 50
0.99999 0.99999
4 5

0.9999 3 50 0.9999
9 10
0.999 2 8 0.999
7
6
5
0.99 4 30 0.99
1
3 20
40
2 15
0.9 c=0 0.9
1 10
0.8 9 0.8
8
0.7 7 0.7
0.6 0 6 0.6
5 30
0.5 4
0.5
0.4 3
0.4
0.3 0.3
2
0.2 0.2
1
0.1 0.1
0

20

Probability of occurrence of c or less defects

Probability of occurrence of c or less defects

0.01 0.01

15
0.001 0.001

0.0001 0.0001
10
0.00001 0.00001
c=0 c=5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 2 3 4 5 6 7 8 9 10 20 30
Value of pn

FIGURE 3.6 
Thorndyke chart. (From Dodge, H.F. and Romig, H.G., Sampling Inspection Tables: Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 35.
With permission.)
Acceptance Sampling in Quality Control
Probability Functions 49

f-Binomial Distribution
The f-binomial distribution is well known as an approximation of the binomial; however,
it is useful as a distribution in its own right. It describes the distribution of defects in ran-
dom samples without replacement from a finite population containing a known number
of defects (Schilling 2005). Here
N is the lot size, N > 0
n is the sample size, n > 0
D is the number of defects in the lot, D ≥ 0
x is the number of occurrences, 0 ≤ x ≤ D
Its frequency function is
x D- x
æ D öæ n ö æ N - n ö
f (x) = ç ÷ç ÷ ç ÷
è x øè N ø è N ø

A recursion formula to obtain successive values of the f-binomial distribution is

æ D - x öæ n ö
f ( x + 1) = ç ÷ç ÷ f (x) ,
è x + 1 øè N - n ø

where, of course,

D
æ N -nö
f (0) = ç ÷
è N ø

For the f-binomial distribution, the mean is

Dn
m=
N

and the standard deviation is

N-n
s= m ,
N

where ( N - n )/N acts as a finite population correction factor to the usual Poisson infinite
population standard deviation, m.
The properties of the f-binomial can be found from the conventional binomial ­distribution
using the following relations:

Binomial f-Binomial
n D
p n/N
x x
50 Acceptance Sampling in Quality Control

As an example, suppose there are two defects in a lot of 10. A sample of 4 is taken from the
lot. Then, using the recursion formula where N = 10, n = 4, and D = 2, we have
D 2
æ N -nö æ 10 - 4 ö
f (0) = ç ÷ =ç ÷ = 0.36
è N ø è 10 ø
f ( 1) = f ( 0 + 1)
æ D - x öæ n ö
=ç ÷ç ÷ f (0)
è x + 1 øè N - n ø
æ 2 - 0 öæ 4 ö
=ç ÷ç ÷ ( 0.36 )
è 0 + 1 ø è 10 - 4 ø
æ 2 öæ 4 ö
= ç ÷ ç ÷ ( 0.36 )
è 1 øè 6 ø
= 0.48
f ( 2 ) = f ( 1 + 1)
æ 2 -1 öæ 4 ö
=ç ÷ç ÷ ( 0.48 )
è 1 + 1 ø è 10 - 4 ø
æ 1 öæ 4 ö
= ç ÷ ç ÷ ( 0.48 )
è 2 øè 6 ø
= 0.16

Furthermore, the mean of this distribution is

D

m= åxf ( x ) = 0 ( 0.36 ) + 1( 0.48 ) + 2 ( 0.16 ) = 0.80
x =0

or

Dn ( 2 )( 4 )
m= = = 0.80
N 10

and the variance is

D

å( x - m) f ( x ) = ( 0 - 0.80 ) ( 0.36 ) + ( 1 - 0.80 ) ( 0.48 ) + ( 2 - 0.80 ) ( 0.16 ) = 0.48

2 2 2 2
s2 =
x =0

or

N-n 10 - 4
s= m = 0.80 = 0.6928
N 10

and

s2 = ( 0.6928 ) = 0.48
2

Probability Functions 51

Negative Binomial Distribution

It is sometimes necessary to determine the number of random trials required to obtain
a given number of defectives. The negative binomial is the probability distribution that
is used to obtain the probability of a given number of trials up to and including the xth
defective. The parameters of the negative binomial distribution are similar to those of the
binomial itself. Here
n is the number of trials to and including the xth defective, n ≥ x
p is the proportion defective, 0 ≤ p ≤ 1
q is the proportion effective, q = 1 − p
x is the number of occurrences x = 1, 2, …
Its frequency function may be represented as b−1(n|x, p) so that

( )
f ( n ) = b -1 n x , p = Cxn--11 p x qn - x

with a mean of

x
m=
p

and a standard deviation of

xq
s=
p

Successive values of the negative binomial distribution may be calculated using the recur-
sion relation:
n
f ( n + 1) = q f ( n)
n- x +1

The negative binomial gives the number of trials to a fixed number of successes, rather
than the number of successes in a fixed number of trials as does the binomial. The term
negative binomial relates to the fact that the successive values of the frequency function
can be determined from an expansion of
-x
æ1 qö
ç - ÷
èp pø

which, of course, is of the binomial form.

Sometimes called the Pascal or Pólya distribution, the cumulative negative binomial is
related to the cumulative binomial by the following relation:
n x -1

åi=x
Cxi--11 p x qi - x = 1 - åC p q
i =0
n i n-i
i

52 Acceptance Sampling in Quality Control

which shows that the negative binomial distribution function for up to n trials to obtain
x successes is equal to the complement of the binomial distribution function for x − 1
successes in n trials. Using B−1(n|x, p) for the negative binomial distribution function and
B(x|n, p) for the binomial distribution function, we have

B-1(n x , p) = 1 - B( x - 1 n, p)

and individual terms are simply

x

(
b -1 n x , p = ) n
(
b x n, p

)
Consider the sampling plan that was used earlier to illustrate the binomial distribution
when p = .01, namely, n = 5 and c = 0. Suppose we wish to calculate the probability of 1, 2,
3, 4, 5, or more trials before finding a defective in random samples from a large lot. Using
the negative binomial distribution, we have

f ( 1) = C11--11 ( .01) ( .99 ) = C00 ( .01)( 1) = 1 ( .01)( 1) = .01

1 1- 1

Furthermore,

f ( 2 ) = C12--11 ( .01) ( .99 ) = C01 ( .01)( .99 ) = 1 ( .01)( .99 ) = .0099

1 2 -1

and tabulate the probability of 3 and 4 trials before finding a defective using the recursion
formula:

2
f ( 3 ) = ( .99 ) (.0099 ) = .0098
2 -1+1
2
f ( 4 ) = ( .99 ) (.0098 ) = .0097
3 -1+1

and finally

f ( 5 ) = C15--11 ( .01) ( .99 ) = 1 ( .01) ( .99 ) = .0096

1 4 1 4

so that

P ( >5 ) = 1 – .01 – .0099 – .0098 – .0097 – .0096 = .951

and the distribution is

n F(n)
1 .0100
2 .0099
3 .0098
4 .0097
5 .0096
>5 .9510
Probability Functions 53

The expected number of trials to a defective is

1
m= = 100
.01

and the standard deviation of the number of trials on which the first defective occurs is

1 ( .99 )
s= = 99.5
.01

which indicates a large spread in the number of trials to a defective.

As a check on the probability distribution, we may observe

(
B-1 ( n x , p ) = 1 - B x - 1 n, p )
B -1
( 5 1, .01) = 1 - B ( 0 5, .01)
.049 = 1 - .951
= .049

where the binomial probability was calculated previously in the discussion of binomial
probabilities.

Exponential and Continuous Distributions

Continuous distributions are extremely useful in acceptance sampling, although some-
what more complicated and restrictive in their use. Sampling plans based on attribute
data are typically nonparametric by nature, that is, it is not necessary to know the shape
or parameters of the distribution of any measurements involved to use an attribute plan.
This is not generally true for the variable sampling plans, which are based on measure-
ment (variable) data. These distributions are usually continuous and require specification
of shape and parameters, such as measures of location and spread. We will consider two
such distributions, the exponential and the normal. We shall also consider the Weibull
family of distributions that includes the exponential as a special case and that can be used
to approximate the normal.
The exponential distribution is used extensively in evaluating acceptance plans
for ­reliability and life testing. It is distinguished by a constant failure rate. That is, the
­probability of future failure is constant regardless of how long a unit has been in o
­ peration.
The parameter of this function is simply μ. Here
μ is the mean of the distribution, μ > 0
x is the measurement distributed, x ≥ 0
and its frequency (density) function is

1 - x/m
f (x) = e
m
54 Acceptance Sampling in Quality Control

The density is not as useful as the frequency function for discrete distributions. As a mat-
ter of fact, evaluation of this function will not lead to the probability associated with any
given point in the continuum, since the probability of a point is zero. Consequently, the
density function must be integrated over a range of possible values of the argument to
obtain a probability. This may be expressed in terms of the distribution function. The
­distribution function for the exponential distribution is

x x
1 -t/m -m -t/m

F (x) =
ò
0
m
e dt =
m
e
0
= -e - x/m + 1

and

F ( x ) = 1 – e – x/m

The mean of the exponential distribution is, of course, μ = μ, while, simply enough, its
standard deviation is σ = μ also. Many problems involving the exponential distribution are
couched in terms of its (constant) failure rate λ, which is simply

1
l=
m

Consider a requirement that the mean life of a power transistor must be greater than 5000 h.
That is, its failure rate (h−1) must be less than

1
l= = .0002
5000

Suppose a random unit is tested for 500 h. What is the probability that it will fail in that
period if it comes from a process with a mean life of exactly 5000 h?

F ( 500 ) = 1 - e -500/5000
= 1 - e -0.1
= 1 - .905 = .095,

where the value of e−0.1 was obtained from Appendix Table T3.1. Both the mean and the
standard deviation of this distribution are 5000 h.
The values of the distribution function of the exponential distribution may be found by
using the Thorndyke chart (see Figure 3.6) for the Poisson distribution. Enter with a value
of μ on the horizontal axis equal to the absolute value of the exponent in the exponential
distribution and read the cumulative probability associated with x = 0. This is e–μ, which
when subtracted from one gives the exponential distribution function. A check of that
chart will show that upon entering the x axis with a value of 0.1 and reading the y value
for the curve c = 0, a value of roughly .905 is obtained, and subtracting from 1 gives .095.
This agrees with the previous calculation.
Probability Functions 55

Weibull Distribution
The Weibull distribution may be thought of as a generalization of the exponential distribu-
tion incorporating parameters for location, spread, and shape. The distribution is defined
for positive values of x, starting at zero. The location parameter γ adjusts the distribution to
start at a value γ, other than zero. The scale parameter η or characteristic life is the x value
for which F(x – γ) = .6321 for any Weibull shape. The shape parameter β gives the distribu-
tion flexibility in shape so that it can be used to fit a variety of empirical and theoretical
failure distributions. Here
γ is the location (minimum life) parameter, γ > 0
η is the scale parameter, η > 0
β is the shape parameter, β > 0
x is the measurement distributed, x ≥ γ
The frequency (density) function is

b -1
bæ x-gö - éë( x - g )/hùû
b
f (x) = ç ÷ e
hè h ø

with distribution function

b
- éë( x - g )/hùû
F (x) = 1 - e

The mean of the distribution is

æ 1ö
m = g + hG ç 1 + ÷
è b ø

and its standard deviation is

æ 2ö æ 1ö
s = h G ç 1 + ÷ - G2 ç 1 + ÷
è bø è b ø

where Γ(x) is the gamma function such that

0
ò
G ( x ) = e -tt x -1dt = ( x - 1) !

the factorial function for x – 1, if x is a positive integer.

The exponential distribution is a special case of the Weibull distribution when β = 1.
It will be seen that when this is the case and the location parameter γ = 0, and μ = η, the
mean occurs at the point at which F(x) = .6321. Moreover, σ = η.
56 Acceptance Sampling in Quality Control

2.0

β=4

β=3
f(x)

1.0

β=2

β=1
β = 0.5
0
0 1.0 2.0 3.0 x

FIGURE 3.7 
The Weibull frequency (density) function for various values of shape parameters, β(γ = 0, η = 1).

The shape parameter β allows the distribution to take on a variety of shapes as shown in
Figure 3.7. Specifically, we have
Exponential distribution: β = 1
Rayleigh distribution: β = 2
Approximate normal distribution: β = 3.44
The value β = 3.44 is given as the value of the shape parameter approximating a normal
distribution in the sense that when β = 3.44 the median and the mean of the distribution
are equal to each other. When γ = 0 and η = 1, this distribution has a mean μ = .899 and
standard deviation σ = .289. Thus, normal data x′ with mean μ′ and standard deviation
σ′ should be plotted approximately as a straight line on a Weibull probability paper with
β = 3.44 when transformed using

x¢ - m¢
x = .289 + .889
s¢

with inverse

æ x - .899 ö ¢
x¢ = s¢ ç ÷+m
è .289 ø

The failure rate of the exponential distribution is constant. The failure rate of the Weibull
distribution is decreasing for values of the shape parameter β < 1 and increases for β > 1.
Of course, when β = 1 the failure rate is constant. Since the failure rate changes over the
possible values of x (life), it is quoted in terms of the instantaneous failure rate at any
­chosen value of x. This is called the hazard rate h(x), where

b -1
bæ x-gö
h(x) = ç ÷
hè h ø
Probability Functions 57

Reliability specifications for use in acceptance sampling are sometimes written in terms
of the hazard rate.
Suppose the example given for the exponential distribution is regarded as a special case
of the Weibull distribution. The specified mean in that case was 5000 h, and we have

γ = 0,  η = 5000,  and  β = 1

so that the probability of a failure before 500 h is

1
- éë( 500 - 0 )/5000 ùû
F ( 500 ) = 1 - e = .095

as before.

Normal Distribution
No area of statistics seems to have escaped the impact of the normal distribution. This is
certainly true of acceptance sampling where it forms the basis of a large number of vari-
able acceptance sampling plans. It has pervaded other areas of acceptance sampling as
well.
The normal distribution is completely specified by two parameters μ and σ. Here
μ is the the mean, –∞ < μ < ∞
σ is the the standard deviation, σ > 0
x is the measurement distributed, –∞ < x < ∞
Its frequency function is
1 2
1 - éë( x - m )/s ùû
f (x) = e 2
s 2p

Unlike the exponential and the Weibull distributions, no closed-form formula can be
obtained for the distribution function. Expressed as an integral, it is
x 1 2
1 - ëé( t - m )/s ùû

F (x) =
s 2p
-¥
ò e 2 dt

and is shown cumulated over the standard normal frequency function in Figure 3.8.

F(x)

0 x t

FIGURE 3.8 
Standard normal frequency (density) function (μ = 0, σ = 1).
58 Acceptance Sampling in Quality Control

Fortunately, values of the distribution function may be obtained from tables of the stan-
dard normal distribution such as is given in Appendix T3.2. The table is for the specific
standard normal distribution with μ = 0 and σ = 1. It is tabulated in terms of standard
normal deviates, z, which are simply the x values for the standard normal distribution. To
use the table to obtain probabilities at specific values of x for other normal distributions
(i.e., with different mean values and standard deviations), it is necessary to transform the
x values into the z values given by the table using the formula

x -m
z=
s

Similarly, if an x value is desired, that has a given probability, the probability may be
found in the body of the table in terms of z and the x value will be obtained using the
transformation

x = m + zs

For example, suppose bolts are manufactured by a process having a mean of 50 mm and
a standard deviation of .1 mm. The distribution of bolt lengths conforms to the normal
distribution, that is,

m = 50 mm , s = 0.1 mm

If it is desired to determine what proportion of the bolts have lengths less than 49.8 mm
(which is, of course, the probability of obtaining such a bolt in a random sample), we have

49.8 - 50
z= = -2
0.1

and using Appendix Table T3.2

P ( Z £ –2 ) = .0228

so

P ( X £ 49.8 ) = .0228

Similarly, to determine what length is exceeded by 10% of the bolts, we have

P ( Z £ 1.282 ) = .90

and this would give 10% above the value z = 1.282. Accordingly,

x = 50 + 1.282 ( .1) = 50.1282

Probability Functions 59

The probability of obtaining a result between any two specified values may be found by
subtracting cumulative probabilities. For example, 80% of the bolts (symmetric about the
mean) lie between

z = 1.282 ( 90% below )

and

z = –1.282 (10% below )

or between 49.8718 and 50.1282 mm. We also find that the proportion of bolts between
49.9 mm (z = –1) and 50.3 mm (z = +3) is

z Cumulative Probability

3 .9987
−1 .1587
.8400

or 84% of the bolts.

Certainly, a large share of the importance of the normal distribution in statistics lies in
the central limit theorem, which can be stated as follows:

Central limit theorem: Let f(x) be any frequency (density) function of a population with
finite mean μ and standard deviation σ. Let X be the mean of a random sample of n
from the population. Then the frequency function of X approaches the normal dis-
tribution with mean μ and standard deviation s/ n as n increases without bound.
The theorem is proved in most basic mathematical statistics texts, such as Mood and
Graybill (1973).
It is important to realize that the population distribution is unspecified—the theorem
holds for any underlying population having a finite mean and standard deviation. Thus,
for any population we can say that the distribution of sample mean values will be approxi-
mately normal with

mX = m
s
sX =
n

as the sample size n becomes large. How closely the distribution of sample mean values
is said to approach normality depends, of course, on the shape of the underlying distribu-
tion and the magnitude of n. Shewhart (1931) has demonstrated empirically and Schilling
and Nelson (1976) have shown mathematically that in many applications a sample size of
5 is adequate, 9 is good, and 25 is excellent in assuring a normal distribution of sample
mean values from a variety of fairly well-behaved underlying distributions. Naturally,
when the underlying distribution is normal, the normality of the distribution of sample
mean values is assured.
60 Acceptance Sampling in Quality Control

Suppose samples of size n = 25 are taken from the population of bolts (μ = 50 mm, σ = 0.1)
mentioned previously. Then

m X = 50
0.1
sX = = 0.02
25

and, for instance, we can state that 95% of the possible sample averages from this popula-
tion will be less than

x = 50 + 1.65 ( 0.02 ) = 50.03 mm

Summary of Distributions
A summary of the probability distributions presented in this chapter is given in Table 3.1
for quick reference, the table shows the frequency (density) function, distribution func-
tion, mean, standard deviation, restrictions on the parameters, domain, and use of each
distribution.

Tables of Distributions
Many useful tables have been generated for the evaluation of the probability distribution
shown here. A convenient notation for the range of the argument of the tables is x(y)z,
which indicates the values move from x in increments of y up to z. A few of the tables
applicable in acceptance sampling are the following.

Hypergeometric Tables
Lieberman and Owen (1961) give tables of the hypergeometric frequency and distribution
functions tabulated in the following notation:
N is the lot size
n is the sample size
k is the number defectives in lot (Np here)
x is the argument
P(x) is the value of distribution function (F(x) here)
p(x) is the value of frequency function (f(x) here)
The tables are complete for N ≤ 50 and require interpolation thereafter up to a maximum
lot size of N = 2000. Sufficient values are tabulated through N = 50 to use the relation

F ( N, n, k , x ) = F ( N, k , n, x )
TABLE 3.1
Distributions Useful in Acceptance Sampling
Distribution
Distribution Frequency Function Function Mean Standard Deviation Restrictions Use
x
CxNpCnNq-x N-n
Hypergeometric f (x) = F (x) = å f (i) np npq N>0 Sampling defectives from
CnN i =0
N -1 n = 1, 2, …, N finite lot without
Probability Functions

p = 0, 1/N, 2/N, …, 1 replacement

q=1–p
x = 0, 1, 2, …, n
x
Binomial f ( x ) = Cxn p x qn-x F (x) = np npq n>0 Sampling defectives from
å f (i)
i =0 0≤p≤1 infinite lot or process
q=1–p Sampling defectives from
x = 0, 1, 2, …, n finite lot with
replacement
x
m x e -m
Poisson f (x) = F (x) = å f (i) μ m μ>0 Sampling defects from
x! i =0 x = 0, 1, 2, … area of infinite
opportunity for
occurrence, with mean
occurrence μ
x D- x x
æ nö æ nö Dn N-n
f-Binomial f ( n ) = CxD ç ÷ ç 1 - ÷ F (x) = å f (i) m N>0 Sampling defects without
èNø è Nø i =0
N N n>0 replacement from finite
D≥0 area of opportunity
0≤x≤D containing D defects
x
x xq
Negative binomial f ( n ) = Cxn--11 p x qn-x F (x) = å f (i) n≥x For number units, n,
i ==0
p p 0≤p≤1 sampled up to and
q=1–p including the xth
x = 0. 1, 2, … success
1 -( x/m )
Exponential f (x) = e F(x) = 1 − e−(x/μ) μ μ μ≥0 Life distribution for units
m x≥0 with constant failure
rate λ = 1/μ
(Continued)
61
 62

TABLE 3.1 (Continued)

Distributions Useful in Acceptance Sampling
Distribution
Distribution Frequency Function Function Mean Standard Deviation Restrictions Use
b-1
b
bæ x-g ö - ( x - g )/h
( ) b æ 1ö æ 2ö æ 1ö
Weibull f (x) = ç ÷ e F( x) = 1 - e- (( x-g )/h) g + hG ç 1 + ÷ h G ç 1 + ÷ - G2 ç 1 + ÷ γ>0 Family of life distribution
hè h ø è b ø è b ø è b ø η>0 with decreasing (β < 1),
β>0 constant (β = 1), or
x≤γ increasing (β > 1) hazard
b-1
bæ x-g ö
rate h ( x ) = ç ÷
hè h ø
x
2
1 -(1/2 )( ( x -m )/s )
Normal f (x) = e F (x) = μ σ −∞ < μ < ∞ Common underlying
f 2p ò f (t ) dt
-¥ σ>0 measurement
−∞ < x < ∞ distribution
Distribution of sample
mean values
Useful as approximation
Acceptance Sampling in Quality Control
Probability Functions 63

that is,
x x
Cik CnN--i k CinCkN--i n

åi =0
CnN
= å
i =0
CkN

to allow reversing the roles of k and n in obtaining values from the tables. Other symme-
tries that may be utilized are

F( N , n , k , x ) = F( N , N - n , N - k , N - n - k + x )

= 1 - F( N , n, N - k , n - x - 1)

= 1 - F( N , N - n, k , k - x - 1)

All these relationships apply to the frequency functions also.

Binomial Tables
The three tables that cover the binomial distributions are

1. U. S. Department of Commerce Applied Mathematics Series No. 6 (1950) gives the

values of the binomial frequency function as

f ( r ) = Crn p r qn - r

and the cumulative probabilities from r to n, that is,

n

1 - F ( r - 1) = åC p q
s=r
n s n-s
s

These are given for the following values:

p =  . 01(.01) . 50,  q = 1 − p,  n = 2(1)49,  r = 0(1)(n − 1)

2. Romig (1953) gives values of f(x) and F(x) for p = .01(.01).50,

p = 1 − p,  n = 50(5)100,  and  x  up  to  F(x) =  . 99999

3. Harvard Computation Laboratory (1955) also presents tables of the cumulative

binomial distribution from r to n, that is,
n

1 – F ( r – 1) = B ( r, n, p ) = åC p (1 - p )
n x n- x
x
x =r

for ranges of p between .01 and .50 with r = 0(1)n and n = 1(1)50(2)100(10)200
(20)500(50)1000.
Procedures and examples useful in applying binomial tables have been given
by Nelson (1974). Of course, the binomial tables may easily be used to determine
values of the f-binomial distribution using the relation of the parameters given
earlier in the discussion of the f-binomial.
64 Acceptance Sampling in Quality Control

Poisson Tables
Molina (1942) has tabulated the Poisson distribution in terms of its frequency function
axe - a
f (x) =
x!

and cumulative distribution from c to ∞

¥
axe - a

1 - F ( c - 1) = å
x =c
x!

for

m = a = .001 ( .001) .01 ( .01) .30 ( .10 ) 15 ( 1) 100

Another useful set of tables has been prepared by the Defense Systems Department,
General Electric Company (1962). They tabulate
U x e -U
f (x) =
x!

and the cumulative distribution

x
U r e -U

F (x) = år =0
r!

over an extensive range from U = .0000001 to U = 205.

Negative Binomial Tables

Williamson and Bretherton (1963) present values of the negative binomial in terms of the
following:
k is the number of successes (x here)
n is the number of nonsuccesses (n – x here)
p is the probability of success
for a total of n* = n + k trials to reach the kth success, where n* is the values of n as presented
in the formula for the negative binomial distribution given here. They give the frequency
function

f ( n* - k ) = P ( n ) = Ckn-+1k -1 pi qn

and distribution function

n

F ( n* - k ) = F ( n ) = åC
r =0
r + k -1 k r
k -1 pq

Thus, to find the number of trials n* to get the kth success, it is necessary to look up the
probability under p, k, and n = n* – k.
Probability Functions 65

Probabilities associated with successive values of n are given for selected combinations
of p and k from p = .05, k = 0.1(0.1)0.5 up to p = .95, k = 2(2)50(10)200.

Exponential and Weibull Tables

Since the exponential and the Weibull are continuous distributions with an explicit
­distribution function in closed form,

Exponential: 
F(x) = 1 – e–x/μ
b
F( x ) = 1 - e [ ] - ( x - g )/h
Weibull: 

It is only necessary to obtain tables of e–x to evaluate them. Such tables are available in any
mathematical handbook such as the U.S. Department of Commerce Applied Mathematics
Series No. 55 (1964). Many hand calculators have such values built-in. Appendix Table T3.1
gives selected values of e–x.

Normal Distribution Tables

The U.S. Department of Commerce Applied Mathematics Series No. 23 (1953) gives extensive
tables of the standard normal distribution showing values of the frequency (density) function
1 - x2/2
f (x) = e
2p

and cumulative probabilities from –x to +x as

x
1
F ( x ) - F ( -x ) =
òe
- x 2/ 2
dx
2p
-x

for x = 0(.0001)1(.001)7.800 and above and also from x to ∞

¥
1
1 - F (x) =
ò
2
e - x /2dx
2p
x

for x = 6(.01)10.

Summary
Examples of some of these tables are given in the appendix, and they include the following:

Appendix Table Distribution Source Shows Range

T3.2 Normal Burr (1953) F(z) z = –3.59(.01)3.59

T3.3 Hypergeometric Lieberman and Owen (1961) F(x), f(x) N = 1(1)10
T3.4 Binomial Harvard Tables (1955) 1 – F(r – 1) n = 1(1)33
T3.5 Poisson Molina (1942) 1 – F(c – 1) np = .001 to 10.0
66 Acceptance Sampling in Quality Control

If possible, the reader should learn to use the tables cited or similar tables in conjunction
with work in acceptance sampling. The small set of tables compiled by Odeh et al. (1977)
will be found particularly useful. Many other tables present values of these and other
probability distributions together with information useful in acceptance sampling. These
include Owen (1962), Beyer (1968), and Burington and May (1970), among others. While
computers and hand calculators will readily produce specific values, considerable insight
into the nature of these distributions can be had by reference to these tables. Furthermore,
the introductory material in the tables frequently contains information on the distribu-
tions not readily available elsewhere.

Useful Approximations
The complexity of the hypergeometric distribution, and to some extent the binomial,
makes it necessary to approximate these distributions at times with other, more tractable
distributions. Fortunately, rules have been derived, that, when adhered to, ensure that
reasonably good approximations will be obtained. Naturally, such rules depend upon just
how close one distribution is expected to come to another. A schematic chart showing
some distribution functions approximating the hypergeometric and the binomial distribu-
tions is presented in Figure 3.9.
The hypergeometric may be approximated by the ordinary p-binomial when the sam-
ple size is less than 10% of the population size. When the sample represents more than
10% of the population, the f-binomial may be used for calculations involving a proportion

Hypergeometric
Nq
C xNp Cn–x
f(x) =
CN
n

n n
≤ 0.1 > 0.1
N N

p-Binomial
p ≤ 0.1 p > 0.1
f(x) = C nx p x qn–x

f-Binomial
x Np–x None
np < 5 np ≥ 5 n n
f(x) = CxNp 1–
N N

Normal
Poisson µ = np
µ = np
σ = √npq

FIGURE 3.9 
Distributions approximating the hypergeometric and binomial. (Note: For a population of size N containing
M defectives, p = M/N, and q = 1 – M/N.)
Probability Functions 67

defective, p < .10. The f-binomial is the standard p-binomial with the sampling proportion
f = n/N used as p and the number of defectives in the population Np used as the sample
size n. The frequency function then becomes
x Np - x
ænö æ nö
f ( x ) = CxNp ç ÷ ç 1 - ÷
èNø è Nø

Probabilities may be obtained using tables for the standard p-binomial with

n
p= n = Np
N

The rationale for the f-binomial approximation may be developed as follows:

CXNpCnN--xNp
f (x) =
CnN

= CxNp
( N - Np ) ! n !( N - n)!
( n - x ) ! ( N - Np - n + x ) ! N !
( N - n ) ( N - n - 1)¼( N - n - ( Np - x ) + 1) n !
= CxNp
N ( N - 1)¼( N - Np + 1) ( n - x )!
æ N - n N - n - 1 N - n - ( Np - x ) + 1 ö æ n n -1 n- x +1 ö
= CxNp ç  ÷ ç  ÷
ç N
è N -1 N - ( Np - x ) + 1 ÷ø çè N - ( Np - x ) N - ( Np - x ) - 1 N - Np + 1 ÷ø

since it can be shown that for b > a

a a -1
>
b b -1

Substitute the first ratio for each succeeding ratio in the first brackets and n/N for each ratio
in the second brackets to obtain
Np - x x
æ N-nö ænö
f ( x ) < CxNp ç ÷ ç ÷
è N ø èNø

or
Np - x x
æ nö ænö
f ( x ) < CxNp ç 1 - ÷ ç ÷
è Nø èNø

which is the f-binomial.

Note that the approximation could be improved by substituting the ratio n/(N − Np) in
the second brackets, but the binomial tables could no longer be used in determining the
relevant probabilities.
For proportion defective greater than a tenth when the sampling proportion is greater
than 10% of the population, the hypergeometric itself should be used.
68 Acceptance Sampling in Quality Control

It has been pointed out by Guenther (1973) that the Wise (1954) approximation can be
used effectively with binomial tables in the derivation of hypergeometric sampling plans.
This approximation to the hypergeometric consists of using the cumulative binomial dis-
tribution with

2Np - x
p=
2N - n + 1

to come very close to the hypergeometric values. Details of its use in the development of a
sampling plan will be found in Guenther (1977). Another excellent approximation is that
of Sandiford (1960).
In turn, the binomial distribution may be approximated by the Poisson distribution for
p small and n large (roughly when the product np is less than 5). This is done by looking
up Poisson probabilities of x successes when the mean of the Poisson distribution is μ = np.
When the product np is greater than 5, the binomial distribution may be approximated
by the normal distribution with

m = np, s = npq

Note that nq must also be greater than 5. However, in acceptance sampling, it is usually
the case that q > p. Here, the normal cumulative probability is taken over a region cor-
responding to the number of successes desired. In approximating a discrete distribution,
such as the binomial, with a continuous distribution, such as the normal, it is necessary
to use a “continuity” correction. Since the probability of a point in a continuous distribu-
tion is zero, it is necessary to approximate each discrete number of successes by a band on
the x-axis going out from the number one-half units on each side as shown in Figure 3.10.
Thus, the probability of x or less successes would be found as the area up to x + 1/2 under
the normal curve. The probability of x or more successes would be the area above x −1/2
and so on.
To illustrate these approximations, let us take a case where the sampling proportion is
equal to a tenth. Suppose the lot size is 100, sample size is 10, p = 0.1, and we desire the
probability of 2 or fewer defectives. Using appropriate formulas, tables, or a computer,
we get

Hypergeometric: Owen (N = 100, n = 10, k = 1, x = 2)

F ( 2 ) = .93998

Probability
of exactly
x successes

x
x – 1/2 x + 1/2

FIGURE 3.10 
Continuity correction.
Probability Functions 69

Binomial: Harvard (n = 10, p = .10, x = 2)

F ( 2 ) = .92981

Poisson: Molina (np = 10(.1) = 1, x = 2)

F ( 2 ) = .91970

f-Binomial:
2

F ( 2) = åC (.1) (.9 )
10 i 10 - i
i
i =0

Harvard (n = 10, p = .10, x = 2)

F ( 2 ) = .92981

Normal:

m = np = 10 ( .1) = 1
s = npq = 10. ( .1)( .9 ) = .95
2.5 - 1
z= = 1.58
.95
F ( 2 ) = .9428

All the approximations were fairly close to the hypergeometric value. Actually, the normal
was used for illustrative purposes only since it should not usually be used to approx-
imate the binomial when np is less than 5. This is an indication of the utility of these
approximations.

Tests of Fit
It is not enough to assume a distribution to hold in real-life applications of statistics.
Statistics is not like mathematics where correct answers are derived from the assump-
tions. In statistics, the assumptions must be correct and must describe the physical situa-
tion adequately before correct answers will be obtained. For this reason, it is not enough to
assume a distribution holds to be correct. Enough real data should be analyzed to assume
the assumption is correct.
Frequently, the underlying distributions of measurements are characterized by prob-
ability models. In these cases, it is necessary to assure that the data conform to the model
used. Methods have been developed to test if particular distributions are applicable.
These include probability plots, χ2 tests of goodness of fit, the Kolmogorov–Smirnov test,
and the Wilk–Shapiro test, and others. Sample size considerations are quite important in
70 Acceptance Sampling in Quality Control

99

95
90

80

50
P

20
10
5

2σ
1

620 630 640 650 660 670 680 x

L µ U

FIGURE 3.11 
Probability plot.

acceptance sampling since large sample sizes are needed to detect aberrations in the tails
of the distribution where the defective material is likely to be found. It is important that
those applying acceptance quality control procedures be familiar with these tests and pro-
cedures and they should apply them to real data before assuming any distribution shape
applies. They are discussed in most basic texts on applied statistics.
The probability plot is one of the most useful and versatile of the tests of fit (see
Figure  3.11). It involves plotting the ordered observations from a sample on a special
paper against the cumulative percentage at which the individual ordered observations
stand in the sample. In this way, an empirical cumulative probability distribution plot
for the sample is obtained. Estimates can be made from this plot and its shape can be
used as an indication of the underlying probability distribution, which gave rise to the
sample. Special probability papers transform the axis representing cumulative percent-
age in such a way that if the sample came from the distribution represented by the paper
selected, the points will plot roughly in a straight line. The papers can be obtained to
represent a variety of distributions, the normal and Weibull probably being the most
common. Directions for the construction of a normal probability paper have been given
by Nelson (1976).
Plotting positions can readily be determined using the formula
i - 1/2
P̂( i ) = (100 )
n

which gives the approximate probability of obtaining values less than x(i). Then, the
individual ordered points x(i) are plotted against their empirical cumulative frequency
Probability Functions 71

(usually in percent) estimated by Pˆ( i ). A straight-line plot is an indication that the under-
lying ­distribution of measurements is that of the paper on which the points are plotted.
Substantial departures from a straight-line plot indicate that the distribution for the paper
may not apply. A straight-line fit through a straight-line plot can be used to make estimates
of the parameters of the underlying distribution. On the normal probability paper, the mean
is estimated from the 50th percentile of the empirical plot. Similarly, the standard deviation
can be obtained as half the difference between the 16th and the 84th percentile values.
For example, consider the following data taken from MIL-STD-414. The specification for
electrical resistance of a certain electrical component is 650.0 ± 30 Ω. Suppose the values
of sample resistance in a sample of 10 are as follows: 643, 651, 619, 627, 658, 670, 673, 641,
638, and 680 Ω. A probability plot for these data appears in Figure 3.11, which plots the
­following points:

Order(i) x(i) Pˆ( i )

1 619 5
2 627 15
3 638 25
4 641 35
5 643 45
6 650 55
7 651 65
8 658 75
9 670 85
10 673 95

A straight-line plot is obtained. A line drawn through the points allows the following
­estimates to be made (on a plot having a more detailed resistance scale):
m̂ is the mean, 647.5
ŝ is the standard deviation, 17.5
PL is the percent below lower spec, 6%
PU is the percent above upper spec, 3%
PT is the total out of spec, 9%
Actually, for these data

x = 647.0 s = 17.2

and so the probability plot provided very good estimates of the population parameters in
this case, close to those obtained by the usual computational methods. Using x and s to
estimate the percent out of specification limits

U - x 680 - 647
zU = = = 1.92
s 17.2

giving 2.74% above the upper limit and defining zL as in MIL-STD-414

x - L 648 - 620
zL = = = 1.57
s 17.2
72 Acceptance Sampling in Quality Control

giving 5.82% below the lower limit. This gives a point estimate of 8.56% out of specifica-
tions, which is very close to that obtained from the probability plot.
The Weibull distribution is particularly useful in reliability analysis and associated
sampling plans. Weibull (1951) first used plots of the Weibull distribution. Later versions
and refinements in analysis were developed by Kao (1959), Nelson (1967), and Nelson and
Thompson (1971). The reader is referred to these papers for a discussion of the Weibull
probability plot and its uses. An extensive book on probability papers has been pre-
pared by King (1971). Shapiro (1980) has prepared an in-depth manual on testing normal-
ity and other distributional assumptions for the American Society for Quality Control.
An ­excellent introductory text on probability plots has been written by Nelson (1979) and
appears in the same series.

Software Applications
Excel
Excel provides users with some useful probability distribution functions that are appli-
cable to acceptance sampling problems. They are discussed here in the order they were
introduced in this chapter, along with a corresponding example:
• Hypergeometric
• Hypgeom.dist(x,n,Np,N,cumulative?)
where
x is the # occurrences in the sample (x = 0, 1, 2, …, n)
n is the sample size (n = 0, 1, 2, …, N)
N is the lot size (N > 0)
Np is the # occurrences in the lot

Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Probability of a four-card hand of aces (N = 52, p = 4/52, n = 4, x = 4)
f(4) = Hypgeom.dist(4,4,4,52,FALSE) = 3.69379E-06 (= 1/270725)
• Binomial
• Binom.dist(x,n,p,cumulative?)
where
x is the # occurrences in the sample (x = 0, 1, 2, …, n)
n is the sample size (n > 0)
p is the proportion defective (0 ≤ p ≤ 1)

Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Sampling plan (x = 0, n = 5, c = 0, p = .01)
f(0) = Binom.dist(0,5,.01,FALSE) = 0.95099005 ~ 0.951
Probability Functions 73

• Poisson
• Poisson.dist(x, μ, cumulative?)
where
x is the # occurrences in the sample (x = 0, 1, 2, …)
μ is the mean number of defects (μ > 0)

Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Bubbles in imported glassware (x = 1, μ = 2)
f(1) = Poisson.dist(1,2,FALSE) = 0.270670566 ~ 0.2707
• f-Binomial
• Excel doesn’t have a function for this distribution, but we can use the relations
to the binomial to solve for these probabilities using the binomial function
• Binom.dist(x,D,n/N,cumulative?)
where
x is the # occurrences in the sample (x = 0, 1, 2, …, n)
D is the number of defects in the lot (D ≥ 0)
n is the sample size (n > 0)
N is the lot size (N > 0)
Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Sampling plan (x = 1, D = 2, n = 4, N = 10)
f(1) = Binom.dist(1,2,.4,FALSE) = 0.48
• Negative binomial
• Negbinom.dist(n,1,p,cumulative?)
where
n is the number of trials up to and including the xth defective
p is the proportion defective (0 ≤ p ≤ 1)
q is the proportion effective, or good (1 − p)
x is the # occurrences in the sample (x = 0, 1, 2, …)
Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Sampling plan where the first success occurs with the first defective
found on the second sample (x = 2, p = .01)
f(2) = Negbinom.dist(2,1,.01,FALSE) = 0.009801 ~ 0.0098
• Exponential
• Expon.dist(x,λ,cumulative?)
where
x is the measurement distributed (x > 0)
μ is the mean of the distribution (μ > 0)
λ is the 1/μ,which is the failure rate
74 Acceptance Sampling in Quality Control

Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: A probability power transistor will fail at 500 h if it comes from a pro-
cess with a mean life of 5000 h (x = 500, μ = 5000)
F(500) = Expon.dist(500,1/5000,TRUE) = 0.095162582 ~ 0.095
• Weibull
• Weibull.dist(x,η,β,cumulative?) [NOTE: Excel does not consider γ > 0.]
where
x is the measurement distributed (x > 0)
η = α = scale parameter (η > 0) [NOTE: Excel calls this parameter β]
β is the shape parameter (β > 0) [NOTE: Excel calls this parameter α]

Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: A probability power transistor will fail at 500 h if it comes from a pro-
cess with a mean life of 5000 h (x = 500, η = 5000, β = 1 [exponential], γ = 0).
F(500) = Weibull.dist(500,1,5000,TRUE) = 0.095162582 ~ 0.095
• Normal
• Norm.dist(x,μ,σ,cumulative?)
where
x is the measurement distributed (−∞ < x > ∞)
μ is the mean (−∞ < μ > ∞)
σ is the standard deviation (σ > 0)
Cumulative is a logical value that determines the form of the function. If cumu-
lative is TRUE, then the function returns the cumulative distribution function;
if FALSE, it returns the probability mass function.
• Example: Proportion of bolts with lengths less than 49.8 mm manufactured by a
process having a mean of 50 mm and a standard deviation of 0.1 mm (x = 49.8,
μ = 50, σ = 0.1)
F(500) = Norm.dist(49.8,50,0.1,TRUE) = 0.022750132 ~ 0.0228

Minitab
As one would expect, Minitab can provide a variety of probability values for a wide range
of distributions, including all of those discussed in this chapter (except for the f-binomial).
Again, they are discussed here in the order they were introduced in this chapter, along
with the corresponding example used in Excel in the preceding section:
• Hypergeometric
Choose Calc > Probability Distributions > Hypergeometric and complete the
­dialog box as shown in Figure 3.12.
Probability Density Function
Hypergeometric with N = 52, M = 4, and n = 4
x P(X = x)
0.0000037
Probability Functions 75

FIGURE 3.12 
Minitab dialog box to get a hypergeometric probability.

• Binomial
Choose Calc > Probability Distributions > Binomial and complete the dialog box
as shown in Figure 3.13.
Probability Density Function
Binomial with n = 5 and p = 0.01
x P(X = x)
0 0.950990

• Poisson
Choose Calc > Probability Distributions > Poisson and complete the dialog box as
shown in Figure 3.14.
Probability Density Function
Poisson with mean = 2
x P(X = x)
1 0.270671

• Negative binomial
Choose Calc > Probability Distributions > Negative Binomial and complete the
dialog box as shown in Figure 3.15.
Probability Density Function
Negative binomial with p = 0.01 and r = 1
x P(X = x)
2 0.0099
* NOTE * X = total number of trials.
76 Acceptance Sampling in Quality Control

FIGURE 3.13 
Minitab dialog box to get a binomial probability.

FIGURE 3.14 
Minitab dialog box to get a Poisson probability.

• Exponential
Choose Calc > Probability Distributions > Exponential and complete the dialog
box as shown in Figure 3.16.
Probability Density Function
Exponential with mean = 5000
x P(X ≤ x)
500 0.0951626
Probability Functions 77

FIGURE 3.15 
Minitab dialog box to get a negative binomial probability.

FIGURE 3.16 
Minitab dialog box to get an exponential probability.

• Weibull
Choose Calc > Probability Distributions > Weibull and complete the dialog box as
shown in Figure 3.17.
Probability Density Function
Weibull with shape = 1 and scale = 5000
x P(X ≤ x)
500 0.0951626
78 Acceptance Sampling in Quality Control

FIGURE 3.17 
Minitab dialog box to get a Weibull probability.

• Normal
Choose Calc > Probability Distributions > Normal and complete the dialog box as
shown in Figure 3.18.
Probability Density Function
Normal with mean = 50 and standard deviation = 0.1
x P(X ≤ x)
49.8 0.0227501

FIGURE 3.18 
Minitab dialog box to get a normal probability.
Probability Functions 79

Problems
1. A trial lot of 100 candles is received from a new supplier. Five candles are checked to
be sure that the wick extends properly above the body. One is found to be defective.
What is the probability of 1 or fewer defective candles in a sample of 5 if 2 candles
were defective in the lot? How many would you expect? What would be the standard
deviation of the number observed in a sample of 5?
2. If the vendor in Problem 1 maintains a quality of 2% defective, what is the probabil-
ity of 1 or fewer defective candles in a sample of 5 from a very large lot? How many
would you expect? What would be the standard deviation of the number observed in
a sample of 5? Check your answer with the Larson nomograph.
3. A spot welder is expected to produce not more than 2 defective welds in a shift’s pro-
duction. If the process average were actually 2 defective welds per shift, what is the
probability of obtaining 2 or fewer bad welds on a given shift? Check your answer
with the Thorndyke chart. Why is the answer not one-half? What would be the stan-
dard deviation of the number of bad welds per shift?
4. A continuous sampling plan starts by inspecting i successive units. If no defectives
are found, a switch to sampling inspection is made. If i = 5, what is the probability of
finding a defective on the fifth trial if the proportion defective submitted to the plan
is .05? What is the mean number of trials to the second defective?
5. The life of a transistor follows the exponential distribution with a mean life
μ = 10,000 h. What is the probability of a unit failing before 20,000 h? What is the
standard deviation of the transistor’s lifetime?
6. Express Problem 5 in terms of the parameters of the Weibull distribution. What is the
probability of a lifetime less than or equal to 10,000 h?
7. Bottles are to be filled with 1 L of liquid. The amount of fill is normally distributed
with standard deviation σ = 0.01 L. The mean fill is set at μ = 1.03 L to minimize the
possibility of underfill. To check on the overfill, the contents are poured into a con-
tainer marked with a “narrow limit” at 1.005 L. What is the probability of observing
a fill less than 1.005 L when the process mean is actually at 1.03 L? What is the prob-
ability of one such indication in a sample of 3?
8. A new supplier submits a test lot of rods whose length is specified to average 3 ± .001 m.
The rods are to be welded together so that average length is important. The distribu-
tion of lengths is unknown, but a sample of 9 rods yields a mean X = 3.001 m. Is such
a result likely if the standard deviation of this type of product is σ = .0003 m and the
mean is 3 m?
9. Suppose samples of 5 are to be taken from a lot of 25 for a simple nondestruc-
tive test. If the fraction defective is 0.08, what approximation is appropriate for the
hypergeometric distribution? What is the probability of 1 or fewer defectives in a
sample of 5?
10. A probability plot is to be made of the weight of 500 pieces to check for normality.
The 80th ordered observation is 24 while observation 420 is 48. If the fitted line
passes through both these points, estimate the standard deviation. Estimate the
mean.
80 Acceptance Sampling in Quality Control

References
Beyer, W. H., 1968, Handbook of Tables for Probability and Statistics, 2nd ed., CRC Press, Cleveland, OH.
Burington, R. S. and D. C. May, 1970, Handbook of Probability and Statistics with Tables, 2nd ed.,
McGraw-Hill, New York.
Burr, I. W., 1953, Engineering Statistics and Quality Control, McGraw-Hill, New York.
Defense Systems Department, General Electric Company, 1962, Tables of Individual and Cumulative
Terms of the Poisson Distribution, Van Nostrand, Princeton, NJ.
Dodge, H. F. and H. G. Romig, 1941, Single sampling and double sampling inspection tables, The Bell
System Technical Journal, 20(1): 1–61.
Dodge, H. F. and H. G. Romig, 1959, Sampling Inspection Tables: Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York, p. 35.
Guenther, W. C., 1973, A sample size formula for the hypergeometric, Journal of Quality Technology,
5(4): 167–173.
Guenther, W. C., 1977, Sampling Inspection in Statistical Quality Control, Macmillan, New York.
Harvard University Computing Laboratory, 1955, Tables of the Cumulative Binomial Probability
Distribution, Harvard University Press, Cambridge, MA.
Kao, J. H. K., 1959, A graphical estimation of mixed Weibull parameters in the life testing election
tubes, Technometrics, 1(4): 389–407.
King, J. R., 1971, Probability Charts for Decision Making, Industrial Press, New York.
Larson, H. R., 1966, A nomograph of the cumulative binomial distribution, Industrial Quality Control,
23(6): 270–278.
Lieberman, G. J. and D. B. Owen, 1961, Tables of the Hypergeometric Probability Distribution, Stanford
University Press, Stanford, CA.
Molina, E. C., 1942, Poisson’s Exponential Binomial Limit, Van Nostrand, New York.
Mood, A. M. and F. A. Graybill, 1973, Introduction to the Theory of Statistics, 3rd ed., McGraw-Hill,
New York.
Nelson, L. S., 1967, Weibull probability paper, Industrial Quality Control, 23(9): 452–453.
Nelson, L. S., 1974, Using tables of the cumulative binomial distribution, Journal of Quality Technology,
6(2): 116–118 [Addendum, 7(1): 49 (1975)].
Nelson, L. S., 1976, Constructing normal probability paper, Journal of Quality Technology, 8(1): 56–57.
Nelson, W. B., 1979, How to analyze data with simple plots, Vol. 1, in The ASQC Basic References
in Quality Control: Statistical Techniques (E.J. Dudewicz, Ed.), American Society for Quality
Control, Milwaukee, WI.
Nelson, W. B. and V. C. Thompson, 1971, Weibull probability papers, Journal of Quality Technology,
3(2): 45–50.
Odeh, R. E., D. B. Owen, Z. W. Birnbaum, and L. Fisher, 1977, Pocket Book of Statistical Tables, Marcel
Dekker, New York.
Owen, D. B., 1962, Handbook of Statistical Tables, Addison-Wesley, Reading, MA.
Romig, H. G., 1953, 50–100 Binomial Tables, John Wiley & Sons, New York.
Sandiford, P. J., 1960, A new binomial approximation for use in sampling from finite populations,
Journal of the American Statistical Association, 55(292): 718–722.
Schilling, E. G., 2005, Average run length and the OC curve of sampling plans, Quality Engineering,
17(3): 399–404.
Schilling, E. G. and P. R. Nelson, 1976, The effect of non-normality on the control limits of X charts,
Journal of Quality Technology, 8(4): 183–188.
Shapiro, S. S., 1980, How to test normality and other distributional assumptions, Vol. 4, in The ASQC
Basic References in Quality Control Statistical Techniques (E.J. Dudewicz, Ed.), American Society
for Quality control, Milwaukee, WI.
Shewhart, W. A., 1931, Economic Control of Quality of Manufactured Product, Van Nostrand, New York.
Probability Functions 81

Thorndyke, F., 1926, Applications of Poisson’s probability summation, The Bell System Technical
Journal, 5: 604–624.
United States Department of Commerce, 1950, Tables of the Binomial Probability Distribution, National
Bureau of Standards, Applied Mathematics Series No. 6, US Government Printing Office,
Washington, DC.
United States Department of Commerce, 1953, Tables of the Normal Probability Functions, National
Bureau of Standards, Applied Mathematics Series No. 23, US Government Printing Office,
Washington, DC.
United States Department of Commerce, 1964, Handbook of Mathematical Functions, National Bureau of
Standards, Applied Mathematics Series No. 55, US Government Printing Office, Washington,
DC.
Weibull, W., 1951, A statistical distribution function of wide applicability, Journal of Applied Mechanics,
18: 293–297.
Williamson, E. and M. H. Bretherton, 1963, Tables of the Negative Binomial Probability Distribution,
John Wiley & Sons, New York.
Wise, M. E., 1954, A quickly convergent expansion for cumulative hypergeometric probabilities,
direct and inverse, Biometrika, 41(3): 317–329.
4
Concepts and Terminology

The fundamental tool for analysis of a sampling plan is the operating characteristic (OC)
curve. Two types of curves are recognized:

1.
Type A: Sampling from an individual (or isolated) lot, showing probability that the
lot will be accepted plotted against lot proportion defective.
2.
Type B: Sampling from a process (such as the producer’s process, which produced
the lot), showing the proportion of lots that will be accepted plotted against pro-
cess proportion defective.

Naturally, the probability distributions utilized in plotting these types of OC curves are
inherently different. They also depend upon the measure in which quality is expressed.
These include

• Attributes: A dichotomous (two-class) classification of units into defective and non-

defective, for example, the number of defective units in a sample of 100 units
• Counting: An enumeration of occurrences of a given characteristic per given number
of units counted, for example, the number of defects per 100 units in the population
• Variables: The measurement of some characteristic along a continuous scale, for
example, the diameter of a circular casting as measured in centimeters

The distinction is made between defect (an imperfection great enough to be counted) and
defective (a unit containing one or more defects, which could be rejected for any one of them).
The probability distributions appropriate for the derivation of OC curves of the two
types are shown in Table 4.1. The form of these distributions and their properties are
shown in Table 3.1.
For variables data, the applicable distribution is that of the variable as it would appear to
the inspector, that is, including piece-to-piece variation, measurement error, and changes
in environmental conditions. There are means available for separating these sources of
error and controlling them. Such methods are addressed in texts on design of experiments,
such as Hicks (1999) or Anderson and McLean (1974), and in texts on process quality con-
trol, such as Ott et al. (2005).
Differences in the OC curves associated with these different type plans may be i­ llustrated
by the sampling plan

N is the lot size, which is 20.

n is the sample size, which is 10.
c is the acceptance number , which is 1.

Using the methods in Chapter 2, it is possible to compute the different OC curves. The
results are shown in Table 4.2.

83
84 Acceptance Sampling in Quality Control

TABLE 4.1
Probability Distributions for Operating Characteristic Curves
Characteristic Type A Type B

Attribute Hypergeometric Binomial

Count f-binomial Poisson
Measurement Applicable continuous distribution
of measurement involved

TABLE 4.2
Probabilities of Acceptance for Hypergeometric, f-Binomial, Binomial, and Poisson
(N = 20, n = 10, c = 1)
Percent Defective Type A Hypergeometric Type A f-Binomial Type B Binomial Type B Poisson

5 1.000 1.000 0.914 0.910

10 0.763 0.750 0.736 0.736
15 0.500 0.500 0.544 0.558
20 0.291 0.312 0.376 0.406
25 0.152 0.188 0.244 0.287
30 0.070 0.109 0.149 0.199
35 0.029 0.062 0.086 0.136
40 0.010 0.035 0.046 0.092
45 0.003 0.020 0.023 0.061
50 0.001 0.011 0.011 0.040
55 0.005 0.027
60 0.002 0.017
65 0.001 0.011
70 0.007

We see from the Type A values in Table 4.2 that with the hypergeometric and f-binomial
sampling plans n = 10, c = 1, it is impossible to fail if the lot is 5% defective or 5 defects per
100 units when the lot size is 20. Why? Because for the lot of 20 to be 5% defective it would
contain just one defective, and one defective is allowable under the plan. In fact, for a finite
lot size, only a limited number of percent defective or defects per 100 units can be formed,
in this case 0, 5, 10, 15, …, 95, 100.
The Type B OC curve is not so restricted. In fact, the producer’s process could have
been running at any percent defective when the lot of 20 was formed. The Type B OC
curve views the lot of 20 as a sample from the producer’s process and the sample of 10 as
a subsample of the same process. In this way, it is reasonable to address the probability of
acceptance for any percent defective from 0 to 100 when using a Type B OC curve.
Finally, if the number of defects is counted in a lot of 20 items the count could easily exceed
20 since one item can have one or more defects. Note that this is not the case for either of
the Type A sampling situations that deal with defectives or with defects. Neither the num-
ber of defectives could exceed 20, the lot size, nor could the number of defects exceed some
finite number. A count of defects is often expressed in terms of “defects per 100 units.” In this
form, the measure of quality is analogous to “percent defective,” which is also based on 100.
However, as noted, defects per 100 units may exceed 100 in Type B situations. For instance, if
we knew the lot of 20 had 3 defective pieces in it with 4, 10, and 12 defects each, the lot would,
Concepts and Terminology 85

in total, contain 26 defects—even though it composed of only 20 pieces and had only 3 defec-
tives. The mean number of defects per unit in such a situation would be 1.3. The mean num-
ber of defects per 10 pieces would be 13, since a sample size of 10 was specified. The Poisson
distribution with a mean μ = 13 would be used to calculate the probability of acceptance in
such a case. It should be pointed out that the defects per unit probabilities of acceptance in
Table 4.2 were computed using a mean value of
p ´ 10
m=
100
since the value p is interpreted as defects per 100 units so that p = 5 defects per 100 units
implies p/100 = .05 defects per unit, which gives
10 p
= 0.5 defects per 10 units
100
A plot of the four OC curves is given in Figure 4.1, identified by the first letter of their names.

1.00

Hypergeometric
0.90 f-Binomial
Binomial
Poisson

0.80

0.70

0.60

0.50
Pa

0.40

0.30

0.20

0.10

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50
p

FIGURE 4.1 
Types A, B, and defects OC curves.
86 Acceptance Sampling in Quality Control

Note how the curves are fairly well superimposed for small p and then diverge as p
becomes large. This shows the use of these distributions as approximations when p is
small. Also the shape of the Type A (hypergeometric and f-binomial) curves is quite differ-
ent from that of the Type B (binomial and Poisson) curves, since the sample represents a
large proportion (50%) of the lot. These curves also illustrate the conservative nature of the
approximations since they tend to underestimate for high probability of acceptance and
overestimate for low probability of acceptance.

Average Run Length of Type B Plans

Average run length (ARL) has been used extensively to describe the effectiveness of ­process
control procedures. It gives an indication of the expected number of samples until a
­decision is made. Schilling (2005) has proposed that this measure is a natural descriptor of
Type B sampling plans as well. The OC curve is easily augmented by the addition of ARL
values corresponding to the selected “percent lots accepted” shown on the y-axis or by
construction of an ARL axis parallel to it (Figure 4.2).
ARL is easily calculated as a function of probability of acceptance (Pa) using the mean of
the geometric distribution of run length:

1
ARL =
1 - Pa

Once calculated, these values can then be exhibited in a table of the corresponding values
of Pa and ARL, such as Table 4.3 provided by Schilling (2005), or from few key values com-
mitted to memory. Table 4.4 is such a table. Note that it applies to OC curves of process
control procedures as well.

100.0 14
Percent lots accepted, Pa
90.0
Average run length (ARL) 12
80.0
Percent lots accepted Pa (%)

Average run length (ARL)

70.0 10

60.0
8
50.0
6
40.0

30.0 4
20.0
2
10.0

0.0 0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Proportion defective (p)

FIGURE 4.2 
OC and ARL curves for Type B plan (n = 0, c = 1 using binomial distribution).
Concepts and Terminology 87

TABLE 4.3
Conversion of Pa to Average Run Length
Pa (%) ARL

99 100.00
95 20.00
90 10.00
80 5.00
70 3.33
60 2.50
50 2.00
40 1.67
30 1.43
20 1.25
10 1.11
5 1.05
1 1.01

TABLE 4.4
Type A Probability of Acceptance Using
the Hypergeometric Distribution (n = 10, c = 1)
% N = 20 N = 60 N = 100 N=∞

 5 1.000 0.931 0.923 0.914

10 0.763 0.741 0.738 0.736
15 0.500 0.533 0.538 0.544
20 0.291 0.354 0.363 0.376
25 0.152 0.219 0.229 0.244
30 0.070 0.126 0.136 0.149
35 0.029 0.067 0.075 0.086
40 0.010 0.033 0.039 0.046

The addition of ARL values to Type B OC curves should help distinguish between Type
A and Type B applications, since the presence of ARL values in Type B OC curves empha-
sizes the unique nature of such plans when applied to a series of lots. This should contrib-
ute greatly to an understanding of these considerations in the selection of a sampling plan.

Sample Size and Lot Size

Reference to the formulas for the binomial, the Poisson, and the various continuous distri-
butions will indicate that they do not contain any reference to lot size. Only in the hyper-
geometric and f-binomial distributions will such a parameter be found. Even then, the
effect of lot size on the OC curve is minimal when a small proportion (say, less than 10%)
of the lot is used up in taking the sample. This is illustrated in Table 4.3, which shows the
probability of acceptance for the plan n = 10, c = 1 when the lot size is 20, 60, 100, and ∞.
88 Acceptance Sampling in Quality Control

A plot of the OC curves would show little difference among the curves except when
N = 20. It is apparent that the probabilities change substantially with lot size only when the
sample represents a large portion of the lot (say, from 1/2 to 1/6). Changes are slight when
the sample is a small fraction of the lot (for 1/10 and less). Also note that the values for
N = ∞, calculated from the binomial distribution, are conservative in approximating the
hypergeometric for smaller lot sizes in that they underestimate Pa for lower percent defec-
tive and overestimate for high percent defective. Thus, use of the binomial distribution in
constructing a Type A OC curve when the fraction of the lot sampled is reasonably low
will not only give fairly close answers but tends to be conservative as well. Similar results
can be obtained for the Poisson approximation to the binomial distribution and f-binomial
approximation to the hypergeometric distribution.
As might be expected, changes in sample size for a given lot size will, however, have a
substantial effect on the protection afforded by a plan. Table 4.5 shows the probability of
acceptance for plans with sample sizes 5, 10, and 15 from a lot of size 20 with acceptance
number c = 1 for Type A (hypergeometric) probabilities.
The effect of sample size is somewhat less pronounced for Type B (binomial) probabili-
ties; however, Table 4.6 gives an example similar to Table 4.4 but with infinite lot size.
The most dramatic effect on the probability of acceptance, however, comes with chang-
ing acceptance numbers. Even the inherent shape of the OC curve is changed in going from

TABLE 4.5
Type A Probability of Acceptance Using
the Hypergeometric Distribution (N = 20, c = 1)
% n=5 n = 10 n = 15

5 1.000 1.000 1.000

10 0.947 0.763 0.447
15 0.860 0.500 0.140
20 0.751 0.291 0.032
25 0.634 0.152 0.005
30 0.517 0.070
35 0.406 0.029
40 0.307 0.010

TABLE 4.6
Type B Probability of Acceptance Using
the Binomial Distribution (c = 1)
% n=5 n = 10 n = 15
5 0.977 0.914 0.829
10 0.919 0.736 0.549
15 0.835 0.544 0.319
20 0.737 0.376 0.167
25 0.633 0.244 0.080
30 0.528 0.149 0.035
35 0.428 0.086 0.014
40 0.337 0.046 0.005
Concepts and Terminology 89

TABLE 4.7
Type A Probability of Acceptance Using
the Hypergeometric Distribution (N = 20, n = 10)
% c=0 c=1 c=2

5 0.500 1.000 1.000

10 0.237 0.763 1.000
15 0.105 0.500 0.895
20 0.043 0.291 0.709
25 0.016 0.152 0.500
30 0.005 0.070 0.314
35 0.002 0.029 0.175
40 0.010 0.085

one acceptance number to another. This can be seen in Table 4.7, which shows the effect of
changing the acceptance number for a plan with N = 20, n = 10.
It should be clear, then, that the two principal determinants of the OCs of a sampling plan
are acceptance number and sample size. Lot size plays a very minor role in determining
protection even when sampling sizable proportions of the lot. This is contrary to intuitive
belief and should be constantly borne in mind by the practicing quality control engineer in
setting up sampling plans and sampling schemes. Often a relationship of lot size to sample
size is specified (even by MIL-STD-105E), but this is for logistic and economic purposes
and not primarily for purposes of enhancing the protection afforded by the plan.
We may ask what is the effect of maintaining the acceptance number as a constant
­proportion of the sample size. Table 4.8 compares three plans that keep the acceptance num-
ber at 10% of the sample size. These plans are in no sense equivalent. The protection afforded
by n = 40, c = 4 is much higher than the other plans. This can be seen by comparing the plans’
protection at, say, 20% defective. An acceptance number of 4 can, in fact, give more protec-
tion than an acceptance number of 2 provided that the sample size is increased accordingly.
In some operations, it has become customary to specify sample size as a proportion of
the lot size. Take a 10% sample, let us say, usually with an acceptance number, c = 0. Since
the OCs of a plan are dependent principally on sample size, not lot size, this means that
large lots with large samples will be accepted much less often than small lots with small
sample sizes at the same percent defective.

TABLE 4.8
Type B Probability of Acceptance Using
the Binomial Distribution (c/n = .1)
% n = 10, c = 1 n = 20, c = 2 n = 40, c = 4

5 0.914 0.925 0.952

10 0.736 0.677 0.629
15 0.544 0.405 0.263
20 0.376 0.206 0.076
25 0.244 0.091 0.016
30 0.149 0.035 0.003
35 0.086 0.012
40 0.046 0.004
90 Acceptance Sampling in Quality Control

TABLE 4.9
Type A Probability of Acceptance Using the Hypergeometric Distribution
when Sample Size is Proportionate to Lot Size
% N = 20, n = 10, c = 0 N = 40, n = 20, c = 0 N = 100, n = 50, c = 0

5 0.500 0.244 0.028

10 0.237 0.053 0.001
15 0.105 0.010
20 0.043 0.002
25 0.016
30 0.005
35 0.002

TABLE 4.10
Effect of Lot Size on Acceptance (Order of 1000)
Proportion Lots Expected Number Expected Pieces
Lot Size Number Lots Accepted Lots Accepted Accepted

20 50 0.500 25 500
40 25 0.243 6.075 243
100 10 0.028 0.280 28

This is illustrated in Table 4.9 that shows the protection afforded by such a plan. For
lots of size 20, 5% defective material has a 50% chance to be accepted, while for lots of 100
only a 2.8% chance of acceptance is provided. An unscrupulous supplier has an incentive
to provide small lots as can be seen in Table 4.10, which shows the results of shipping 5%
defective material in different lot sizes.
The protection afforded both parties by a plan such as this is clearly dependent upon lot
size and is not a rationally determined criterion for protection.

Effect of Inspection Error

No inspection is perfect all the time. Indeed, it is generally recognized that 100% inspection
is much less than 100% effective in screening out defective items. Studies have indicated
that, in the face of monotony and fatigue, only about 80% of the defectives will be detected
(Juran and Gryna 1970). The reasons for inspection inaccuracy have been detailed by Juran
(1962, pp. 8–25) as follows:

1. Willful errors that include

a. Criminal acts such as fraud and collusion
b. Falsification for personal convenience of the inspector
2. Intermediate errors due to bias, rounding off, overzealousness, etc.
3. Involuntary errors due to blunder, fatigue, and other forms of human ­imperfection
Concepts and Terminology 91

TABLE 4.11
Proportions Defective Misclassified
Inspector Classification
Actual Condition Nondefective Defective

Nondefectives 1 – p1 p1
Defectives p2 1 – p2
Notes: p1, proportion nondefective classified as ­efectives;
p2, proportion defective classified as non­defectives.

In particular, flinching or failure to call a defect when it is close to the specification is a com-
mon source of error of what Juran calls the intermediate type.
It should be pointed out that errors can go either way. An overzealous inspector can eas-
ily flinch by calling a good product bad. Harsh supervision, the mood of the moment, and
the psychological and even physical environment can cause marginal and even less than
marginal decisions to be incorrectly made.
Sample inspection is also subject to the same type of inspection error. While an advan-
tage of sampling is a reduction of the number of pieces subject to repetitive examination,
the same circumstances and motivations exist, which may lead to inspector inaccuracy.
The result is an inaccurate representation of the quality submitted.
Suppose a product is submitted that is of fraction defective p. The inspector misclassifies
the product as shown in Table 4.11.
The Statistical Research Group, Columbia University (1948, p. 23), presents the following
formula for the apparent level of quality p* when the true incoming level defective is p:

p* = p1 ( 1 - p ) + p ( 1 - p2 )

This follows from the rightmost column of Table 4.11. When the true fraction defective is
small and the proportion of defectives that are missed is not large, we have

p* = p1 + p

Similarly, when the chance of misclassification of a nondefective item is small,

p * = p ( 1 - p2 )

It is often the case that errors will go one way or the other, although both types of misclas-
sifications at the same time are possible. For example, suppose due to an error in configu-
ration control the inspector received a print of a symmetric part, which was reversed from
left to right. Unfortunately, the written material was on another sheet and the mistake went
undetected so that the area of acceptance became that of rejection and vice versa. Then,

p1 = 1, p2 = 1

and

p * = 1 ( 1 - p ) + p ( 1 - 1) = 1 - p

That is, the apparent level of defective material would be the actual proportion nondefective.
92 Acceptance Sampling in Quality Control

The formula works equally well for screening or sampling inspection. In screening, it gives
the apparent level of quality after 100% inspection. In sampling, it gives the apparent level of
quality as seen by the inspector. The OC curve can be entered in terms of p* rather than p to find
the probability of acceptance in the face of inspection error. Unfortunately, p1 and p2 are rarely
known but provide a means for analysis of the possible effect of this type of error. Methods of
estimating inspector bias in visual inspection have been discussed by Schilling (1961).

Rectification
Much of the effect of the imposition of a sampling plan depends upon the disposition of
the product after it is inspected. Accepted lots go to the consumer. Rejected lots may be
handled in a number of ways as follows:

Destroyed: There will be no effect on overall quality if the producer continues to sub-
mit at a constant level of quality and a positive effect if quality levels fluctuate
nonrandomly from lot to lot.
Resubmitted: There will be no effect on the overall quality if the producer continues to
submit at a constant level of quality.
Screened: The quality of rejected lots improved within the limits of inspection error.
Properly done 100% inspection of rejected lots would transform each rejected lot
into a perfect one. As a result the overall level of quality as seen by the consumer
would improve.

Acceptance sampling schemes that incorporate 100% inspection of rejected lots are called “rec-
tification” schemes. Formulas are available for calculating the average outgoing quality (AOQ)
from such schemes. This is the long-run average quality shipped to the consumer under 100%
inspection of rejected lots, assuming any defective item found is replaced by a good one. The
average is taken over all lots, good and bad, so that assuming no inspection error,
æ nö
AOQ = pPa ç 1 - ÷
è N ø
since the only defectives transmitted to the consumer would be in the accepted lots
(rejected lots having been made perfect). The average proportion defective the consumer
would receive then is made up of fraction defective p that received a proportion Pa of the
time and fraction defective 0 that received a proportion 1 − Pa of the time. But for all lots,
defective items found in the sampling inspection are also replaced by good ones so that the
remaining proportion defective is
æ N-nö
pç ÷
è N ø
and
æ N-nö æ N-nö
AOQ = pPa ç ÷ + 0 ( 1 - Pa ) ç ÷
è N ø è N ø
æ N -nö
= pPa ç ÷
è N ø
æ nö
= pPa ç 1 - ÷
è Nø
Concepts and Terminology 93

and when the sample size is very small in proportion to the lot size n/N ~ 0, so that the
formula becomes

AOQ = pPa

The maximum value of the AOQ over all possible values of fraction defective, which might
be submitted, is called the AOQ limit (AOQL). It represents the maximum long-term aver-
age fraction defective that the consumer can see under operation of the rectification plan.
It is sometimes necessary to determine the average amount of inspection per lot in the
application of such rectification schemes, including 100% inspection of rejected lots. This
average, called the average total inspection (ATI), is made up of the sample size n on every
lot plus the remaining (N − n) units on the rejected lots so that

ATI = n + ( 1 - Pa ) ( N - n )
= Pa n + ( 1 - Pa ) N

Consider the sampling plan n = 10, c = 1 used on a continuing supply of lots of size 20
from the same producer, that is, in a Type B sampling situation. Clearly, rectification plans
are meaningless on isolated lots, even though they might be 100% inspected if rejected,
because there is no long-term average involved. The Type B probabilities of acceptance
have already been calculated and are listed in Table 4.12, which shows the calculation of
the AOQ and the ATI. The operations involved are indicated in the last row.
It is apparent that the ATI curve starts at 10, the sample size, when p = 0 since no
lots are 100% inspected and rises to 20 when p = 1.0 since all lots will be rejected and 100%
inspected when the lots are completely defective. The ATI curve is shown in Figure 4.3.
The AOQ curve starts at 0 when p = 0 since no rectification is necessary. It rises to a maxi-
mum of around 4.1% defective and then declines as more and more 100% inspection takes
place. When lots are completely defective, they are all rectified and the AOQ is again zero.
The AOQL for this plan can be seen to be around 4.1% defective. We define pM as the
incoming defective at which the AOQ reaches its maximum, that is, the AOQL occurs
when the incoming fraction defective is pM. Then examining the region close to p = .15

TABLE 4.12
Calculation of Average Outgoing Quality and Average Total Inspection
(N = 20, n = 10, c = 1)
p Pa (1 – n/N) AOQ (1 – Pa)(N – n) ATI

0.00 1.000 0.5 0.000 0.00 10.00

0.05 0.914 0.5 0.023 0.86 10.86
0.10 0.736 0.5 0.037 2.64 12.64
0.15 0.544 0.5 0.041 4.56 14.56
0.20 0.376 0.5 0.038 6.24 16.24
0.25 0.244 0.5 0.030 7.56 17.56
0.30 0.149 0.5 0.022 8.51 18.51
0.35 0.086 0.5 0.015 9.14 19.14
0.40 0.046 0.5 0.009 9.54 19.54
0.45 0.023 0.5 0.005 9.77 19.77
0.50 0.011 0.5 0.003 9.89 19.89
(A) (B) (C) (D) = (A)(B)(C) (E) (F) = n + (E)
94 Acceptance Sampling in Quality Control

20

18

16

14

12
ATI
10

0
0 0.10 0.20 0.30 0.40 0.50
p

FIGURE 4.3 
ATI curve (N = 20, n = 10, c = 1).

TABLE 4.13
Determination of Average Outgoing Quality Limit
p Pa (1 – n/N) AOQ

0.13 0.620 0.5 0.0403

0.14 0.582 0.5 0.0407
0.15 0.544 0.5 0.0408
0.16 0.508 0.5 0.0406
0.17 0.473 0.5 0.0402
(A) (B) (C) (D) = (A)(B)(C)

as in Table 4.13, it is apparent that the AOQL is .041 to three-place accuracy and it occurs
at pM  = .15. The consumer will never experience a long-term average fraction defective
greater than .41, although the average may be considerably higher in the short run. The
AOQ curve is given in Figure 4.4.

Curtailment
Just as there are many procedures for disposing of a lot, there are different ways to treat the
sample itself. Consider the following possibilities for a single-sampling plan with sample
size n and acceptance number c:

1.
Complete inspection: All items in the sample of n are inspected.
2.
Semicurtailed inspection: The inspection is stopped when the number of defectives
found exceeds the acceptance number. All units are inspected if the lot is accepted.
Concepts and Terminology 95

0.05

AOQL = .041
0.04

0.03
AOQ

0.02

0.01 pM = 0.15

0
0 0.10 0.20 0.30 0.40 0.50
p

FIGURE 4.4 
AOQ curve (N = 20, n = 10, c = 1).

3.
Fully curtailed inspection: The inspection is stopped when the number of defectives
found exceeds the acceptance number c or the number of nondefectives is found
to exceed n − c. In short, the inspection is stopped once a decision can be made.

Under curtailment, the number of units actually inspected becomes a random variable. There
are formulas that can be used to determine the average sample number (ASN) for such proce-
dures. This is the mean number of items inspected per lot. The formulas for a single-sampling
plan as given by the Statistical Research Group, Columbia University (1942, p. 212), are

1. Semicurtailed

c+1
ASN c = nF ( c n ) +
p
(
1 - F ( c + 1 n + 1) )

2. Fully curtailed

n-c c+1
ASN fc = F ( c n + 1) + (1 - F ( c + 1 n + 1)
q p

where F(x|n) denotes the probability of x or fewer defectives in a sample of n.

Curtailment of single-sampling plans is usually not recommended because of the diffi-

culty in estimating the process average from such data. Such procedures are quite common
with double- or multiple-sampling plans where the first sample may be inspected fully
and later samples curtailed.
96 Acceptance Sampling in Quality Control

An unbiased process average proportion defective from fully curtailed single-sample

data may be estimated using the method of Girshick et al. (1946) as follows:

c
Lot rejected: p =
U -1
d
Lot accepted: p =
U - 1

where
c is the acceptance number
d is the number defectives found
U is the number units inspected

With semicurtailed inspection, the formula becomes

c
Lot rejected: p =
U -1
d
Lot accepted: p =
U

For example, suppose the sampling plan n = 10, c = 1 was to be used with semicurtailed
inspection and the second defective was found as the sixth item inspected. Inspection
would stop since it is obvious that the lot would be rejected. An estimate of the process
average would be

1
p = = .20
6 -1

Calculation of the ASN, if the fraction defective were actually .20, using the binomial dis-
tribution gives

2
ASN c = 10 F ( 110
| )+
.20
(1 - F ( 2|11) )
= 10 ( .3758 ) + 10 ( .3826 ) = 7.584

This indicates that semicurtailment would give an average saving of 2.416 units per inspec-
tion at the cost of some precision in estimating the process average.
The concept of ASN is very useful in determining the average number of samples that
will be inspected in using more advanced sampling plans. In double-sampling plans,
for example, the second sample is taken only if the results from the first sample are not
sufficiently definitive to lead to acceptance or rejection outright. In such a situation,
the inspection may be concluded after either one or two samples are taken and so the
concept of the ASN is necessary to evaluate the average magnitude of inspection in the
long run.
Concepts and Terminology 97

Tolerance and Confidence Intervals

Specifications are sometimes written in terms of tolerance intervals. This is particularly true
in applications of acceptance sampling in the reliability and life-testing areas. Tolerance
intervals specify limits that are estimated to contain a specified proportion of the popu-
lation π with a given confidence γ. Confidence intervals merely estimate a range within
which a population parameter is expected to lie with a given confidence γ. Both kinds of
intervals are available for attributes- and variables-type sampling.
An example of a specification of this type is the following: at least a proportion π of the
population must be acceptable with a confidence level (coefficient) of γ (or, more simply
stated, 100π percent reliability with 100γ percent confidence).
In testing a lot to this type of specification, Type B probabilities are used, since the speci-
fication refers to the population produced by the producer’s process—not the specific lot.
Now just one test of conformance to the specification will be made to accept or reject the
lot. The term “confidence” is taken to mean that whatever method is used, it is to give the
correct result in approving a lot, as equaling or exceeding the specified reliability (100π
percent), more than 100γ percent of the time in repeated applications. The test on a particu-
lar lot will be either correct or incorrect in a single application. But in the long run, it will
accept lots having at least a proportion π nondefective at least a proportion γ of the time. In
this sense an accepted lot can be viewed with γ confidence as having a π or greater propor-
tion of units conforming to the specification.
To illustrate, consider an example given by Mann et al. (1974, p. 374) rephrased as fol-
lows: Suppose that n = 20 and the observed number of failures is x = 1. What is the reli-
ability π of the units sampled with 90% confidence? Here π is unknown and γ is to be .90.
It is necessary to obtain

Pr éë q¢ ³ p0 ùû ³ .90

where π0 is a lower tolerance limit of q′, the fraction conforming in the population. For a
binomial distribution such as this, the tolerance limit problem resolves itself into finding a
lower confidence limit on q′ in the population sampled. The binomial tables give

20

åC
i=2
20 i 20 - i
i pq = .900 +

with p = .181 and with corresponding q = .819. So

P ( q ³ .819 ) = .90

and the lower tolerance limit is π = .819.

An alternative approach is to find the reliability or the confidence desired directly from
the OC curve. This may be done through the use of the relation

p = 1 - p g = 1 - Pa

The Type B OC curve for the plan n = 20, c = 1 is shown in Figure 4.5.
98 Acceptance Sampling in Quality Control

1.00

0.90

0.80

0.70

0.60

0.50
Pa

0.40

0.30

0.20

0.10

0
0 0.05 0.10 0.15 0.20 0.25 0.30
p

FIGURE 4.5 
Type B OC curve (n = 20, c = 1).

We see that for

Pa = 1 - g = 1 - .90 = .10

the corresponding p value from the OC curve is .181. Therefore, as before, the estimated
reliability from the sample is

p = 1 - p = 1 - .181 = .819

These relationships can also be employed to find a sampling plan to be used with specifi-
cations of the tolerance interval type. Suppose the life of a tire is specified to be such that
87% of the population must last more than 20,000 mi. with 75% confidence. Here, we have

Pa = 1 - g = 1 - .75 = .25

and

p = 1 - p = 1 - .87 = .13

Then it is clear that the plan n = 20, c = 1 would satisfy this requirement since for p = .129
the value of Pa = .25.
Note that when specifications are stated in terms of tolerance intervals, only one point
on the OC curve is specified. Thus, the plan n = 10, c = 0 also satisfies the requirements
Concepts and Terminology 99

of the tire example but does not offer the producer as much protection against good lots
being rejected.
It should be pointed out that this type of problem may also be solved using measure-
ments and a variables sampling plan with a reduction in sample size. The procedure
involved is much the same and will be discussed under variables sampling plans.

Levels and Risks

It is usually desirable to set up a sampling plan with both the producer’s and consumer’s inter-
ests in mind. This benefits both since their interests are not mutually exclusive and are in fact
to a large extent compatible as seen in Table 4.14. While the producer and consumer risks are
fairly well defined in terms of good products being rejected and bad products being accepted,
respectively, each has an interest in having reasonable levels maintained for the other.
Since specification of two points may be used to define an OC curve, it is often desir-
able to specify these points in terms of the producer and the consumer. Thus, we have the
following:

Producer’s quality level (PQL): It is a level of quality that should be passed most of
the time. The state of the art almost always prohibits this from being a fraction
­defective of zero.
Producer’s risk (PR): It is the risk of having PQL material rejected by the plan.
Consumer’s quality level (CQL): It is a level of quality that should be rejected most of
the time.
Consumer’s risk (CR): It is the risk of having CQL material accepted by the plan.

It is customary (though not necessary) to designate the PR as .05 indicating Pa = .95 at the
PQL and the CR as .10 to give Pa = .10 at the CQL. This value of the CQL in percent is called
the “lot tolerance percent defective” (LTPD) for 10% limiting quality of the plan. Figure 4.6
shows the location of these points on the OC curve for n = 20, c = 1. While the PR and
CR may take on any values, if the traditional values are taken, we have the following for
n = 20, c = 1: PQL = .018, PR = .05, CQL = .181, and CR = .10.
In addition to these points, a third important point on the curve is defined as follows:

Indifference quality (IQ) level: It is the point where the producer and the consumer
share a 50% probability of acceptance.

This point characterizes the plan in the sense of equal risk, although it is unlikely that the
producer will stay in business if 50% of the lots are rejected. Rather, the IQ quantifies the
area of vagueness or indifference between the consumer and the producer.

TABLE 4.14
Producer and Consumer Interests
Producer Consumer

Good lots rejected Good product lost (producer risk) Potential higher cost
Bad lots accepted Potential customer dissatisfaction Paid for bad product (consumer risk)
100 Acceptance Sampling in Quality Control

1.00
PR
1–α

0.50
Pa

β
CR
0 p
PQL IQ CQL

FIGURE 4.6 
PQL, CQL, PR, CR (n = 20, c = 1).

Use of these points to describe sampling plans allows for their ready develop-
ment and characterization. This is particularly true when the nature of Type A and
Type B plans is considered. The principal measure of Type A plans is the IQ at which
the  probability of acceptance is split equally between the interests of the producer
and the consumer. The slope of the OC curve is then a subsidiary but important con-
sideration measuring the ­discrimination of the plan. This can be seen in the work of
Hamaker (1950).
Type B plans, however, are driven by their process orientation. As such, the ARL becomes
an important consideration along with percent lots accepted. Experience over the years
has led to values of 95% and 10% for PR and CR, respectively. This amounts to 1 lot in
20 rejected in error when quality is good and 1 lot in 10 accepted in error when quality
deteriorates. Corresponding ARLs are an average of 20 lots inspected until a lot at the PQL
is rejected in error and an average of 1.11 lots inspected until a lot at the CQL is accepted.
These values are summarized as follows:

Percent of Lots Accepted ARL

PQL 95 20
IQ 50 2
CQL 10 1.05

It will be seen then that they represent the producer’s and consumer’s interests quite
well and that they give good coverage of the region of interest.
Concepts and Terminology 101

Choosing Quality Levels

Consider the term acceptance sampling as it applies to choosing quality levels and take a par-
ticular note of its reference to acceptance, not rejection. One of the common tenets of sampling
is that the supplier is honorable and wishes to do the right thing within the bounds of ordinary
commerce. It is assumed that quality is generally good and that bad lots are an aberration and
not a practice usually. If this were not so, the consumer needs a new supplier and possibly
100% inspection until one is found. Sampling plans are intended to validate the acceptability
of the lot and alert those concerned of any departure from acceptable norms so that corrective
action can be undertaken. This is an important consideration in determining where quality
levels should be set. Certain levels of risk have become more and more common.
The CR is of primary interest in Type A situations and is often set at 10% at the CQL. That level
of risk has borne the test of time having been incorporated as the LTPD as early as 1923 in the
development of sampling plans at Western Electric. Indeed, Dodge and Romig (1944, p. 32) point
out that their tables “are based on a consumer’s risk of 0.10, a value found in most practices.”
The PR is of prime importance in developing and evaluating Type B plans. Use of 5%
for the PR at the AQL has become quite conventional, also meeting the test of time. This
goes back at least to the Statistical Research Group, Columbia University (1948, p. 145), in
acknowledging “use of the 95% point is quite common.” When a flow of lots is considered,
this implies a rejection rate, in error, of 1 lot in 20 at the PQL, which seems quite reasonable
when other alternatives are investigated.
The remaining risks are often carried over between the types. While a CR of 10% is used
as a key measure of Type A plans and a PR of 5% is key to Type B plans, the choice of 5%
PR for Type A plans and 10% CR for Type B plans has become quite common.
Once values are assigned, PQL and CQL split the OC curve into three regions as in
Figure 4.6. The PQL is the maximum value in a region of acceptance in which the PR is less
than or equal to the value specified. The CQL is the minimum in a region of rejection for
which the CR is less than or equal to the value specified. Between these two regions is an
area of indifference in which acceptance or rejection is largely a matter of chance. The PQL
then represents the worst quality that can be tolerated on a long-term basis, while the CQL
represents the lower limit on quality, which must be rejected. The PQL should be set at the
state of the art or worse if that can be tolerated. The CQL should be set with regard to the
conditions surrounding the potentially defective product. Thus, the PR may be in parts per
million, but the CR in percent if parts per million failures would be expected to generate
defective products in the magnitude of percent—for example, failure of one head on a five-
headed parts per million machine would produce 20% defective.
The four values (PQL, PR, CQL, and CR) provide a useful characterization of sampling plans
and can be regarded as particularly important points on the OC curve. Nevertheless, it is
important to point out that other values of risk may be desirable. For example, values of CR of
5% are sometimes used in reliability plans to match customary ways to express specifications
in that field. The choice of levels should always be made with the resulting OC curve in mind.
The choice of quality levels (CQL and PQL) with which to construct a sampling plan
must be made considering the seriousness of the defects to which it is applied, OCs of the
resulting sampling plan, economic consequences in terms of sample size, ability of the
producer to meet the levels, and needs of the consumer that must be met. The construction
of any sampling plan involves a trade-off of these items.
As mentioned earlier, no acceptance control procedure should be instituted without ref-
erence to as much process control information as is available. Necessary and obtainable
102 Acceptance Sampling in Quality Control

quality levels must be determined so that the acceptance sampling scheme employed is a
cost-effective compromise in the interest of both the producer and the consumer. The best
tool in choosing quality levels is a well-designed control chart and possibly process opti-
mization studies to see what levels can economically be met.
Acceptance control should not be thought of as a policing operation but rather as the first
step toward mutually acceptable process controls to maximize the cost-effectiveness of
both approaches. This may allow an eventual use of surveillance inspection to detect any
departure from agreed-on levels at minimal cost to both parties.
Further discussion of quality levels and risks can be found in Chapter 19 on the
­administration of acceptance sampling plans.

Classification of Defects
Defect types are not all of the same concern. Dodge and Torrey (1956) have pointed out
that this is because

1. Defects of different kinds are not all equally serious

2. Defects of the same kind differ in seriousness according to the extent of departure
from specified limits or standards

Accordingly, defects are sometimes classified into groups that reflect their seriousness.
Quality levels for sampling inspection are set accordingly. One such classification has been
given by the Statistical Research Group (1948, p. 82).
Major: Will cause failure of the item to function as intended
Minor: Will impair the efficiency, shorten the lifetime, or otherwise reduce the value
of the item
Irregularity: A departure from good workmanship not affecting the performance or
life of an item
Sometimes an additional category is added reflecting concern for the product safety.
Using the definition of MIL-STD-414, for example, this may be

Critical: Could result in hazardous or unsafe conditions for individuals using or

maintaining the product

A leak or a flat spot might be a major defect in a tire. A blemish could be a major defect
while an illegible letter in the brand name may be an irregularity. Clearly, a weak spot or
damaged cord that could lead to a blowout would be a critical defect. It is essential that any
classification of defects be carefully and explicitly defined before it is used.

Measures of Sampling Plans: Terminology

A distinction may be made between the product that is definitely objectionable to the
­consumer on the one hand and the product that fails to meet specifications on the other.
While it is hoped that these categories overlap, they need not always coincide. This can
Concepts and Terminology 103

TABLE 4.15
Specifications and Defects
Product Performance

Specifications Nondefective Defective

Met OK Need tighter specifications
Not met May loosen specifications OK

be seen in Table 4.15. Also specifications imposed on a product, its subassemblies, and
constituents may be essentials to the production operation and its efficiency but have no
relation to the quality of the product as perceived by the consumer. Thus, parts may be
restricted to certain dimensions for the efficient operation of a feed mechanism in pro-
duction but have no relation whatsoever to the quality as measured by the ultimate con-
sumer. For this reason, recent documents dealing with terminology in quality control have
attempted to make a distinction between satisfying the ultimate user and satisfying the
specifications. For example, the ISO 3534-2 standard (2006) entitled Statistics—Vocabulary
and Symbols—Part 2: Applied Statistics defines the following:

Defect: It is the nonfulfillment of a requirement related to an intended or specified

use, in other words, a departure of a quality characteristic from its intended level
or state that occurs with a severity sufficient to cause an associated product or ser-
vice not to satisfy intended normal, or reasonably foreseeable, usage requirements.
Defective: It is an item with one or more defects, in other words, a unit of product
or service containing at least one defect or having several imperfections that in
combination cause the unit to fail to satisfy intended normal, or reasonably fore-
seeable, usage requirements. (The word defective is appropriate for use when a
unit of product or service is evaluated in terms of customer usage [as contrasted
with conformance to specifications].)
Nonconformity: It is the nonfulfillment of a requirement, in other words, a departure of
a quality characteristic from its intended level or state that occurs with severity suffi-
cient to cause an associated product or service not to meet a specification requirement.
Nonconforming unit: It is a unit with one or more nonconformities, in other words, a
unit of product or service containing at least one nonconformity.

Clearly, the term defective is also appropriate for use in the handling of components and
materials internal to a production operation since one operation supplying material to another
would take on the roles of the producer and the consumer, respectively. In such a situation, a
part not meeting specifications would be viewed as a defective by the consuming operation
while it may be regarded as a nonconforming unit by the supplier, since the same part might go
to a user internally or externally who would find it capable of satisfying usage requirements.
Since acceptance sampling is usually presented in terms of an adversarial relationship
between a producer and a consumer, and since, in most applications, interest in acceptance
sampling is centered on satisfaction of usage requirements on the part of the consumer, the
terms defect and defective will be used here with the understanding that the terms nonconfor-
mity and nonconforming unit should be used when evaluating an item against a specification
when no evaluation is being made of its intended use internally or externally to the producer.
104 Acceptance Sampling in Quality Control

In summary, then, a sampling plan may be assessed, at any given incoming proportion
defective p, by five basic measures as defined in the ISO 3534-2 standard (2006):

1.
Probability of acceptance (Pa): This refers to the probability that when using a given
acceptance sampling plan, a lot will be accepted when the lot or process is of a
specific quality level, in other words, “the probability that a lot will be accepted
under a given sampling plan.” A plot of Pa against p comprises the OC curve. Such
curves are of the two types:
a. Type A: Plots the probability that a lot will be accepted against the proportion
defective in the lot inspected
b. Type B: Plots the proportion of lots that will be accepted against the proportion
defective in the producer’s process, which gives rise to the lot inspected
2.
ASN: This refers to the average sample size inspected per lot in reaching decisions
to accept or not to accept when using a given acceptance sampling plan, in other
words, “the average number of sample units per lot used for making decisions
(acceptance or nonacceptance).” The ASN is meaningful in Type B sampling situa-
tions. A plot of the ASN against p is called the ASN curve for the plan.
3.
AOQ: This refers to the expected average quality level of the outgoing product
for a given value of incoming product quality in other words, “the expected qual-
ity of outgoing product following the use of an acceptance sampling plan for a
given value of incoming product quality.” This is normally calculated only when
rejected lots are 100% inspected since otherwise AOQ = p for a stream of lots of all
of incoming product quality p. The AOQ is meaningful in Type B sampling situa-
tions. A plot of the AOQ against p is called the AOQ curve of the plan.
4.
AOQL: This refers to the maximum AOQ over all possible values of incoming
product quality level for a given acceptance sampling plan and rectification of all
nonaccepted lots unless specified otherwise, in other words, “for a given accep-
tance sampling plan, the maximum AOQ over all possible levels of incoming
quality.” This may be seen as the maximum point on the AOQ curve. The propor-
tion defective at which the AOQL occurs is denoted as pM. AOQL is sometimes
shown as pL.
5.
ATI: This refers to the average number of items inspected per lot including 100%
inspection of items in nonaccepted lots, in other words, “the average number of
units inspected per lot based on the sample size for accepted lots and all inspected
units in nonaccepted lots.” Thus, the ATI is the total average number of units
inspected for lots including sample units and units involved in 100% inspection as
required. The ATI is meaningful in Type B sampling situations. A plot of the ATI
against p is called the ATI curve of the plan.

Graphs of Measures
The principal measures of sampling plans are usually presented in the form of graphs,
which show at a glance how the plan will perform against various possible values of
proportion defective. Since knowledge of the incoming fraction defective is usually not
Concepts and Terminology 105

available (otherwise there would be no sense to sample), the graphs allow for rational
matching of the plan to the sampling situation. They portray performance against good
and bad quality. This allows selection of a plan on the basis of its protection and other
measures of performance without knowing the exact fraction defective to which the plan
will actually be applied.
Four such (Type B) curves are illustrated for the plan N = 20, n = 10, c = 1:

OC curve: It shows probability of acceptance plotted against possible values of propor-

tion defective. It is used in assessing the protection afforded by the plan (Figure 4.7).
ASN curve: It shows the ASN plotted against possible values of proportion defec-
tive. Used with plans involving several sampling stages, it shows how the average
sample size ­varies as incoming quality changes. It is used in assessing the inspec-
tion requirements for the plan in the absence of rectification (Figure 4.8).
AOQ curve: It shows the AOQ plotted against possible values of proportion defective.
The AOQ at the curve’s maximum is the AOQL. The proportion defective at which
it occurs is labeled pM. It is used in evaluating the effect on average quality going to
the consumer after 100% inspection of rejected lots to determine the level of assur-
ance afforded to the consumer by a rectification procedure. The magnitude of the
AOQL is sometimes ­represented as pL (Figure 4.9).
ATI curve: It shows the ATI plotted against possible values of proportion defective.
Used with rectification procedures, it indicates overall inspection requirements for
the total procedure (Figure 4.10).

Knowledge of these component measures of sampling plans allows the quality engineer to
properly prescribe the plan appropriate to the sampling situation.

1.00

0.90

0.80

0.70

0.60

0.50
Pa

0.40

0.30

0.20

0.10

0
0 0.10 0.20 0.30 0.40 0.50
p

FIGURE 4.7 
OC curve (n = 10, c = 1).
106 Acceptance Sampling in Quality Control

20

18

16

14

12
ASN
10

0
0 0.10 0.20 0.30 0.40 0.50
p

FIGURE 4.8 
ASN curve (n = 10, c = 1).

AOQ
0.05

AOQL

0.03

0.02

0.01

0
0 0.10 pM 0.20 0.30 0.40 0.50

FIGURE 4.9 
AOQ curve (N = 20, n = 10, c = 1).
Concepts and Terminology 107

20

18

16

14

12
ATI
10

0
0 0.10 0.20 0.30 0.40 0.50
p

FIGURE 4.10 
ATI curve (N = 20, n = 10, c = 1).

Specifying a Plan
Discriminating use of sampling procedures demands knowledge and specification of the
characteristics of the plans to be employed. A primary consideration is the protection
afforded to both the producer and consumer. Since two points may be used to characterize
the OC curve, it is customary to specify
p1 as the PQL
p2 as the CQL
α as the producer risk
β as the consumer risk
For single-sampling attributes plans, 1 − α and β can be determined directly from the
distribution function of the probability distribution involved. Figure 4.11 shows the rela-
tion of these quantities to the OC curve. Also shown are the regions of acceptance, indiffer-
ence, and rejection defined by these points. Quality levels of p1 or better are expected to be
accepted most of the time (≥1 − α) by the plan depicted. Quality levels of p2 or worse are
expected to be rejected most of the time (≤1 − β), while intermediate levels will experience
decreasing probability of acceptance as levels move from p1 to p2. Occasionally, only one set
of parameters (p1, α) or (p2, β) is specified. When this is done, any plan having an OC curve
passing through the points meets the criterion. A single-sampling attributes plan may be
specified by any two of the following: (p1, α), (p2, β), n, c. It may also be determined by speci-
fying AOQL and one of the other values listed. The operating ratio R = p2/p1 is often used
to characterize sampling plans. The operating ratio varies inversely with the acceptance
number and may be used to derive individual plans for the given values of α and β. Unless
otherwise stated, α is usually taken to be .05 and β taken to be .10 since these values have
become traditional in acceptance sampling having satisfied the test of long-term usage.
108 Acceptance Sampling in Quality Control

1.00
α
1–α

Pa

β
β

0 p1 p2 p

Acceptance Indifference Rejection

FIGURE 4.11 
Relation of p1, p2, 1 − α, and β to the OC curve.

The chapters that follow will be primarily devoted to presenting the operating
­ rocedure for implementing various sampling plans and schemes together with a means
p
for ­determining the Pa, ASN, AOQ, AOQL, and ATI under full and curtailed sampling. In
general, except where explicitly stated, Type B measures will be given since they also act as
conservative approximations to the Type A results.

Software Applications
Minitab
Minitab can create graphical measurements for a finite lot sampling plan using an approach
similar to that used to compare plans in Chapter 2. Rather than specify multiple n and c
values, we will only enter the singular values of each for our finite lot plan along with the
lot size N. To illustrate this approach, suppose that we revisit the finite plan N = 20, n = 10,
c = 1. The Minitab dialog box is completed as shown in Figure 4.12.
Minitab produces a nice 3-in-1 plot as shown in Figure 4.13 with the OC curve, AOQ
curve, and ATI curve. These curves agree with those found in Figures 4.7, 4.9, and 4.10,
respectively. Minitab does not produce an ASN plot for a single-sampling plan as the value
of n does not change with p as shown in Figure 4.8. Note that the AOQL is specified on the
Minitab output and not on the AOQ curve as seen in the following.
Concepts and Terminology 109

FIGURE 4.12 
Minitab dialog box to produce a finite lot sampling plan.

Operating characteristic (OC) curve Average outgoing quality (AOQ) curve

AOQ (percent defective)

4
1.0
3
0.9
2

0.8 1
Probability of acceptance

0
0.7 0 5 10 15 20
Incoming lot percent defective
0.6
Average total inspection (ATI) curve
0.5 18
Average total inspection

16
0.4
14

0.3 12

0 5 10 15 20 10
Lot percent defective 0 5 10 15 20
Sample size = 10, Acceptance number = 1 Lot percent defective

FIGURE 4.13 
Minitab OC curve for finite lot sampling plan N = 20, n = 10, c = 1.
110 Acceptance Sampling in Quality Control

Acceptance Sampling by Attributes

Measurement type: Go/no go
Lot quality in percent defective
Lot size: 20
Use binomial distribution to calculate probability of acceptance
Compare User Defined Plan(s)
Sample Size 10
Acceptance Number 1
Accept lot if defective items in 10 sampled ≤ 1; Otherwise reject.
Average outgoing quality limit (AOQL) = 4.082 at 14.873 percent defective.

Problems
1. A new firm sends a lot of 100 to qualify as a supplier of circuit boards for use in an
electronic assembly. The previous supplier had a process average of 10% defective.
In inspecting the lot, what type of plan should be applied: Type A or Type B? What
probability distribution is the correct one for constructing the OC curve of the plan
selected if specifications are in terms of percent defective? What probability distri-
bution should be used if the specifications are in terms of defects per 100 units?
2. Consider the plan n = 4, c = 0 used on lots of size 8. Draw the Type A and Type B OC
curves. Compute the probability of acceptance at p = .125, .25, .375, .50 as a minimum.
3. In Problem 2, the lot size is raised to 16 but the plan n = 4, c = 0 is retained. What are
the Type A probabilities of acceptance at p = .125, .25, .375, .50? How would these
probabilities change if the lot size were made still larger?
4. A special military radio is specified to have less than 1 defect in 50 units. How
many defects would be expected in a sample size of 100 units? What is the prob-
ability of 0, 1, 2, 3 defects in 100 units if quality were exactly at the specified level?
What is the probability of acceptance if two defects were allowed in a sample
of 100?
5. The plan n = 5, c = 1 is being used on the inspection of tires for a minor defect on a
series of lots. The process average has been 12.5% defective for some time. The defect
is hard to find and is missed 20% of the time. Draw the effective OC curve accounting
for the inspection error. Use the points p = .125, .25, .375, .50 at a minimum for the true
fraction defective.
6. If rejected lots are 100% inspected in Problem 2, draw the AOQ curve. Should
Type A or Type B probabilities be used? Compute AOQ for p = .125, .25, .375, .50 at a
minimum.
7. Draw the ATI curve for Problem 6. Compute ATI for p = .125, .25, .375, .50 at a
minimum.
8. The sampling plan n = 15, c = 2 is being used for inspection of gaskets as received.
If the incoming process fraction defective is .10 and the plan is curtailed only after
finding the third defective gasket, what is the ASN?
Concepts and Terminology 111

9. If a lot is rejected after 11 units are inspected in Problem 8, what is the estimated
­fraction defective.
10. For the plan n = 32, c = 1, 12.2 defects per 100 units will be rejected 90% of the time,
while 1.66 defects per 100 units will be accepted 90% of the time. If PR = CR = .10,
what are the PQL and CQL for this plan expressed in defects per unit.

References
Anderson, V. L. and R. A. McLean, 1974, Design of Experiments, Marcel Dekker, New York.
Dodge, H. F. and H. G. Romig, 1944, Sampling Inspection Tables, John Wiley & Sons, New York.
Dodge, H. F. and M. N. Torrey, 1956, A check inspection and demerit rating plan, Industrial Quality
Control, 13(1): 5–12.
Girshick, M. A., F. Mosteller, and L. J. Savage, 1946, Unbiased estimates for certain binomial sampling
problems with applications, Annals of Mathematical Statistics, 17: 13–23.
Hamaker, H. C., 1950, The theory of sampling inspection plans, Philips Technical Review, 13(9): 260–270.
Hicks, C. R., 1999, Fundamental Concepts in the Design of Experiments, 5th ed., Holt, Rinehart, &
Winston, New York.
International Organization for Standardization, 2006, Statistics—Vocabulary and Symbols—Part 2
Applied Statistics, International Organization for Standardization (ISO), Geneva, Switzerland.
Juran, J. M., 1962, Quality Control Handbook, 2nd ed., McGraw-Hill, New York.
Juran, J. M. and F. M. Gryna, 1970, Quality Planning and Analysis, McGraw-Hill, New York.
Mann, N. R., R. E. Schafer, and N. D. Singpurwalla, 1974, Methods for Statistical Analysis of Reliability
and Life Data, John Wiley & Sons, New York.
Ott, E. R., E. G. Schilling, and D. V. Neubauer, 2005, Process Quality Control, 4th ed., ASQ Quality
Press, Milwaukee, WI.
Schilling, E. G., 1961, The challenge of visual inspection in the electronics industry, Industrial Quality
Control, 18(2): 12–15.
Schilling, E. G., 2005, Average run length and the OC curve of sampling plans, Quality Engineering,
17(3): 399–404.
Statistical Research Group, Columbia University, 1948, Sampling Inspection, McGraw-Hill, New York.
5
Single Sampling by Attributes

The single-sampling plan is basic to all acceptance sampling. The simplest form of such
a plan is single sampling by attributes, which relates to dichotomous situations, that is,
those in which the inspection results can be classified into only two classes of outcomes.
This includes go/no-go gauging procedures as well as other classifications, such as mea-
surements in or out specifications. Applicable to all sampling situations, the attributes
single-sampling plan has become the bench mark against which other sampling plans are
judged. It is employed in inspection by counting the number of defects found in the sam-
ple (Poisson distribution) or evaluating the proportion defective from processes or large
lots (binomial distribution) or from individual lots (hypergeometric distribution). Single
sampling is undoubtedly the most used of any sampling procedures.

Operation
Implementation of an attributes single-sampling plan is very simple. It involves taking a
random sample of size n from a lot of size N. The sample may be intended to represent
the lot itself (Type A sampling) or the process used to produce the lot (Type B sampling).
The number of defectives (or defects) d found is compared to an acceptance number c.
If the number found is less than or equal to c, the lot is accepted. If the number found is
greater than c, the lot is rejected. The operation of the plan is illustrated in Figure 5.1.

Selection
Tables of single-sampling attributes plans are available. Perhaps, the two best-known
sources are military standard MIL-STD-105E (1989) and its derivatives and the Dodge and
Romig tables (1959). These will be discussed in later chapters. The use of such tables as
a  collection of individual plans provides ease of selection on the basis of the operating
characteristics (OCs) and other measures classified therein.
Analytic procedures are also available for determining the so-called two-point single-
sampling plans for specified values of

p1 as the producer’s quality level (PQL)

p2 as the consumer’s quality level (CQL)
α as the producer’s risk (PR)
β as the consumer’s risk (CR)

113
114 Acceptance Sampling in Quality Control

Sample n

d≤c d>c

Accept Reject

FIGURE 5.1
Procedures for single sampling by attributes.

TABLE 5.1
Procedures for Determining Single-Sampling Plans
Type Plan Method Use

Type B (defectives) Tables of Poisson unity Tables for derivation of plan given operating ratio R for
(defects) values tabulated values of α, β, and c. Poisson approximation to
binomial for defectives. May be used as exact for defects.
Type B (defectives) Binomial nomograph Nomograph for derivation of plan given α, β, p1, p2. Uses
binomial distribution directly. Hence, exact for defectives.
Type A (defectives) f-Binomial nomograph Uses binomial nomograph to derive Type A plans given α, β,
p1, p2 through f-binomial approximation to hypergeometric
distribution. Given the lot size, it gives approximate plan for
defectives.
Type B (defects) Thorndyke chart Procedure for use of Thorndyke chart for Poisson distribution
(defectives) to derive plan given α, β, p1, p2. Exact for defects.
Approximate for defectives through Poisson approximation
to the binomial.
Type A (defectives) Hypergeometric tables Iterative procedure for derivation of exact hypergeometric
plan given N, α, β, p1, p2 using the Lieberman-Owen tables of
hypergeometric distributions.

where
p2
R= = operating ratio
p1

Five such procedures will be set forth here. They relate to the derivation of plans as indi-
cated in Table 5.1.
When constructing a plan for defects, rather than defectives, with these procedures, use
p as the number of defects per unit. The tables of unity values and the Thorndyke chart
may then be used directly, where n is simply the sample size, that is, the number of units
sampled.

Tables of Poisson Unity Values

Factors for construction of single-sampling plans are available in the literature, which is
based on the Poisson distribution and which provides excellent approximations to the
binomial sampling situation as well. These include the original approach of Peach and
Littauer (1946) together with the work of Grubbs (1949) and Cameron (1952) and the tabu-
lations by the U.S. Army Chemical Corps Engineering Agency (1953). These so-called unity
Single Sampling by Attributes 115

values are expressed as the product np, where n is the sample size and p is the proportion
defective. When dealing with defects, p is the defects per unit. The unity values can be eas-
ily used to construct and evaluate plans on the basis of the operating ratio.
Appendix Tables T5.1 and T5.2 present the values for single-sampling attributes plans
developed by Cameron (1952). Additional sets of unity values for matched sets of single,
double, and multiple plans have been developed by Schilling and Johnson (1980) and are
presented in Appendix Table T6.1. They may be used in the customary situation in which
α = .05 and β = .10. Other risk levels associated with p1 and p2 will also be found in the
Cameron (1952) tables. The values for other risk levels are used in a manner identical to
those for the conventional levels of α and β. The theory of construction of unity values is
explained by Duncan (1986).
To derive a plan having α = .05 and β = .10, determine the operating ratio R = p2/p1.
Appendix Table T5.1 lists values of R corresponding to various acceptance numbers c and
risks α and β. The value of c tabulated closest to the desired value of R for the indicated
risks is the acceptance number to be used. Choose a value of R from the table equal to or
just less than the value desired, to be conservative, in terms of guaranteeing both risks.
To find the sample size n, divide np1 by p1. Always round up in obtaining the sample size.
Appendix Table T5.2 shows the probability of acceptance associated with various unity
values for the plans and acceptance numbers given in Appendix Table T5.1. This table may
be used to evaluate the OC curve of any single-sampling attributes plan. Unity values are
shown for various acceptance numbers c tabulated in columns by the probability of accep-
tance P(A). Simply divide the unity values for a given acceptance number by the sample
size of the plan to get values of p and find the probability of acceptance for these values
from the column headings.
For example, for α = .05 and β = .10, suppose it is desired to have p1 = .018 and p2 = .18 so
that R = 10. Then for the value of R listed (10.96) in Appendix Table T5.1, the acceptance
number is shown to be c = 1. The sample size is .355/.018 = 19.7, which rounds up to 20.
The plan is n = 20, c = 1. If the probability of acceptance is to be evaluated for Pa = .10, use
Appendix Table T5.2 to find the corresponding p = 3.89/20 = .194. Moreover, the indiffer-
ence quality for this plan is 1.678/20 = .084.
In a similar manner, 13 points on the OC curve can be described using Appendix
Table T5.2 to obtain the following:

Pa p Pa p

.995 .103/20 = .005 .500 1.678/20 = .084

.990 .149/20 = .007 .250 2.693/20 = .135
.975 .242/20 = .012 .100 3.890/20 = .194
.950 .355/20 = .018 .050 4.744/20 = .237
.900 .532/20 = .027 .025 5.572/20 = .279
.750 .961/20 = .048 .010 6.638/20 = .332
.005 7.430/20 = .372

The procedure described holds the PR exactly because the PQL was used to obtain the
sample size, while the CR can vary slightly from the specified value when the sample size
is rounded. It will seldom be possible to hold both risks exactly. If the CR is to be held at
the expense of the PR, obtain np.10 corresponding to P(A) = .10 from Appendix Table T5.2
and divide by p2 to obtain the sample size. If the result differs from the sample size using
p1, use the larger sample size, or if only one p1 and p2 is of primary interest, use the sample
size associated with the value of interest.
116 Acceptance Sampling in Quality Control

Binomial Nomograph
The Larson (1966) nomograph presented earlier as Figure 3.5 can also be used to derive
single-sampling attributes plans. Given p1, p2, α, and β, plot p1 and p2 on the left scale for
proportion defective shown as “probability of occurrences on a single trial (p).” Then plot
1 − α and β on the right scale for probability of acceptance shown as “probability of c or
fewer occurrences in n trials (p).” With a straight edge, connect the points: p1 with 1 − α and
p2 with β. At the intersection of the lines, read the sample size n and the acceptance number
c from the grid.
The nomograph can also be used to evaluate the OC curve of a plan. To do this, plot
the point (n, c) on the grid. Locate the position of each value of probability of acceptance
to be evaluated on the right p scale. Then draw a line from p through (n, c) and read the
corresponding fraction defective p on the left scale. The procedure can be reversed to
find the value of probability of acceptance for a given value of proportion defective set
on the left scale.
For example, suppose α = .05, β = .10, p1 = .018, and p2 = .18. The derivation of the plan
n = 20, c = 1 is shown in Figure 5.2. The dotted line shows an indifference quality of p = .08
for this plan.
The Larson nomograph is based on the binomial distribution and so will allow the direct
evaluation of Type B plans for fraction defective. It allows the derivation and evaluation of
plans for values of probability of acceptance not shown in the Cameron tables. It provides
a reasonable and conservative approximation for the derivation of plans when the hyper-
geometric distribution should apply and the binomial approximation to the hypergeomet-
ric distribution is appropriate.

f-Binomial Nomograph
Ladany (1971) has provided a method for adapting the Larson binomial nomograph for
use in deriving Type A plans for a finite lot of size N when the f-binomial approximation
to the hypergeometric distribution applies. This is when the sampling ratio F = n/N > 0.1

p1 = 0.018

Pa

0.08
0.10 = β

p2 = 0.18 0.50

0.95 = 1 – α

FIGURE 5.2
Larson’s nomograph for n = 20, c = 1.
Single Sampling by Attributes 117

and the fraction defective p < .1. Other approximations are listed in Figure 3.9. A somewhat
more complicated direct method for deriving Type A plans using the Lieberman and Owen
(1961) tables is given later in this chapter.
To use the binomial nomograph in this context, for specified p1, p2, α, and β, determine
two pseudosample sizes:

n1 = p1N , n2 = p2 N

And consider two lines, simultaneously drawn, as in Figure 5.3:

Line A from β on the right scale, through the intersection c (to be determined) with
the n2 line on the grid
Line B from 1 − α on the right scale and the intersection of the same c with n1 line on
the grid

When the two lines intersect on the left scale and have identical values of c at n2 for line A
and n1 for B, respectively, the plan has been determined. For the value of c specified, the
value of the sampling fraction F = n/N can be read at the point of intersection on the left
scale. Multiplication of this value of F by the lot size will give the sample size n.
Ladany (1971) suggests the use of a thread or rubber band at β and 1 − α on the right
scale, which, when looped around a stylus and run up and down the left scale, would pro-
vide a flexible representation of the two lines. In any event, practice with the method soon
makes the use of two transparent rulers adequate for the purpose.
The procedure may, of course, also be used to evaluate the Type A probability of accep-
tance for the plan specified by N, n1, and c, for a given value of p as follows:

1. Locate the point representing c and n = Np on the grid.

2. Locate the value of n/N on the left (p) axis.
3. Draw a line through these points. The intersection of this line with the right (p)
scale will indicate the cumulative probability of acceptance for the value of p.

p 1

Pa
c

n
4 0.30 = β
3 0.50
2
0.76 = 1 – α
A
1
c
B

F = n/N = 0.50

FIGURE 5.3
f-Binomial application of nomograph.
118 Acceptance Sampling in Quality Control

Clearly, the procedure could be reversed to find the value of p corresponding to a given
value of probability of acceptance. This would involve drawing a line from Pa on the right
scale to n/N on the left. The intersection of the line with the curve for the acceptance con-
stant involved gives Np. Division of this value by N gives p.
For example, suppose lot size is N = 20 and it is desired to develop a Type A plan having
α = .24, β = .30, p1 = .10, and p2 = .20. Then we have n1 = 20(.10) = 2 and n2 = 20(.20) = 4. The
binomial nomograph would appear as in Figure 5.3.
Line A passes from p = .30 through (n2 = 4, c = 1) to p = .50. Line B passes from p = .76
through (n1 = 2, c = 1) to p = .50. The lines intersect at .50 on the left axis and so the plan has
sampling fraction F = n/N = .50. Hence, since N = 20, we find n = 10 and the plan is N = 20,
n = 10, and c = 1. The indifference quality level may be evaluated using the dotted line in
Figure 5.3 as Np = 3 so that p = 3/20 = .15. This is exactly the value obtained for the plan in
Table 4.2 using the tables of Lieberman and Owen (1961).
It should be noted that the discrete nature of the hypergeometric distribution precludes
certain fractions defective from occurring. This should be considered throughout in appli-
cation of the binomial nomograph in this way.

Thorndyke Chart
Although somewhat more complicated than Larson’s binomial nomograph, the Thorndyke
(1926) chart, as given in Dodge and Romig (1959), may be used to derive a single-sampling
attributes plan. This chart, presented earlier as Thorndyke chart (Figure 3.6), uses cumula-
tive Poisson probabilities on the ordinate and unity values np on the abscissa. Curves for
various acceptance numbers are shown. The procedure, adapted from Burgess (1948), is as
follows:

1. Project an imaginary horizontal line from β.

2. Place the bottom edge of a piece of paper on the line so that the bottom left corner
of the paper lies above np = 1.
3. Project an imaginary horizontal line from 1 − α on the ordinate and mark its inter-
section with the paper on the vertical left edge.
4. Project an imaginary vertical line from the unity value equal to the operating ratio
R desired up to the bottom edge of the paper and mark the paper at the point of
intersection.
5. Slide the bottom edge of the paper along the line projected from β until a single c
curve most nearly passes through both the 1 − α and the R marks on the paper at
the same time. This is the value of c for the sampling plan.
6. For this value of c, read the value of np corresponding to 1 − α or the value of np
corresponding to β. Division of either of these unity values by p1 or p2, respectively,
will give the sample size for the plan. Unless one risk is specifically to be held, use
the larger of the two sample sizes or choose an appropriate compromise interme-
diate value.

For use in determining the plans for defects, simply substitute the desired values of
defects per unit for p in the previous procedure.
To illustrate this method, suppose it is desired to have α = .05, β = .10, p1 = .018, and
p2 = .18. The corresponding operating ratio is R = 10. The resulting Thorndyke chart is
Single Sampling by Attributes 119

5 6
4
3
2
1
1 – α = 0.95 1–α Slide
0
R
β = 0.10

(a) 1.0 R = 10.0

2
1
1 – α = 0.95 1–α Slide
0 c=1
R
β = 0.10

(b) 0.36 3.8

FIGURE 5.4
Thorndyke chart to derive plan: α = 0.05, β = 0.10, p1 = .018, and p2 = .18. (a) Construction of slide and (b) deriva-
tion of plan.

shown in Figure 5.4 and shows the appropriate acceptance number to be c = 1. The curve
for c = 1 shows np.95 = .36 and np.10 = 3.8. Dividing these by p1 and p2, respectively, we obtain
n = 20 and n = 21. If the PR is to be held, the plan n = 20, c = 1 would be used.
The Thorndyke chart is based on Poisson probabilities and so like Cameron’s tables
serves as a good approximation to the binomial distribution for defectives and is exact
in dealing with defects per unit. It provides a procedure for determining a plan based
on the Poisson distribution for values not tabulated in the tables, which employ unity
values.

Hypergeometric Tables
When sampling from a single (isolated) lot, the construction of a sampling plan is com-
plicated by the computationally cumbersome hypergeometric distribution. The computer
can be a real help here. Alternatively, when the situation is such that the approximations
shown in Figure 3.9 are appropriate, they should be used. This may be thought of as sub-
stituting the appropriate Type B OC curve for the approximating distribution in place of
that of the Type A plan that uses the hypergeometric. Alternatively, Ladany’s approach to
the use of the binomial nomograph may be employed.
If the approximations are not appropriate or an exact solution is required, an iterative
method may be used as outlined in the following. It requires that tables of the hypergeo-
metric distribution, such as those of Lieberman and Owen (1961), be available. If they
are not available, recourse must be made to the computer or in simple problems to hand
120 Acceptance Sampling in Quality Control

calculation. The procedure for selecting a Type A plan using the hypergeometric distribu-
tion is as follows:

1. Specify PQL = p1, PR = α, CQL = p2, and CR = β.

2. Use approximations of Figure 3.9 if appropriate to develop an analogous Type B
plan.
3. Use hypergeometric tables to obtain a plan as follows:
a. Determine a binomial plan n*, c* that meets the desired specifications.
b. For a given lot size, N, determine the number defective in the lot D1 = Np1 cor-
responding to the PQL and D2 = Np2 corresponding to CQL. Start at n = n* and
c = c*, and iterate through the hypergeometric tables, alternating D2 and D1
between successively lower sample sizes and acceptance numbers until a mini-
mum sample size plan meeting the specifications is obtained. This involves
simultaneously satisfying the inequalities:

F ( x ) ³ 1 - a at D1

F ( x ) £ 1 - b at D2

c. Iterate as follows:

N n D2 x F(x) N n D1 x F(x)

N n* D2 c* β1 → N n* D1 c* (1 − α)1
N na D2 ca β2 ← N na D1 ca (1 − α)2
N nb D2 cb β3 → N nb D1 cb (1 − α)3
⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ ⋯
N nk D2 ck βk ≤ β → N nk D1 ck (1 – α)k ≥ 1 − α

Having started with the binomial plan, which because of its conservative nature assures
β1 ≤ β and (1 – α)1 ≥ 1 – α, lower the sample size until the inequalities are violated. Next
lower the acceptance number and again successively drop the sample size until it is con-
firmed that the inequalities do not hold. Repeat this process until an acceptance number
is found for which a sample size cannot be obtained satisfying the inequalities. The plan
identified for the next higher acceptance number is the hypergeometric plan satisfying the
specifications. In general, probability of acceptance is lowered by increasing n and lower-
ing c with consequent decrease in β and 1 – α.
To illustrate the method, suppose a hypergeometric plan is to be selected having p1 = .10,
p2 = .20, α = .24, and β = .30 with a lot size of 20. From the Larson nomograph, the binomial
plan satisfying these specifications is n = 17 and c = 2. Using the Lieberman–Owen (1961)
tables for

D1 = .1 ( 20 ) = 2

D2 = .2 ( 20 ) = 4

the following results are obtained against the specified

b = .30 1 - a = .76
Single Sampling by Attributes 121

Steps N n D2 x F(x) N n D1 x F(x) Plan

1 20 17 4 2 .0877 20 17 2 2 1.0000
2 20 15 4 2 .2487 20 15 2 2 1.0000 n = 15, c = 2
3 20 14 4 2 .3426 20 14 2 2 1.0000
4 20 15 4 1 .0320 20 15 2 1 .4474
5 20 12 4 1 .1531 20 12 2 1 .6526
6 20 9 4 1 .3746 20 9 2 1 .8105
7 20 10 4 1 .2910 20 10 2 1 .7632 n = 10, c = 1
8 20 10 4 0 .0433 20 10 2 0 .2368
9 20 9 4 0 .0681 20 9 2 0 .2895
10 20 8 4 0 .1022 20 8 2 0 .3474

The plan is n = 10, c = 1. This is the same plan that is obtained by the Ladany f-binomial
adaptation of the Larson nomograph. In step 1, the binomial plan was used. Sample size was
reduced to obtain the plan n = 15, c = 2 in step 2, which satisfies the inequalities but which
has not been shown to be optimum in terms of sample size. Step 3 confirms the plan in step 2.
Step 4 lowers the acceptance number, while steps 5 through 8 lead to the plan n = 10, c = 1.
Steps 9 and 10 confirm that no plan exists for the next lower acceptance number.
These results could be obtained using a computer or possibly a programmable calcula-
tor. Furthermore, the strategy employed can be used in the development of other types of
sampling plans. Using a slightly different approach, Guenther (1969) has outlined a gen-
eral iterative procedure by which two-point plans can be obtained from binomial, hyper-
geometric, or Poisson tables.

Measures
The performance of single-sampling attributes plans may be characterized by the measures
given in Table 5.2. These may be evaluated using the binomial or Poisson distributions as
appropriate to the sampling situation. Care must be exercised in the use of the hyper-
geometric distribution due to the depletion of the lot as samples are taken. The formulas
should be modified accordingly. The distributions are listed in Table 3.1. The x and y values
for calculation of the average outgoing quality limit (AOQL) are explained in Chapter 14
and are given in Appendix Table T14.1 based on the Poisson model. The approximation
y ≈ 0.4(1.25c + 1) shown for y was developed by Schilling for acceptance numbers of 5 or
less. The notation F(c|n) indicates the probability of c or fewer defectives in a sample of n.
The frequency function f(c|n) is interpreted accordingly.
To illustrate application of these formulas, suppose the measures of the plan n = 20 and
c = 1 are desired when sampling from a succession of lots of size N = 120. They are to be
calculated using Type B (binomial) probabilities when the incoming proportion defective
is p = .18.
Probability of acceptance:

Pa = F ( 1|20 ) = C020 ( .18 ) ( .82 ) + C120 ( .18 ) ( .82 )

0 20 1 19

= .019 + .083 = .1102

122 Acceptance Sampling in Quality Control

TABLE 5.2
Measures of Single-Sampling Attributes Plans
Measure Formula
Probability of acceptance Pa = F(c|n)
ASN Full inspection
ASN = n
Semicurtailed
c+1 é
( )
ASN c = nF c n +
p ë
(
1- F c +1 n +1 ù
û )
Fully curtailed
n-c c+1 é
ASN fc =
1- p
(
F c n+1 + )
p ë
(
1- F c +1 n +1 ù
û )
AOQ Defectives found replaced
æ N-nö
AOQ = pPa ç ÷
è N ø
Defectives not replaced
æ N-n ö
AOQ = pPa çç ÷÷
è N - np ø
Approximate (n/N small)
AOQ ≅ pPa
AOQLa Defectives found replaced with good
y æ N-nö
AOQL = ç
n è N ÷ø
Defectives not replaced
y æ N-n ö
AOQL = çç ÷
n è N - np ÷ø
Approximate (n/N small)
y .4
AOQL = @
n n
( )
1.25 ( c + 1) for c £ 5

AOQL occurs at
x
pM =
n
ATI ATI = n + (1 − Pa)(N − n) = nPa + N(1 − Pa)
a x, y values given in Appendix Table T14.1.

Average sample number (ASN):

Full inspection, ASN = 20

Semicurtailed
1+1
ASN c = 20 F ( 1|20 ) + é1 - F ( 2|21) ùû
.18 ë
2
= 20 ( .102 ) +
.18
[1 - .244] = 10.44

Fully curtailed
20 - 1 1+1
ASN fc = F ( 1|21) + é1 - F ( 2|21) ùû
1 - .18 .18 ë
20 - 1 1+1
=
1 - .18
(.087 ) +
.18
[1 - .244] = 10.42

Single Sampling by Attributes 123

Average outgoing quality (AOQ):

Defectives found replaced

æ 120 - 20 ö
AOQ = .18 ( .102 ) ç ÷ = .015
è 120 ø
Defectives not replaced
æ 120 - 20 ö
AOQ = .18 ( .102 ) ç = .016
ç 120 - 20 ( .18 ) ÷÷
è ø
Approximate

AOQ = .18 ( .102 ) = .018

AOQL:

Defectives found replaced

.8400 æ 120 - 20 ö
AOQL = ç ÷ = .035
20 è 120 ø
Defectives not replaced
.8400 æ 120 - 20 ö
AOQL = ç ÷ = .036
20 çè 120 - 20 ( .18 ) ÷ø

Approximate
.8400
AOQL = = .042
20
AOQL occurs at
1.62
pM = = .081
20
Average total inspection (ATI):

ATI = 20 + ( 1 - .102 ) ( 120 - 20 ) = 109.8

These calculations illustrate the value of the approximations shown and how little the
measures are affected by the variations shown in the inspection technique. The formulas
could also be evaluated for defects per unit using the Poisson model.

Software Applications
Excel
Use the Excel template Acceptance Sampling for Attributes (2-point plans).xlsm to generate an
acceptance sampling plan based on n and c. Using the earlier example from this chapter,
suppose that α = 0.05 and β = 0.10, and we desire p1 = 0.018 and p2 = 0.18. In Figure 5.5, the
 124

FIGURE 5.5
Excel template Acceptance Sampling for Attributes (2-point plans).xlsm.
Acceptance Sampling in Quality Control
Single Sampling by Attributes 125

1.80% Acceptable quality level, AQL

0.95 Desired probability of accepting lot at AQL ( 1 – α )

OC Curve 18.00% Rejectable quality limit, RQL

for Sample Plan 0.10 Desired probability of accepting lot at RQL (β)
20 n, Sample size necessary to achieve alpha and beta
1 c, Number of acceptable defectives in sample

0.950 Actual probability of acceptance at AQL

0.102 Actual probability of acceptance at RQL
AQL RQL
1.00
95.0%
0.90
Probability of accepting lot, Pa

0.80
0.70
0.60
0.50
0.40
Optimal OC curve
0.30
0.20
0.10 10.0%
0.00
0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 40.00%
Percent nonconforming, P

FIGURE 5.6
OC curve for derived sampling plan n = 20 and c = 1.

user enters 0.018 into cell B3 for the acceptable quality level (AQL) (p1), 0.95 in cell B4 for
1 − α, 0.18 in cell B6 for the rejectable quality level (RQL) or lot tolerance percent defective
(LTPD) (p2), and 0.10 into cell B7 for β. To get the plan, the user clicks the Compute n and c
button. The result of n = 20 and c = 1 agrees with the plan produced by the Cameron tables.
The OC and AOQ curves are available on separate tabs in the worksheet. They appear in
Figures 5.6 and 5.7.

Minitab
It is easy to develop an acceptance sampling plan in Minitab using the four speci-
fied values of α, β, p1, and p2. In Minitab, we choose Stat > Quality Tools > Acceptance
Sampling by Attributes, use Create a Sampling Plan from the top dropdown box, and
complete the dialog box as shown in Figure 5.8. The sampling plan is n = 28 and c = 2.
This plan does not agree with the one based on the Cameron tables, which is favorable
to the producer because it actually favors the consumer, that is, the plan has a β value
nearly or equal to the specified β value at the RQL or LTPD. This type of plan is consis-
tent with the plans produced by the Schilling and Johnson tables. The OC curve for this
plan is shown in Figure 5.9.
126 Acceptance Sampling in Quality Control

Average outgoing quality (%defective) given sampling plans

4.500%
Optimal AOQL = 4.14%
4.000%
Percent defective after sampling

3.500%

3.000%
(lower is better)

2.500%

2.000%

1.500%

1.000%

0.500%

0.000%
0.000% 5.000% 10.000% 15.000% 20.000% 25.000% 30.000% 35.000% 40.000%
Pre-inspection percent defective, p

FIGURE 5.7
AOQ curve for derived sampling plan n = 20 and c = 1.

FIGURE 5.8
Minitab dialog for derived sampling plan n = 28 and c = 2.
Single Sampling by Attributes 127

Operating characteristic (OC) curve

Sample size = 28, Acceptance number = 2
1.0
Probability of acceptance

0.8

0.6

0.4

0.2

0.0
0 10 20 30 40
Lot percent defective

FIGURE 5.9
Minitab OC curve for derived sampling plan n = 28 and c = 2.

Acceptance Sampling by Attributes

Measurement type: Go/no go
Lot quality in percent defective
Use binomial distribution to calculate probability of acceptance
Acceptable Quality Level (AQL) 1.8
Producer’s Risk (α) 0.05
Rejectable Quality Level (RQL or LTPD) 18
Consumer’s Risk (β) 0.1
Generated Plan(s)
Sample Size 28
Acceptance Number 2

Accept lot if defective items in 28 sampled ≤ 2; Otherwise reject.

Percent Probability Probability

Defective Accepting Rejecting
1.8 0.986 0.014
18.0 0.098 0.902

Statgraphics
Statgraphics has much more extensive acceptance sampling plan capabilities than Minitab.
It can produce several types of sampling plans discussed here and in later chapters. To dem-
onstrate its ability to produce a single-sampling plan, choose SPC > Acceptance Sampling
> Attributes > OC/AOQL/LTPD Plans from the menu. Using the same example used for
Excel and Minitab, enter 1.8 for the AQL and 18 for the LTPD, and let α and β default to the
given 5% and 10% values as shown in Figure 5.10. Be sure that the Action Create OC plan
128 Acceptance Sampling in Quality Control

FIGURE 5.10
Statgraphics dialog for derived sampling plan n = 28 and c = 2.

is selected. Note that Statgraphics requires a lot size value. Since this is a Type B plan with
a very large virtual “lot size”, that is, continuous series of lots from a process, enter 10,000
to replace the default 1,000 value.
The output produced by Statgraphics is more informative than that given by Minitab as
shown in the following. The plan it generates agrees with the one produced by Minitab,
namely, n = 28 and c = 2. The user can select the OC, AOQ, and ATI curves for this plan.
The OC curve appears in Figure 5.11.
Acceptance Sampling for Attributes
Lot size: 10,000
Desired Features
Producer's risk (alpha): 5.0%
Consumer's risk (beta): 10.0%
Generated Plan
Sample size (n) = 28
Acceptance number (c) = 2
Plan Attributes
Acceptable quality level (AQL): 1.8%
Producer's risk (alpha) = 1.35005%
Lot tolerance percent defective (LTPD): 18.0%
Consumer's risk (beta) = 9.76177%
Single Sampling by Attributes 129

Operating characteristic (OC) curve

n = 28, c = 2
1

Prob. of acceptance 0.8

0.6

0.4

0.2

0
0 10 20 30 40
True percent defective

FIGURE 5.11
Statgraphics OC curve for derived sampling plan n = 28 and c = 2.

AOQL = 4.86495% at 7.81493% defective

ATI =
162.627 units per lot at the AQL
3774.79 units per lot at the AOQL
9026.56 units per lot at the LTPD

The StatAdvisor
This procedure creates sampling plans for lot inspection. In this case, it has generated
a sampling plan based upon desired PRs and CRs. The plan states that 28 items should
be sampled from each lot of 10,000, and the lot accepted without further inspection
if the number of bad items is less than or equal to 2. Using such a plan, a lot ­containing
1.8% defective items will be rejected only 1.35005% of the time, while a lot c­ ontaining 18.0%
defective items will be accepted only 9.76177% of the time. If rejected lots can be subjected
to 100% inspection and all bad items replaced with good ones, the average percent of
defective items shipped will be no greater than 4.86495% (the AOQL).

The Statgraphics analysis summary displays the following information for the user:

• Desired features: Summarizes the user-specified features upon which the plan is
based. In the example mentioned earlier, the plan was constructed so as to have a
PR of no more than 5% and a CR of no more than 10%.
• Generated plan: Shows the smallest sampling plan that has the desired features. In
the example, n = 28 items are to be sampled from the lot of N = 10,000 and the lot
is accepted if no more than c = 2 items are nonconforming.
• Plan attributes: Exact results for the generated plan. This includes the following:
• Producer’s risk at the AQL: Probability of rejecting a “good” lot.
• Consumer’s risk at the LTPD: Probability of accepting a “bad” lot.
130 Acceptance Sampling in Quality Control

• AOQL: Assuming that rejected lots are 100% inspected and that any noncon-
forming items are replaced by good items, this is the maximum fraction of
nonconforming items that are accepted.
• ATI: Assuming that rejected lots are 100% inspected and that any nonconform-
ing items are replaced by good items, this is the average percentage of items in
a lot that will be inspected.
• The StatAdvisor: Explains the output to the user in plain English. For the cur-
rent plan, the alpha and beta risks are smaller than requested, since the sample
size must be an integer. If rejected lots are rectified, the maximum average per-
cent of nonconforming items that will be accepted is approximately 5%, which
would occur if the incoming lots contain 7.8% nonconforming items. For lots
containing exactly 1.8% defective items (“good” lots), approximately 163 items
out of each lot of 10,000 will be inspected on average.

Problems
1. Construct single-sampling plans to the following specifications given in proportion
defective with PR = .05 and CR = .10:
a. PQL = .04, CQL = .21
b. PQL = .03, CQL = .13
c. PQL = .02, CQL = .06
Use the following techniques:
a. Poisson unity values
b. Binomial nomograph
c. Thorndyke chart
Compare results.
2. Plot the Type B OC curve for the following single-sampling plans using Pa = .95, .75,
.50, .25, and .10 at a minimum for plotting positions.
a. n = 13, c = 1
b. n = 32, c = 0
c. n = 1125, c = 2
Use the following techniques:
a. Poisson unity values
b. Binomial nomograph
c. Thorndyke chart
Compare results.
3. Obtain a Type A plan for a lot size of 200 with PQL = .025 and CQL = .125 with risks
α = .05 and β = .10, use
a. Binomial nomograph
b. Hypergeometric tables (optional)
Determine the points Pa = .75, .50, .25 on the OC curve.
Single Sampling by Attributes 131

4. Derive a defect per unit plan having PQL = 1.1 and CQL = 12.2 defects per hundred
units for risks α = .05 and β = .10. What is the indifference quality for this plan?
5. If lots are received in quantities of 1000, obtain the AOQ and ATI at the following
points for the plans given in Problem 2 earlier and find the AOQL of each.
a. Pa = .95
b. Pa = .50
c. Pa = .10
6. Use the binomial nomograph to derive and compare the OC curves of the plans
n = 5 and c = 0, 1, at Pa = .95, .50, .10. Plot the AOQ and ATI curves for these plans for
lots of size 200. Find their AOQL.
7. Use unity values to derive and compare the OC curves for sample sizes 5 and 10 for
c = 1 at Pa = .95, .50, .10. Plot the AOQ and ATI curves for these plans for lots of size
200. Find their AOQL. Use α = .05 and β = .10.
8. Use the Larson nomograph to derive a binomial sampling plan satisfying the specifi-
cations p1 = .03, p2 = .09, α = .05, and β = .10. Then find the appropriate hypergeometric
plan for use with a lot of N = 500. This illustrates the importance of considerations of
lot size when the sampling fraction is high.
9. Using the Thorndyke chart, derive a plan to satisfy the conditions of Problem 8 and
compare with the results for that problem. Why is the sample size highest using the
Thorndyke chart?
10. For even degrees of freedom, it is well known that the complement of the cumulative
distribution function for the Poisson can be determined from χ2/2 with degrees of
freedom 2c + 2. Using this fact, derive the unity values for c = 3? (Hint: See Cameron
[1952].)

References
Burgess, A. R., 1948, A graphical method of determining a single sampling plan, Industrial Quality
Control, 4(6): 25–27.
Cameron, J. M., 1952, Tables for constructing and for computing the operating characteristics of
­single sampling plans, Industrial Quality Control, 9(1): 37–39.
Dodge, H. F. and H. G. Romig, 1959, Sampling Inspection Tables, Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York.
Duncan, A. J., 1986, Quality Control and Industrial Statistics, 5th ed., Richard D Irwin, Homewood, IL.
Grubbs, F. E., 1949, On designing single sampling inspection plans, Annals of Mathematical Statistics,
20: 242–256.
Guenther, W. C., 1969, Use of the binomial, hypergeometric and Poisson tables to obtain sampling
plans, Journal of Quality Technology, 1(2): 105–109.
Ladany, S. P., 1971, Graphical determination of single-sample attribute plans for individual small
lots, Journal of Quality Technology, 3(3): 115–119.
Larson, H. R., 1966, A nomograph of the cumulative binomial distribution, Industrial Quality Control,
23(6): 270–278.
Lieberman, G. J. and D. B. Owen, 1961, Tables of the Hypergeometric Probability Distribution, Stanford
University Press, Stanford, CA.
132 Acceptance Sampling in Quality Control

Peach, P. and S. B. Littauer, 1946, A note on sampling inspection, Annals of Mathematical Statistics,
17: 81–85.
Schilling, E. G. and L. I. Johnson, 1980, Tables for the construction of matched single, double, and
multiple sampling plans with application to MIL-STD-105D, Journal of Quality Technology, 12(4):
220–229.
Thorndyke, F., 1926, Applications of Poisson’s probability summation, The Bell System Technical
Journal, 5: 604–624.
United States Department of Defense, 1989, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes, (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
United States Department of the Army, 1953, Master Sampling Plans for Single, Duplicate, Double, and
Multiple Sampling, Manual No.2, Chemical Corps Engineering Agency/Army Chemical Center,
Edgewood, MD.
6
Double and Multiple Sampling by Attributes

Double- and multiple-sampling plans reflect the tendency of many experienced inspectors
to give a questionable lot an additional chance. Thus, in double sampling, if the results of
the first sample are not definitive in leading with acceptance or rejection, a second sample
is taken that then leads to a decision on the disposition of the lot. This approach makes
sense, not only as a result of experience but also in the mathematical properties of the pro-
cedure. For one thing, the average sample number (ASN) can usually be made to be less for
a double-sampling plan than for a single-sampling plan with the same protection.
A natural extension of double sampling is to allow further additional samples to be taken
to achieve even more discrimination in the disposition of a lot. Such procedures are called
multiple-sampling plans when, as with double sampling, the last sample is constructed to
force a decision at that point. That is, for a specific last sample (say, the kth sample), it is so
arranged that rk = ak + 1, where rk is the rejection number and ak is the acceptance number.
Thus, double sampling is simply a special case of multiple sampling where k = 2.
Multiple-sampling plans allow even more flexibility and still further reduction in the
average sample size over double-sampling plans but are often found to be difficult to
administer because of the complexity of handling and recording all the samples required.
As an example of the reduction in sample size that can be obtained, MIL-STD-105E (Code H,
1.5 acceptable quality level [AQL], normal inspection) shows that for plans matched to the
single-sampling plan n = 50, c = 2, the ASN at the 95th percentile is as follows:

Plan ASN

Single 50
Double 43
Multiple 35

This is typical of the efficiency in sampling that may be generated by the use of double-
and multiple-sampling procedures. Efficiency of this sort may be costly, however, in terms of
administration, since there is an increasingly variable workload in going from single to double
to multiple sampling. These plans offer an additional dimension to the application of sampling
plans, however, by providing increased economy and flexibility when properly applied.
Double- and multiple-sampling plans are said to be matched to single-sampling plans
when their operating characteristic (OC) curves coincide. The inherent shape of a multiple-
sampling OC curve is, however, different from that of a single-sampling OC curve. Hence,
plans are often matched at two points, usually at p.95 and p.10.
Inspection is often curtailed, that is, inspection is stopped after reaching a decision, or
semicurtailed, that is, it is stopped only on a decision to reject. In either case, the first ­sample
is almost always inspected in full so that estimates and records kept on the first sample will
have a consistent sample size. Usually, the ASN is assessed at the producer’s quality level
(PQL), since this should be the sustained normal level of the operation of the plan if no
problems occur.

133
134 Acceptance Sampling in Quality Control

Operation
Double Sampling
Application of a double-sampling plan requires that a first sample of size n1 be drawn at
random from the lot (usually assumed large). The number of defectives d1 is counted and
compared to the first sample acceptance number a1 and rejection number r1.

If d1 ≤ a1, the lot is accepted.

If d1 ≥ r1, the lot is rejected.
If a1 < d1 < r1, a second sample is taken.

If needed, a second sample of size n2 is drawn. The number of defectives d2 contained in the
second sample is determined. The total number of defectives

D2 = d1 + d2

is compared to the acceptance number a2 and the rejection number r2 for the second sam-
ple. In double sampling, r2 = a2 + 1 to insure a decision on the second sample.

If D2 ≤ a2, the lot is accepted.

If D2 ≥ r2, the lot is rejected.

The operation of the plan is shown in Figure 6.1.

Multiple Sampling
Multiple sampling involves the inspection of specific lots on the basis of 1 to k succes-
sive samples as needed to make a decision. In MIL-STD-105E, k is taken as 7, that is, the

Sample n1

d1 ≤ a1 a1 < d1 < r1 d1 ≥ r1

Sample n2

D2 = d1 + d2 ≤ a2 D2 = d1 + d2 ≥ r2

Accept Reject

FIGURE 6.1
Procedure for double sampling by attributes.
Double and Multiple Sampling by Attributes 135

multiple-sampling plans contained therein must reach a decision by the seventh sample.
Multiple-sampling plans are usually presented in tabular form:

Sample Sample Size Cumulative Sample Size Acceptance Number Rejection Number

1 n1 n1 a1 r1
2 n2 n1 + n2 a2 r2
⋮ ⋮ ⋮ ⋮ ⋮
k nk n1 + n2 +  ⋯  + nk ak rk = ak + 1

To start the procedure, a sample of n1 is randomly drawn from a lot of size N and the
­number of defectives d1 in the sample is counted.
If d1 ≤ a1, the lot is accepted.
If d1 ≥ r1, the lot is rejected.
If a1 < d1 < r1, another sample is taken.
If subsequent samples are needed, the first sample procedure is repeated sample by sam-
ple. For each sample, the total number of defectives found at any stage (say the ith)

Di = åd
j =1
j

is compared with the acceptance number ai and the rejection number ri for that stage until a
decision is made. Since, for the last (kth) sample, rk = ak + 1, a decision must be made by the
kth sample. Sometimes acceptance is not allowed at the early stages of a multiple-sampling
plan; however, rejection can take place at any stage. When acceptance is not allowed, the
symbol # is used for the acceptance number. The operation of the plan is shown in Figure 6.2.

Selection
A convenient source of single-, double-, and multiple-sampling plans will be found in the
MIL-STD-105E tables and its derivatives. The OC curves and other measures presented
in these tables can be used to select an appropriate plan. Matched single- and double-
sampling plans are also given in the Dodge and Romig (1959) tables. These sets of tables
will be discussed in later chapters.
Procedures are also available for determining double- and multiple-sampling plans
using Poisson unity values in a manner similar to single-sampling plans. These require the
specification of p1 (PQL), p2 (CQL), α (producer’s risk), and β (consumer’s risk) and the cal-
culation of the operating ratio R = p2/p1. Double- and multiple-sampling plans also require
specification of the relationship of successive sample sizes, that is, a multiple m where for
double sampling n2 = mn1.
Duncan (1986) has provided a compilation of unity values and operating ratios for dou-
ble and multiple sampling as developed by the U.S. Army Chemical Corps Engineering
Agency (1953). The double-sampling plans are for m = 1 and m = 2, respectively, with the
136 Acceptance Sampling in Quality Control

Sample n1

d1 ≥ r1
d1 ≤ a1 a1 < d1 < r1

Sample n2

D2 = d1 + d2 ≤ a2 a1 < D2 < r2 D = d1 + d2 ≥ r2

Sample n1
i i
Di = ∑ dj ≤ a1 a1 < Di < ri Di = ∑ dj ≥ ri
j=1 j=1

Sample nk
k k
Dk = ∑ dj ≤ ak Dk = ∑ dj ≥ rk
j=1 j=1

Accept Reject

FIGURE 6.2
Procedure for multiple sampling by attributes.

rejection numbers constant for both samples. That is,

r1 = r2 = a2 + 1

The multiple-sampling unity values are presented in terms of plans in which the sample
size at each stage is equal, that is,

n1 = n2 =  = ni =  = nk

The Chemical Corps plans are not in matched sets and do not utilize acceptance numbers
corresponding to MIL-STD-105E.
Appendix Table T6.1 was developed by Schilling and Johnson (1980) for the construction
and evaluation of matched sets of single-, double-, and multiple-sampling plans. It may
be used to derive individual plans to meet specified values of fraction defective and prob-
ability of acceptance. It may also be employed to match the scheme performance of the
MIL-STD-105E system to that of an individual plan. The tables extend into the range of low
probability of acceptance useful in reliability, safety, and compliance testing.
The unity values np listed in Appendix Table T6.1 are based on MIL-STD-105E accep-
tance and rejection numbers. Values of n are for the first sample sizes; succeeding sam-
ples in double and multiple plans are all equal and of size n. The plans are numbered
by the corresponding single-sampling acceptance number and a letter (S, D, M) show-
ing the type of plan: single, double, and multiple. Two sets of double and multiple
plans are included in addition to those from MIL-STD-105E to cover operating ratios
R = 33 and R = 22 to facilitate matching an individual plan to the MIL-STD-105E system.
Double and Multiple Sampling by Attributes 137

No single-sampling plan has an operating ratio in this range. To obtain the operating ratio
R = 20, the ­double-sampling plan requires n2 = 5n1 and is the only plan in Appendix Table
T6.1 in which the first and second sample sizes are not equal. Supplementary plans, not in
MIL-STD-105E, are included to provide a complete set of plans to match single-sampling
acceptance numbers from 0 to 15. The table is for α = .05 and β = .10. Its use is similar to
that of the Cameron (1952) tables for single sampling including applications to plans to
inspect defects per unit.
To construct a given plan, follow these steps:

1. Decide if single, double, or multiple sampling is to be used.

2. Specify

p1 = PQL ( Pa = .95 )

p2 = CQL ( Pa = .10 )
3. Form the operating ratio
p2
R=
p1

4. Choose a plan having acceptance numbers associated with an operating ratio just
less than or equal to R.
5. Determine the sample size as
np2
n=
p2

Round up in determining the sample size.

6. The plan consists of the acceptance numbers and sample size chosen.
7. The OC curve may be drawn by dividing the values of np shown for the plan by
the sample size to obtain values of p associated with the values of probability of
acceptance listed.
8. The ASN curve may be drawn by multiplying the values of ASN/n shown by
the sample size and plotting the resulting values of ASN against the p values
obtained for the corresponding probability of acceptance.

The formula for sample size n is presented showing division by p2 to ensure the consumer’s
risk is maintained. Alternatively, n = np1/p1 as in the Cameron tables. The value of np1 can
be found under a probability of acceptance of .95 for the plan. If values of sample size dif-
fer between these two formulas, the probability of acceptance will be exact at the value of
p associated with the value of n actually used and approximate for the other value of p.
Sometimes a convenient intermediate sample size may be chosen.
For example, suppose a double-sampling plan is desired having 95% probability of
acceptance at p1 = .01 and 10% probability of acceptance at p2 = .05.

1. Double sampling is to be used.

2. p1 = .01, p2 = .05.
138 Acceptance Sampling in Quality Control

.05
3. R = = 5.
.01
4. The operating ratio is R = 4.40 for the plan, giving acceptance numbers

Ac 1, 4; Re 4, 5

5. So

4.398
n1 =  87.96 ~ 88
.05

This holds p2 exactly while p1 = .011. Note that if p1 is to be held

1.000
n1 = = 100
.01

and at this sample size p2 will be .044.

6. Using n = 88, the plan is as follows:

Cumulative
Sample Sample Size Sample Size Ac Re

1 88 88 1 4
2 88 176 4 5

7. OC curve is found by dividing the values of np shown by 88 to obtain the following:

Pa p Pa p
.99 .007 .10 .050
.95 .011 .05 .058
.90 .014 .01 .077
.75 .020 .005 .086
.50 .028 .001 .105
.25 .038 .0005 .114
.0001 .134

8. The ASN curve is found by multiplying the values of ASN/n1 shown by 88 and
plotting against the values of p obtained for the corresponding probabilities of
acceptance in step 7 to obtain the following:

p ASN p ASN

.007 94.4 .050 113.8

.011 109.6 .058 106.6
.014 115.8 .077 95.4
.020 125.0 .086 92.7
.028 129.4 .105 89.5
.038 124.4 .114 88.9
.134 88.3
Double and Multiple Sampling by Attributes 139

Appendix Table T6.1 can be used to find matching single (R = 4.89) and multiple (R = 4.67)
plans. They are as follows:

Sample Sample Size Cumulative Sample Size Ac Re

Single
1 134 134 3 4
Multiple
1 33 33 # 3
2 33 66 0 3
3 33 99 1 4
4 33 132 2 5
5 33 165 3 6
6 33 198 4 6
7 33 231 6 7

where # indicates no acceptance allowed on the first sample. These plans are matched
about as well as those in MIL-STD-105E. Their OC and ASN curves are found in a similar
manner to those of the single-sampling plan.
The OC curves of these single-, double-, and multiple-sampling plans are shown in
Figure 6.3. Their ASN curves are presented in Figure 6.4.
As a further illustration, suppose a plan is desired having a PQL of 1.78 defects per
100 units and a CQL of 19.5 defects per 100 units with risk α = .05 and β = .10. Converting

1.00

0.90 Single

Double
0.80 Multiple

0.70

0.60

0.50
Pa

0.40

0.30

0.20

0.10

0
0 0.01 0.02 0.03 0.04 0.05
p

FIGURE 6.3
OC curves for single, double, and multiple matched plans.
140 Acceptance Sampling in Quality Control

150

Single

120 Double

90 Multiple
ASN

60

30

0
0 0.01 0.02 0.03 0.04 0.05
p

FIGURE 6.4
ASN curves for single, double, and multiple matched plans.

to defects per unit, p1 = .0178 and p2 = .195, so that

.195
R= = 10.96
.0178
Appendix Table T6.1 shows that the plans 1S, 1D, and 1M are an appropriate set of matched
plans.
Single sampling: n = 3.89/.195 = 19.95 ~ 20, Ac = 1, and Re = 2.
Double sampling: n = 2.49/.195 = 12.77 ~ 13, Ac = 0, 1, and Re = 2, 2.
Multiple sampling: n = .917/.195 = 4.70 ~ 5, Ac = #, #, 0, 0, 1, 1, 2, and Re = 2, 2, 2, 3, 3, 3, 3.
Furthermore, the indifference quality level occurs at the following values of defects per
unit with the associated ASN shown:
Single sampling: p.50 = 1.678/20 = .084, ASN = 1(20) = 20.
Double sampling: p.50 = 1.006/13 = .077, ASN = 1.368(13) = 17.8.
Multiple sampling: p.50 = .416/5 = .083, ASN = 3.640(5) = 18.2.
The PQLs and CQLs for these matched plans are the following:

PQL = p.95 CQL = p.10

Single .355/20 = .018 3.890/20 = .195

Double .207/13 = .016 2.490/13 = .192
Multiple .103/5 = .021 .917/5 = .183

The plans shown correspond directly to the MIL-STD-105E, Code F, 2.5 AQL normal plan.
Double and Multiple Sampling by Attributes 141

Measures
Double Sampling
The measures of performance of double-sampling plans are given in Table 6.1. Binomial
or Poisson probabilities are appropriate for their evaluation depending on the sampling
situation. These probability distributions are listed in Table 3.1. As an illustration of appli-
cation of these formulas, suppose the measures of the plan n1 = n2 = 13; a1 = 0, a2 = 1; and
r1 = r2 = 2 are to be evaluated when sampling from a succession of lots of size N = 120.
Calculations are to be made using the binomial distribution in a Type B sampling situation
when the incoming proportion defective is p = .18.

TABLE 6.1
Measures of Double-Sampling Attributes Plans
Measure Formula

Probability of acceptance, Pa Pa = Pa1 + Pa2 , where Pai is Pa for ith sample

r1 -1

(
= F a1 n1 + ) å f (d n ) F (a - d n )
d1 = a1 +1
1 1 2 1 2

ASN Full inspection

ASN = n1 + n2(F(r1 − 1|n1) − F(a1|n1))
Semicurtailed
é r2 - d1 ù
r1 -1 ê p + n2 F a2 - d1 n2( ) ú
ASN c = n1 + å ( f d1 n1 êê )
æ r -d ö
ú
ú
d1 = a1 +1
êë è p ø
(
ê - çç 2 1 ÷÷ F r2 - d1 n2 + 1 ) ú
úû
AOQ Defectives found replaced with good
æ N - n1 ö æ N - n1 - n2 ö
AOQ = pPa1 ç ÷ + pPa2 ç ÷
è N ø è N ø
Defectives not replaced
æ N - n1 ö æ N - n1 - n2 ö
AOQ = pPa1 çç ÷÷ + pPa2 çç ÷÷
è N - n1 p ø è N - n1 p - n2 p ø
Approximate (n/N small)
AOQ ≅ pPa
AOQL (Approximate)a Determine n* and c* as follows:
1. Average cumulative sample sizes to obtain n*.
2. Average all acceptance and rejection numbers to
obtain c*.
Obtain value of y using c*:
y æ N - n* ö
AOQL 
n* èç N ÷ø
ATI ATI = n1P1 + ( n1 + n2 ) P2 + N ( 1 - Pa )
= n1 + n2 ( 1 - Pa1 ) + ( N - n1 - n2 ) ( 1 - Pa )
a Approximation matching double to single plans given in Schilling et al. (1978). Value
of y given in Appendix Table T14.1.
142 Acceptance Sampling in Quality Control

Probability of acceptance:

Pa = F(0|13) + f (1|13)F(0|13)
= C013 .180 ( .82 ) + C113 .181 ( .82 ) éC013 .180 ( .82 ) ù
13 12 13
ë û
= .076 + .216 [.076 ]
= .076 + .016 ( Note: Thus, Pa1 = .76 and Pa2 = .016 )
= .092

Average sample number:

Full inspection
ASN = 13 + 13 ( F(1|13) - F(0|13) )
= 13 + 13 ( .292 - .076 )
= 15.8

Semicurtailed
é2-1 2-1 ù
ASN c = 13 + f ( 113
| )ê + 13 F ( 0|13 ) - F(114
| )ú
ë .18 .18 û
= 13 + .216 ë 5.56 + 13 ( .076 ) - 5.56 ( .253 ) û
é ù

= 14.1

Average outgoing quality (AOQ):

Defectives found replaced

æ 120 - 13 ö æ 120 - 13 - 13 ö
AOQ = .18 ( .076 ) ç ÷ + .18 ( .016 ) ç ÷
è 120 ø è 120 ø
= .012 + .002 = 0.014

Defectives not replaced

æ 120 - 13 ö æ 120 - 13 - 13 ö
AOQ = .18 ( .076 ) ç + .18 ( .016 ) ç
ç 120 - 13 ( .18 ) ÷÷ ç 1220 - 13 ( .18 ) - 13 ( .18 ) ÷÷
è ø è ø
= .012 + .002 = 0.014

Approximate

AOQ @ .18 ( .092 ) @ .017

Average outgoing quality limit (AOQL):

13 + 26
n* = = 19.5 ~ 20
2
0 + 2 +1+ 2
c* = = 1.25 ~ 1
4
Double and Multiple Sampling by Attributes 143

y = .8400 (From Appendix Table T14.1)

.8400 æ 120 - 20 ö
AOQL @ ç ÷ @ .035
20 è 120 ø

Average total inspection (ATI):

ATI = 13 ( .076 ) + ( 13 + 13 )( .016 ) + 120 ( 1 - .092 )

= .99 + .42 + 108.96 = 110.337

These measures are useful in the characterization of this double-sampling plan for p = .18.
Repeated calculations for various values of proportion defective would allow construction
of the curves describing plan performance.

Multiple Sampling
The measures of performance of multiple-sampling plans are given in Table 6.2. They apply,
of course, to double-sampling plans as well. The binomial or Poisson probability distribu-
tions are appropriate for their evaluation depending on the sampling situation involved and
the degree of approximation desired. These probability distributions are listed in Table 3.1.
As an illustration of the application of these formulas, consider the plan:

Stage 1:  n1 = 10   a1 = #   r1 = 2

Stage 2:  n2 = 10   a2 = 0   r2 = 2
Stage 3:  n3 = 10   a3 = 1   r3 = 2

where # denotes no acceptance allowed at the first stage. Suppose the plan is to be evalu-
ated at p = .01 for application to lots of size 350. Calculations are to be made using Poisson
probabilities as an approximation to the binomial for a Type B sampling situation. Note
that for np = (10) (.01) = 0.1,
f ( 0|10 ) = .905 F ( 0|10 ) = .905 1 - F ( 0|10 ) = .095
f ( 110
| ) = .090 F ( 110
| ) = .995 1 - F ( 110
| ) = .005

Probably, the best way to portray the evaluation of a multiple-sampling plan is with a
probability tree as shown in Figure 6.5.
The results of Figure 6.5 give the following probabilities:

Stages Accept, Aj Reject, Rj Terminate, Tj Indecision, Ij

1 0 .005 .005 .995

2 .819 .013 .832 .162
3 .148 .016 .164 0
Total .967 .034 1.001 x

This listing of the probabilities associated with the tree shows the probability of accept-
ing (Aj), rejecting (Rj), terminating (Tj), and indecision (Ij) at each stage. It is constructed
simply as the totals of the acceptance, rejection, and indecision probabilities shown at that
stage of the tree. Termination is the sum of acceptance and rejection and acts as a check
since the termination column must sum to one.
144 Acceptance Sampling in Quality Control

TABLE 6.2
Measures of Multiple-Sampling Attributes Plans
Measure Formula
k
Probability of acceptance, Pa Pa = åA
j =1
j

where Aj is the probability of acceptance on the jth stage

(for explicit formula, see Table 6.3)
ASN Full inspection
k j

ASN = åån T where T is the probability of termination on

j =1 m=1
m j j

the jth stage

(for semicurtailed, see Table 6.3)
AOQ Defectives found replaced with good
k æ N-
å
nm ö÷
j
ç
AOQ =
j =1
åç
è N
m=1
÷ pA j
ø
Defectives not replaced
k æ N-
å
nm ö÷
j
ç
AOQ = çå m=1
÷ pA j
å
j
j =1 ç N - p nm ÷
è m=1 ø
Approximate (n/N small)
AOQ ≅ pPa
AOQL (Approximate)a Determine n* and c* as follows:

1. Average cumulative sample sizes to obtain n*.

2. Average all acceptance and rejection numbers to obtain c*.

(Use –1 when acceptance is not allowed at a stage.)

Obtain value of y using c*:
y æ N -n* ö
AOQL 
n * èç N ÷ø
k j

ATI ATI = N ( 1 - Pa ) + åån A

j =1 m=1
m j

a Approximation matching multiple to single plans given in Schilling et al. (1978). Value of y given
in Appendix Table T14.1.

Using the tree and the formulas of Table 6.2 for p = .01, it is possible to obtain the following:

Probability of acceptance:

Pa = A1 + A2 + A3
= 0 + .819 + .148 = .967

Average sample number:

ASN = n1T1 + ( n1 + n2 ) T2 + ( n1 + n2 + n3 ) T3
= 10 ( .005 ) + 20 ( .833 ) + 30 ( .164 ) = 21.6
Double and Multiple Sampling by Attributes 145

Stages Start

0 1 ≥2
(0.905) (0.090) (0.005)
1 Indecision Indecision Reject
0.905 0.090 0.005

0 1 ≥2 0 ≥1
(0.905) (0.090) (0.005) (0.905) (0.095)
Accept Indecision Reject Indecision Reject
2
0.819 0.081 0.005 0.081 0.009

0 ≥1 0 ≥1
(0.905) (0.095) (0.905) (0.095)

3 Accept Reject Accept Reject

0.074 0.008 0.074 0.008

FIGURE 6.5
Probability tree evaluating a multiple sampling plan.

Average outgoing quality:

Defectives found replaced with good

éæ N - n1 ö æ N - n1 - n2 ö æ N - n1 - n2 - n3 ö ù
AOQ = p êç ÷ A1 + ç ÷ A2 + ç ÷ A3 ú
ë è N ø è N ø è N ø û
éæ 350 - 10 ö æ 350 - 20 ö æ 350 - 30 ö ù
= .01 êç ÷0 + ç ÷ .819 + ç ÷ .148 ú
ëè 350 ø è 350 ø è 350 ø û

= .01[.772 + .135] = .009

Defectives not replaced

éæ N - n1 ö æ N - n1 - n2 ö æ N - n1 - n2 - n3 ö ù
AOQ = p êç ÷ A1 + çç ÷÷ A2 + çç ÷÷ A3 ú
êëè N - pn1 ø è N - p ( n1 + n2 ) ø è N - p ( n1 + n2 + n3 ) ø úû
éæ 350 - 10 ö æ 350 - 20 ö æ 350 - 30 ö ù
= .01 êç ÷÷ 0 + çç ÷÷ .819 + çç ÷÷ .148 ú
ç
êëè 350 - .01 ( 10 ) ø è 350 - .01 ( 20 ) ø è 350 - .01 ( 30 ) ø úû

= .01[.773 + .135] = .009

Approximate

AOQ = pPa = .01 ( .967 ) = .0097

146 Acceptance Sampling in Quality Control

Average outgoing quality limit:

10 + 20 + 30
n* = = 20
3
-1 + 0 + 1 + 2 + 2 + 2 6
c* = = =1
6 6
y æ N - n* ö .8400 æ 350 - 20 ö
AOQL = = ç ÷ = .040
n* èç N ÷ø 20 è 350 ø
Average total inspection:

ATI = N ( 1 - Pa ) + n1 A1 + ( n1 + n2 ) A2 + ( n1 + n2 + n3 ) A3
= 350 ( 1 - .967 ) + 10 ( 0 ) + 20 ( .819 ) + 30 ( .148 )
= 11.55 + 16.38 + 4.44 = 32.37

These values measure the performance of this multiple-sampling plan when p = .01. From
calculations at other levels of proportion defective, the relevant curves characterizing the
plan can be drawn.
The OCs and other measures of double- and multiple-sampling plans can also be com-
puted using the control table concept originated by the Statistics Research Group (1948).
The control table for a plan having acceptance numbers

Sample Ac Re

1 # 2
2 0 3
3 2 4
4 3 4

is shown in Figure 6.6. Sample numbers are listed across the top, while the accumulated
­number of defectives is shown on the side. The squares represent possible events in the opera-
tion of the sampling plan. The boundary of the figure is comprised of the squares leading to
acceptance or rejection. If no acceptance decision is possible at a state, the double boundary
square is omitted as in the first sample of Figure 6.6. The top right square in Figure 6.6, for
example, represents accumulation of at least four defectives on the fourth sample. A double
square represents a state at which termination of the plan would occur with acceptance or
rejection. Bottom double squares show acceptance, whereas top double squares show rejec-
tion. Since top square probabilities indicate rejection, they are cumulative at or exceeding the
number shown. They are shown as capital letters (P), as opposed to individual probabilities
shown as small letters (p). They are the only squares that use cumulative probabilities.
To fill out the figure, proceed as follows:

1. Under sample 1, fill in the appropriate probabilities of each event.

Defective Probability Action Symbol for State

0 f(0|n1) No decision p01

1 f(1|n1) No decision p11
≥2 1 − F(1|n1) Reject P21
Double and Multiple Sampling by Attributes 147

Sample
1 2 3 4
Defectives
p12(1 – F(2|n3)) +
p22(1 – F(1|n3)) = p33(1 – F(1|n4)) =
4
P43 P44

p01(1 – F(2|n2)) + p12 f(2|n3) +

p22 f(1|n3) = p33 f(0|n4) =
3 p11(1 – F(1|n2)) =
P32 p33 p34

p01 f(2|n2) + p12 f(1|n3) +

2 1 – F(1|n1) = p11 f(1|n2) = p22 f(0|n3) =
P21 p22 p23

Plan
1 p01 f(1|n2) +
f(1|n1) = n Ac Re
p11 f(0|n2) =
p11 p12 n1 # 2
n2 0 3
n3 2 4
0 f(0|n1) = p01 f(0|n2) = n4 3 4

p01 p02

Sample
1 2 3 4 Total

Accept Aj A1 = 0 A2 = p02 A3 = p23 A4 = p34 ∑Aj

Reject Rj R1 = P21 R2 = P32 R3 = P43 R4 = P44 ∑Rj

Terminate Tj T1 = A1 + R1 T2 = A2 + R2 T3 = A3 + R3 T4 = A4 + R4 1.0

Indecision Ij p01+ p11 p12 + p22 p33 0 x

FIGURE 6.6
Control table format.

2. Under sample 2, fill in the appropriate probabilities of each event as the sum of the
joint probabilities of events leading to that state.

Approach
Defective First Sample Second Sample Symbol Probability Action

0 0 0 p02 = p01f(0|n2) Accept

1 0 1 p12 = p01f(1|n2) + p11f(0|n2) No decision
1 0
2 0 2 p22 = p01f(2|n2) + p11f(1|n2) No decision
1 1
3 0 ≥3 P32 = p01(1 − F(2|n2)) + p11(1 − F(1|n2)) Reject
1 ≥2
148 Acceptance Sampling in Quality Control

3. Under sample 3, fill in the appropriate probabilities of each event as the sum of the
joint probabilities of events leading to that state.

Approach
Second Third
Defective Sample Sample Symbol Probability Action

2 1 1 p23 = p12  f(1|n3) + p22   f(0|n3) Accept

2 0
3 1 2 p33 = p12  f(2|n3) + p22   f(1|n3) No decision
2 1
4 1 ≥3 P43 = p12(1 − F(2|n3)) + p22(1 − F(1|n3)) Reject
2 ≥2

4. Under sample 4, fill in the appropriate probabilities of each event as the sum of the
joint probabilities of events leading to that state.

Approach
Defective Third Sample Fourth Sample Symbol Probability Action
3 3 0 p34 = p33  f(0|n4) Accept
4 3 ≥1 P44 = p33(1 − F(0|n4)) Reject

The probability of acceptance is simply the sum of the probabilities shown in the squares
leading to acceptance:

Pa = p02 + p23 + p34

The probability of rejection can similarly be found as the sum of the probabilities of the
squares leading to rejection:

P ( rej ) = P21 + P32 + P43 + P44

Clearly, the probability of acceptance and rejection must add to one.

A summary table similar to the listing of probabilities for the probability tree approach to
the evaluation of the multiple plan may be developed for the control table and is shown at the
bottom of Figure 6.6. Using the summary table, it is possible to evaluate the measures given
in Table 6.2 just as was done with the probability tree. The control table may also be used to
evaluate explicit formulas for probability of acceptance and ASN as given in Table 6.3.
As an example of the control table approach, consider again the multiple plan:

n Ac Re

10 # 2
10 0 2
10 1 2

The appropriate control table is shown in Figure 6.7.

The results of the summary table are within the rounding error of those for the listing
of probabilities obtained from the probability tree and will give the same results for the
Double and Multiple Sampling by Attributes 149

TABLE 6.3
Explicit Formulas for Measures of Multiple-Sampling Plans
Measure Formula
r1 -1 é r2 -1 é
Probability of
acceptance, Pa
(
Pa = F a1 n1 + ) å f ( d n ) êêëF ( a - d n ) + å f ( d n ) êêë
d1 = a1 +1
1 1 2 1 2
d2 = a2 -d1 +1
2 2

é r2 -1 é
ê
(
 ê F a j - Dj-1 n j + ) å ( )
f dj nj ê
ê
ë d j = a j -D j -1 +1 ë
é rk -1 -1 é ù ù ù
(
 ê F ak -1 - Dk -2 nk -1 +
êë
) å f (d k -1 ) ( )
nk -1 ê F ak - Dk -1 nk  ú ú ú
ê ú ú ú
dk -1 = ak -1 -Dk - 2 +1 ë û úû úû

å
j
where Dj = di and cumulative probabilities with negative arguments are taken to be zero
i =1

(Use ai = −1 if no acceptance is allowed at stage i.)

ASN Full inspection
ASN = n1 + I1[n2 + I2[n3 + I3[⋯[nk + 0]⋯]]]
where Ij is the probability of no decision at stage j
Semicurtailed inspection (equal sample sizes)
k k -1 rj -1

ASN c = å ( nj ) Tj + åå ( nj ) pij (1 - F ( rj+1 - i - 1n - 1))

j =1 j =1 i = a j +1
k rj -1 -1

- å å ( nj ) p ( ) (1 - F ( r - i - 1n - 1))
j =2 i = a j -1 +1
i j -1 j

k -1 rj -1

+ ååp ij
rj+1 - i
p
( (
1 - F rj+1 - i n ))
j =1 i = a j +1

where Tj and pij are taken from the control table and the first sample is always fully
inspected. (Use ai = −1 if no acceptance is allowed at a stage.)

measures as given in Table 6.2. Using the control table to evaluate the explicit formulas
given in Table 6.3 for this plan, we obtain the following:
Probability of acceptance:
r1 -1 é r2 -1 ù
Pa = F ( #|n1 ) + å
d1 = a1 + 1
f ( d1|n1 ) ê F ( a2 - d1|n2 ) +
êë d2 = a2 - d1 + 1
å
f ( d2|n2 ) éë F ( a3 - D2|n3 ) ùû ú
úû
2 -1 é 2 - 1 ù
= F ( -110
| )+ å
d1 =-1+ 1
f ( d1|10 ) ê F ( 0 - d1|10 ) +
êë d2 = 0 - d1 + 1
å
f ( d2|10 ) éë F ( 1 - D2|10 ) ùû ú
úû
1 é 1 ù
= 0+ å
d1 = 0
f ( d1|10 ) ê F ( 0 - d1|10 ) +
êë d2 =1- d1
å
f ( d2|10 ) éë F ( 1 - D2|10 ) ùû ú
úû
= 0 + f ( 0|10 ) F ( 0|10 ) + f ( 0|10 ) f ( 110 | ) F ( 1 - 110 | )
+ f ( 110
| ) F ( -110
| ) + f ( 110
| ) f ( 0|10 ) F ( 1 - 110
| )
+ f ( 110
| ) f ( 110
| ) F ( -110
| )
= 0 + ( .905 )( .905 ) + ( .905 )( .090 )( .905 )
+ 0 + ( .090 )( .905 )( .905 ) + 0
= .
819 + . 074
 + .074

A2 A3
= .967
150 Acceptance Sampling in Quality Control

Sample
1 2 3
Defectives
p01(1 – F(1|10)) +
p11(1 – F(0|10)) = p12(1 – F(0|10)) =
2 0.905 (0.005) + 0.163(0.095) =
0.090 (0.095) =
P21 = 0.005 P22 = 0.013 P23 = 0.015

p01 f (1|10) +
p11 f (0|10) = p12 f (0|10) =
1 0.163(0.905) =
(0.905) (0.090) +
(0.090) (0.905) =
p11 = 0.090 p13 = 0.148
P12= 0.163

p01 f (0|10) =
0 (0.905) (0.905) =

p01 = 0.905 p0 2 = 0.819

Sample
1 2 3 Total

Accept Aj A1 = 0 A2 = 0.819 A3 = 0.148 0.967

Reject Rj R1 = 0.005 R2 = 0.013 R3 = 0.015 0.033

Terminate Tj T1 = 0.005 T2 = 0.832 T3 = 0.163 1.000

Indecision Ij I1 = 0.995 I2 = 0.163 I3 = 0 x

FIGURE 6.7
Control table illustration.

Average sample number:

Full inspection

ASN = n1 + I1 éë n2 + I 2 [ n3 ]ùû
= 10 + .995 éë10 + .163 [10 ]ùû

= 21.57

Semicurtailed inspection

3 2 rj -1 3 rj-1 -1

ASN c = å j =1
(10 j ) Tj + åå
j =1 i = a j + 1
(
(10 j ) pij 1 - F ( rj +1 - i - 19
|) - ) å å (10 j ) p ( ) (1 - F ( r - i - 19
j = 2 i = a j-1 + 1
i j -1 | )) j

2 rj -1
æ rj +1 - i ö
+ å å p çè ij
p ø
(
÷ 1 - F ( rj +1 - i|10 ) )
j =1 i = a j + 1
Double and Multiple Sampling by Attributes 151

= [10T1 + 20T2 + 30T3 ] + éë10 p01 ( 1 - F ( 19

| ) ) + 10 p11 ( 1 - F ( 0|9 ) ) + 20 p12 ( 1 - F ( 0|9 ) ) ùû

- éë 20 p01 ( 1 - F ( 19
| ) ) + 20 p11 ( 1 - F ( 0|9 ) ) + 30 p12 ( 1 - F ( 0|9 ) ) ùû
é æ2ö æ1ö æ1ö ù
+ ê p01 ç ÷ ( 1 - F ( 2|10 ) ) + p11 ç ÷ ( 1 - F ( 110
| ) ) + p12 ç ÷ ( 1 - F ( 110
| ) )ú
êë è p ø èpø èpø úû
= éë10( .005 ) + 20( .832 ) + 30( .163 ) ùû + éë10( .905 )( .0038 ) + 10 ( .090 )( .0861) + 20( .163 )( .0861) ùû
- éë 20 ( .905 )( .0038 ) + 20 ( .090 )( .0861) + 30 ( .163 )( .0861) ùû
é æ 2 ö æ 1 ö æ 1 ö ù
+ ê( .905 ) ç ÷ ( .0002 ) + .090 ç ÷ ( .0047 ) + .163 ç ÷ ( .0047 ) ú
ë è .01 ø è .01 ø è .01 ø û
= 21.51

Further Considerations
Unity values np presented in Appendix Table T6.1 were derived by Schilling and Johnson
(1980) using the theory of unity values as presented by Duncan (1974, pp. 187–188). They
are based on the Poisson distribution and can be used to approximate binomial-sampling
plans where the Poisson approximation to the binomial distribution applies. Since the
probability of acceptance and the ASN can be shown to be a function of np for a given set
of acceptance criteria and ratio of subsample sizes, it is possible to vary np1 and np2 in such
a way that the ratio R = p2/p1 remains unchanged while the value of n changes. Thus, any
member of the set of plans having operating ratio R may be used to generate unity values,
values of np when n = 1, by simply dividing the values of np associated with its OC by n.
A similar argument holds for the values ASN/n.

Software Applications
Excel
Use the Excel template Double Sampling Plan for Attributes (including hypergeometric).xlsm to
evaluate a double-sampling plan. For Type B plans based on the Poisson or binomial, you
can also compare the double-sampling plan to the corresponding single-sampling plan.
Furthermore, since the Type B plans assume an infinite lot size, it is prudent to use a very
large lot size, for example, 1,000,000, to get accurate probabilities.
Using the earlier example of a double-sampling plan that was determined to be n1  =
n2 = 88, c1 = 1, c2 = 4, r1 = 4, and r2 = 5, we enter the values into cells C7, C8, C10, C11,
and C13, respectively (cell C14 is automatically computed), in the first tab entitled
Calculations—Hypergeometric (Figure 6.8). Enter 1,000,000 into cell C5 for the infinite lot
size and indicate that rectifying inspection is to be used in cell C17 (this will allow the AOQ
and ATI curves to be created). All of this information gets carried over to the tab entitled
Calculations—Poisson, Binomial (Figure 6.9). On this latter tab, we will indicate that we
 152

FIGURE 6.8
Excel template Double Sampling Plan for Attributes (including hypergeometric).xlsm Type A plan entry form for derived double sampling plan n1 = n2 = 88, c1 = 1, c2 = 4,
Acceptance Sampling in Quality Control

r1 = 4, and r2 = 5.
 Double and Multiple Sampling by Attributes

FIGURE 6.9
Excel template Double Sampling Plan for Attributes (including hypergeometric).xlsm Type B plan entry form for derived double sampling plan n1 = n2 = 88, c1 = 1, c2 = 4, r1 = 4,
and r2 = 5 with added single-sampling plan n = 134 and c = 3.
153
154 Acceptance Sampling in Quality Control

Operating characteristic curve

Double sampling plan
Pa1 = P(accept on first sample), Pr1 = P(reject on first sample),
Pa = P(accept on both samples)
Assumed distribution = binomial
100% 100%

90% 90%

80% Sampling 80%

plan:
Probability of accepting lot (%)

Probability of rejecting lot (%)

70% N = 1000 70%
n1 = 88
n2 = 88
60% c1 = 1 60%
c2 = 4
50% r1 = 4 50%
r2 = 5
40% 40%

30% 30%

20% Pa1 = P(x1 < c1|n1,c1) 20%

Pa = Pa1 + Pa2
10% Pr1 = P(x1 > r1|n1,c1) 10%

0% 0%
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24%
Fraction defective in lot, p

FIGURE 6.10
Excel Type B OC curve for derived double sampling plan n1 = n2 = 88, c1 = 1, c2 = 4, r1 = 4, and r2 = 5 with added
single-sampling plan n = 134 and c = 3.

wish to use the binomial probability distribution by entering 2 into cell F4. Note that on
this tab we can also enter the corresponding single sampling plan n = 134 and c = 3 into
cells B20 and B22 (cell B24 is automatically computed).
The Type B OC, AOQ, ASN, and ATI curves based on the binomial are shown in
Figures 6.10 through 6.13, respectively.

Minitab
Minitab does not produce either double- or multiple-sampling plans.

Statgraphics
Statgraphics has the ability to produce single-, double-, and multiple-sampling plans.
However, they are only available through SPC > Acceptance Sampling > Attributes > MIL-
STD-105E (ANSI Z1.4/ISO 2859), that is, the produced plans agree with those found in
these published attributes sampling standards. Since these standards will be discussed in
Chapter 11, Statgraphics will be demonstrated for these types of plans at the end of that
chapter.
Double and Multiple Sampling by Attributes 155

Average outgoing quality curve

AOQ(R) = [Avg. outgoing quality for] double sampling plan with rectification
Assumed distribution = binomial
1.4%

1.2%
Average fraction defective of outgoing lots (%)

1.0% Sampling
plan:
N = 1000
0.8% n1 = 88
n2 = 88
c1 = 1
c2 = 4
0.6% r1 = 4

0.4%
AOQ(R) = p[(Pa1((N – n1)/N) + Pa2(N – n1 – n2))/N]
AOQL = 0.0132
0.2%

0.0%
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24%
Fraction defective in incoming lots, p

FIGURE 6.11
Excel Type B AOQ curve for derived double-sampling plan n1 = n2 = 88, c1 = 1, c2 = 4, r1 = 4, and r2 = 5 with recti-
fication of rejected lots.

Average sample number curve

Double sampling plan
ASN = [Avg sample number for] complete inspection, ASNc = curtailed inspection
Assumed distribution = binomial
160
Sampling
plan:
140 N = 1000
n1 = 88
Average sample number (ASN)

20 n2 = 88
c1 = 1
c2 = 4
100
r1 = 4

80

60

40
ASN = n1 + n2[1 – (Pa1 + Pr1)]
20 ASNc = n1(Pa1 + Pr1) + SP(n1,j)[n2 – (n2 – n1)(1 – P(n2 – n1 – 1,c2 – j))+
((c2 – j+1)/p)(1 – P(n2 – n1,c2 – j+1))]
0
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24%
Fraction defective in lot, p

FIGURE 6.12
Excel Type B ASN curves for derived double-sampling plan n1 = n2 = 88, c1 = 1, c2 = 4, r1 = 4, and r2 = 5 showing
complete and curtailed inspection.
156 Acceptance Sampling in Quality Control

Average total inspection curve

Double sampling plan
ATI = [Avg sample number for] double sampling plan
Assumed distribution = binomial
1200

1000
Average total inspection (ATI)

Sampling
800 plan:
N = 1000
n1 = 88
600 n2 = 88
c1 = 1
c2 = 4
400 r1 = 4
r2 = 5

200
ATI = n1Pa1 + (n1 + n2)Pa2 + N(1 – Pa)
0
0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24%
Fraction defective in lot, p

FIGURE 6.13
Excel Type B ATI curve for derived double-sampling plan n1 = n2 = 88, c1 = 1, c2 = 4, r1 = 4, and r2 = 5.

Problems
1. Construct double-sampling plans to the following specifications given in proportion
defective with producer’s risk .05 and consumer’s risk .10.
a. PQL = .039, CQL = .210
b. PQL = .029, CQL = .130
c. PQL = .019, CQL = .060
2. Construct the multiple-sampling plans corresponding to those obtained in Problem 1.
Comment on the match.
3. Plot the Type B OC curve for the following double-sampling plan using Pa = .95, .50,
and .10 at a minimum for plotting positions.
n: 8, 8
Ac: 0, 1
Re: 2, 2
4. Plot the Type B OC curve for the following multiple-sampling plan using Pa = .95, .50,
and .10 at a minimum for plotting positions.
n: 3, 3, 3, 3, 3, 3, 3
Ac: #, #, 0, 0, 1, 1, 2
Re: 2, 2, 2, 3, 3, 3, 3
Double and Multiple Sampling by Attributes 157

5. If lots are received in quantities of 1000, obtain ASN, AOQ, and ATI at the ­minimum
plotting positions for the plan given in Problem 3 and draw the curves.
6. If lots are received in quantities of 1000, obtain ASN, AOQ, and ATI at the minimum
plotting positions for the plan given in Problem 4 and draw the curves.
7. At present, the single-sampling plan n = 35, c = 3 is being used in the acceptance
inspection of incoming material from a very good supplier. What double and multiple
plans may be substituted? How much would be gained thereby?
8. Use the approximation for determining AOQL to find what single-sampling plan
matches the double- and multiple-sampling plans of Problems 3 and 4. Is this
confirmed by the Schilling–Johnson table?
9. What is the ASN under curtailed inspection at p = .025 for the plan given in Problem 3?
Using the formula for a double-sampling plan, is the curtailment worthwhile at this
fraction defective?
10. Regard the plan given in Problem 3 as a multiple-sampling plan and construct the
control table at p = .025 to evaluate the probability of acceptance and the ASN.

References
Cameron, J. M., 1952, Tables for constructing and for computing the operating characteristics of
­single sampling plans, Industrial Quality Control, 9(1, Part I): 37–39.
Dodge, H. F. and H. G. Romig, 1959, Sampling Inspection Tables, Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York.
Duncan, A. J., 1986, Quality Control and Industrial Statistics, 5th ed., Richard D Irwin, Homewood, IL.
Schilling, E. G. and L. I. Johnson, 1980, Tables for the construction of matched single, double, and
multiple sampling plans with application to MIL-STD-105D, Journal of Quality Technology, 12(4):
220–229.
Schilling, E. G., J. H. Sheesley, and P. R. Nelson, 1978, GRASP: A general routine for attribute sam-
pling plan evaluation, Journal of Quality Technology, 10(3): 125–130.
Statistical Research Group, Columbia University, 1948, Sampling Inspection, McGraw-Hill, New York.
U.S. Department of the Army, 1953, Master Sampling Plans for Single, Duplicate, Double, and Multiple
Sampling, Manual No 2, Chemical Corps Engineering Agency/Army Chemical Center,
Edgewood, MD.
7
Sequential Sampling by Attributes

Single, double, and multiple plans assess one or more successive samples to determine lot
acceptability. The most discriminating acceptance sampling procedure involves making a
decision as to disposition of the lot or resample successively as each item of the sample is
taken. Called sequential sampling, these methods may be regarded as multiple-sampling
plans with sample size one and no upper limit on the number of samples to be taken. It can
be shown that the sequential approach provides essentially optimum efficiency in sam-
pling that is an average sample number (ASN) as low as possible (Wald 1947, p. 35). For
example, in comparing average sample sizes for plans matched to the Military Standard
105E (MIL-STD-105E) single-sampling plan n = 50, c = 2, we have at p = .017 (chosen as the
95th percentage point).

Plan ASN

Single 50
Double 43
Multiple 35
Sequential 33.5

Sequential plans are often applied where sample size is critical so that a minimum sample
must be taken. They are somewhat harder to administer than multiple-sampling plans
since in specific applications the amount of inspection effort is not determined until the
sample is taken. The possibility of taking one sample at a time must exist; in some opera-
tions, this would be exceedingly difficult or impossible. Furthermore, the operating proce-
dure requires an astute and trusted inspector since it is somewhat more demanding than
single, double, or multiple sampling.

Operation
Under sequential sampling, samples are taken, one at a time, until a decision is made
on the lot or process sampled. After each item is taken, a decision is made to (1) accept,
(2) reject, or (3) continue sampling. Samples are taken until an accept or reject decision
is made. Thus, the procedure is open-ended, with the sample size not being determined
until the lot is accepted or rejected. The ASN of the plan provides a benchmark as to the
expected sample size in any given application.

159
160 Acceptance Sampling in Quality Control

y2 = sk + h2
dk
Reject

y1 = sk – h1
h2
Continue

0 k
–h1 Accept

FIGURE 7.1
Sequential acceptance plot.

The plan is often implemented using a chart, as shown in Figure 7.1, in which the cumulative
number of defectives found is plotted against the number of individual samples taken, where

k is the number of sample items taken

dk is the number of defectives found by the kth sample item
Y2 = sk + h2, which is the reject limit at kth sample
Y1 = sk − h1, which is the acceptance limit at kth sample
h1, h2 are intercepts
s is the slope (not a standard deviation)

When the plot of the cumulative number of defectives found crosses the acceptance or
rejection limit lines, the lot is disposed of appropriately. Clearly, no acceptance is possible
until the acceptance line Y1 crosses the k-axis. The operation of the plan is shown diagram-
matically in Figure 7.2.
The procedure described is called unit sequential sampling since items are drawn unit
by unit. Occasionally, group sequential procedures are used in which groups of items are
successively drawn (e.g., 10 at a time), inspected, and assessed against the acceptance and
rejection limits at the successive accumulated values of k. This is often done for inspection
convenience. When the physical circumstances of the inspection do not dictate a group size,
an expeditious approach suggested by Cowden (1957) is to make the group size equal to the

Start

Sample 1 unit

dk ≤ Y1 Y1 < dk < Y2 dk ≥ Y2

Accept Resample Reject

FIGURE 7.2
Procedure for sequential sampling by attributes.
Sequential Sampling by Attributes 161

number of samples necessary to allow the first possibility of acceptance. That is, the value of
k is just beyond the intersection of the acceptance line Y1 with the k-axis in Figure 7.1. Group
sequential plans are often listed in the form of a multiple-sampling plan showing cumula-
tive sample size with acceptance and rejection numbers for each group. Of course, the listing
must remain open-ended. When rounding acceptance and rejection numbers obtained from
the sequential acceptance plot, it is desirable to round the acceptance number upward and
the rejection number downward to minimize the difference between the group sequential
plan and the unit sequential plan from which it is derived. Of course, when the acceptance
numbers and rejection numbers are integers for the unit sequential plan at the successive
values of k corresponding to the multiples of group size, the measures of the group and unit
sequential plans will correspond. This will occur when h1, h2, and 1/s are integers.

Selection
Sequential sampling plans have been tabulated by the Statistical Research Group (1945).
A table of plans based on their results is given in Appendix Table T7.1, which shows the
following characteristics for a variety of plans when α = .05 and β = .10:

p1 is the producer’s quality level (PQL)

p2 is the consumer’s quality level (CQL)
h1 is the acceptance intercept
h2 is the rejection intercept
s is the slope
n0 is the number of samples prior to possibility of acceptance
n1 is the number of samples prior to possibility of rejection
np1 is the ASN at PQL
ns is the ASN at proportion defective equal to s
np2 is the ASN at CQL

Using these values, sequential plans for attributes inspection can be readily set up and
characterized.
Formulas for the construction and evaluation of sequential plans for arbitrary values of
p1, p2, α, and β have been derived by Wald (1947) and the Statistical Research Group (1945).
The formulas are as follows:

log éë( 1 - a )/b ùû

h1 =
log ( p2 /p1 ) + log éë( 1 - p1 )/( 1 - p2 ) ùû
log éë( 1 - b ) /a ùû
h2 =
log ( p2 /p1 ) + log éë( 1 - p1 ) / ( 1 - p2 ) ùû
log éë( 1 - p1 )/( 1 - p2 ) ùû
s=
log ( p2 /p1 ) + log éë( 1 - p1 )/( 1 - p2 ) ùû

Either common or natural logarithms can be used in these computations provided they are
consistent.
162 Acceptance Sampling in Quality Control

Then, the acceptance and rejection lines are determined as

Y1 = sk - h1 ( acceptance ) Y2 = sk + h2 ( rejection )

and then plotted as shown in Figure 7.1. These formulas are sometimes expressed as

b b
h1 = =
g1 + g 2 G
a a
h2 = =
g1 + g 2 G
g2 g
s= = 2
g1 + g 2 G

for computational convenience, where

1-b
a = log
a
1- a
b = log
b
p
g1 = log 2
p1
1 - p1
g 2 = log
1 - p2
G = g1 + g 2

Appendix Table T7.2 gives values of a and b tabulated for selected values of α and β.
Appendix Table T7.3 shows values of g1 and g2 for selected values of p1 and p2.
For example, suppose a sequential plan is desired having p1 = .018, p2 = .18, α = .05, and
β = .10. Appendix Table T7.1 does not list such a plan so the formulas must be used. Here

log éë( 1 - .05 )/.10 ùû

h1 =
log ( .18/.018 ) + log éë( 1 - .018 )/( 1 - .18 ) ùû
log 9.5
= = 0.907
log 10 + log 1.1976
log éë( 1 - .10 )/.05 ùû
h2 =
log ( .18/.018 ) + log éë( 1 - .018 )/ ( 1 - .18 ) ùû
log 18
= = 1.164
log 10 + log 1.1976
log éë( 1 - .018 )/( 1 - .18 ) ùû
s=
log ( .18/.018 ) + log éë( 1 - .018 )/( 1 - .18 ) ùû
log 1.1976
= = 0.0726
log 10 + log 1.1976
Sequential Sampling by Attributes 163

dk
Reject y2
2

1 Continue

y1
0
1 2 3 4 5 6 7 8 9 10 11 12 k
Accept
−1

FIGURE 7.3
Sequential graph.

so the lines of acceptance and rejection are

Y1 = 0.0726k - 0.907

Y2 = 0.0726k + 1.164

The plot appears as Figure 7.3 that also shows the plot of sample results if the second and
fifth items were defective leading to rejection and cassation of sampling at the fifth item
sampled.

Measures
Measures of selected values of the ASN are given for the plans in Appendix Table T7.1.
These include n0 , n1 , np1 , ns , and np2 . The operating characteristic (OC) curve may be
sketched from the given values of α, β, p1, and p2. They are based on Type B sampling as are
all of the measures given here.
General formulas exist for probability of acceptance and ASN. To use these formulas, we
employ the auxiliary variable x, where − ∞ < x < ∞, x ≠ 0; then for any arbitrarily selected
value of x, a point (p, Pa) on the OC curve can be calculated (Wald 1947) as

1 - éë( 1 - p2 ) / ( 1 - p1 ) ùû
x

p=
( p2/p1 ) - éë( 1 - p2 ) / ( 1 - p1 ) ùû
x x

with

x
éë( 1 - b ) /a ùû - 1
Pa = x x
éë( 1 - b ) /a ùû - éëb/ ( 1 - a ) ùû

164 Acceptance Sampling in Quality Control

Given the combination p and Pa, we have the general formula for the ASN

Pa log éëb/ ( 1 - a ) ùû + ( 1 - Pa ) log [(1 - b)/a ]

ASN =
p log ( p2 /p1 ) + ( 1 - p ) log éë( 1 - p2 )/( 1 - p1 ) ùû

and for large lot size relative to sample size

AOQ = pPa

Specifically, when x = −∞, −1, 0, 1, ∞, the formulas given in Table 7.1 are obtained.
For example, for the plan p1 = .018, p2 = .180, α = .05, and β = .10 where it was found that
h1 = 0.907, h2 = 1.164, and s = 0.0726, the following measures are obtained using Table 7.1:

p Pa ASN AOQ

0 1 13.5 0
.018 .95 14.7 .017
.0726 .562 15.7 .041
.18 .10 8.9 .018
1 0 1.3 0

These values are usually sufficient for a crude sketch of the OC, ASN, and average ­outgoing
quality (AOQ) curves.

TABLE 7.1
Sequential Sampling by Attributes for
Proportion Defective Points on the Operating
Characteristic, Average Sample Number, and
Average Outgoing Quality Curves
p Pa ASN AOQ

0 1 h1/s 0
p1 1−α (1 - a ) h1 - ah2 (1 − α)p1
s - p1

h2 h1 h2 sh2
s
h1 + h2 s (1 - s ) h1 + h2

p2 β
(1 - b ) h2 - bh1 βp2
p2 - s

1 0 h2/(1 − s) 0
Sequential Sampling by Attributes 165

Sequential Sampling for Defects per Unit

Occasionally, sequential sampling procedures are required for defects per unit. Note that
the unit employed need not be an individual piece but may be several pieces considered
together. MIL-STD-105E is in terms of “defects per hundred.” The following theory applies
whatever the unit as long as it is defined beforehand.
A sequential chart for defects per unit is much like the chart for proportion defective
in that it plots the sum of the defects found against k, the sample number. The PQL and
CQL are, of course, in terms of mean defects per unit, μ1 and μ2, respectively, where μ2 > μ1.
The parameters for the decision lines are as follows using common logarithms:

log éë( 1 - a ) /b ùû
h1 =
log m 2 - log m1
log éë( 1 - b ) /a ùû
h2 =
log m 2 - log m1
m 2 - m1
s=
2.3026 ( log m 2 - log m1 )

Values of operating parameters for such a plan are given in Table 7.2.
The computer is an obvious ally in the application of sequential plans. It can internal-
ize the appropriate cumulative data and evaluate the acceptance or rejection unit by unit
(or group by group) as the data are collected. The operator may be alerted by various sig-
naling devices, or by printing the sequential diagram as needed. Selection and evaluation
of sequential plans are also facilitated by the ease of calculation that the computer supplies.
Thus, the future of sequential sampling is assured by the computer and more extensive
application can be expected.

TABLE 7.2
Sequential Sampling by Attributes for Defects
per Unit Points on the Operating Characteristic
and Average Sample Number Curves
Mean Defects
per Unit Pa ASN

0 1 h1/s
μ1 1−α (1 - a ) h1 - ah2
s - m1
h2 h1 h2
s
h1 + h2 s
μ2 β (1 - b ) h2 - bh1
m2 - s

Note: The form of the chart and its operation are the same
as that shown for proportion defective.
166 Acceptance Sampling in Quality Control

The sequential methods presented here are based on the likelihood ratio test of the sim-
ple hypothesis:

H 0 : p¢ = p1

H1 : p¢ = p2

The theoretical development and application of sequential methods will be found in Wald
(1947) and Wetherill (1986). Proofs associated with these procedures are fairly straightfor-
ward and are developed and presented in detail in these texts.

Software Applications
Excel
Use the Excel template Sequential Sampling.xlsm to generate and evaluate a sequential
sampling plan. To illustrate its use, use the familiar example α = .05, β = .10, p1 = .018,
and p2 = .18, and enter the values into cells B5, B6, B7, and B8, respectively (Figure 7.4).
Based on the formulas, we know that h1 = 0.907, h2 = 1.164, k = 1.078, and s = .0726.
These values agree with those seen in the Excel template in cells B10, B11, B12, and B13.
The accept and reject line equations become Y1 = −.907 + .0726n and Y2 = 1.164 + .0726n.
The measures for this sequential plan based on the equations in Table 7.1 are computed
by the template and are shown in the following, which agree with the estimates previ-
ously presented:

Pa p ASN AOQ

1 0 13.5 0
1−α= 0.95 0.018 = p1 14.7 0.017
h2/(h1+h2) = 0.56 0.0726 =s 15.7 0.041
β= 0.1 0.18 = p2 8.9 0.018
0 1 1.3 0

Note that the Excel template computes and uses the Wald truncation number for n
and the rejection number so that sampling can be terminated, and a decision made. The
template also tells the user when acceptance and rejection are not possible. Suppose
that it was determined that the second and fifth items were found to be defective. What
decision would you make? In order to let the template know which items are defec-
tive, a 1 is entered into cells N6 and N9 with all other cells containing a 0 (good items).
Figure 7.4 shows that once a 1 is entered into cell N9 the decision in cell P9 indicates
Reject <= Stop here!
The sequential sampling chart is given in Figure 7.5. The cumulative number of defects
of 2 exceeds the reject line at sample 5. The OC, AOQ, and ASN curves are shown in
Figures 7.6 through 7.8.
 Sequential Sampling by Attributes

FIGURE 7.4
Excel template Sequential Sampling Plan.xlsx sampling plan entry form.
167
 Sequential sampling plan w/Wald truncation
168

18

Cumulative defects Accept limit

16 Reject limit Truncation reject limit
Truncation accept limit

14

12

Reject

10

Accept

Cumulative defects
6

Item sample number, n

0
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191

–2

FIGURE 7.5
Excel template Sequential Sampling Plan.xlsx sequential sampling chart.
Acceptance Sampling in Quality Control
Sequential Sampling by Attributes 169

Operating characteristic (OC) curve

1.000

0.900

0.800
Lot probability of acceptance, Pa

0.700

0.600

0.500

0.400

0.300

0.200

0.100

0.000
0.000 0.050 0.100 0.150 0.200 0.250
Lot fraction defective, p

FIGURE 7.6
OC curve for sequential sampling plan with α = .05, β = .10, p1 = .018, and p2 = .18.

Average outgoing quality (AOQ) curve

0.045

0.04

0.035
AOQ
Fraction defective in outgoing lots

AOQL = 0.0408
0.03

0.025

0.02

0.015

0.01

0.005

0
0.000 0.050 0.100 0.150 0.200 0.250
Fraction defective in incoming lots

FIGURE 7.7
AOQ curve with AOQL for sequential sampling plan with α = .05, β = .10, p1 = .018, and p2 = .18.
170 Acceptance Sampling in Quality Control

Average sample number (ASN) curve

18

16

14
ASN
12
Average sample size

10

0
0.000 0.050 0.100 0.150 0.200 0.250
Lot fraction defective

FIGURE 7.8
ASN curve for sequential sampling plan with α = .05, β = .10, p1 = .018, and p2 = .18.

Problems
1. Construct a sequential sampling plan such that p1 = .04 and p2 = 20 with α = .05 and β =.10.
2. Construct a sequential sampling plan such that p1 = .07 and p2 = .30 with α = .05 and β = .10.
3. Construct a sequential sampling plan such that p1 = .02 and p2 = .06 with α = .05 and β = .10.
4. Compare the results for Problems 1 through 3. What is the effect of increasing the slope?
What is the effect of decreasing the slope? What is the effect of increasing h2? What is
the effect of increasing h1?
5. Plot the Type B OC curve for the following sequential sampling plan using s and a
minimum of two other plotting points. Assume p1 = .05, p2 = .10, α = .05, and β = .10.
Y2 = 0.0723k + 3.8682
Y1 = 0.0723k − 3.0129
What is the ASN at these points?
6. The plan
Y2 = 0.0656k + 1.9481
Y1 = 0.0656k − 1.5174
has p1 = .03 and p2 = .12 at 1 − α = .95, β = .10. Compute the ASN and AOQ at these points.
Sequential Sampling by Attributes 171

7. MIL-STD-105E, Code L, 1.5 acceptable quality level (AQL) shows μ1 = 1 defect per
100 units at Pa = .95 and μ2 = 5.9 defects per 100 units at Pa = .10. Construct a sequential
chart in terms of defects per 100 units which matches this plan.
8. Determine the ASN at μ1 and μ2 for the plan developed in Problem 7.
9. What is the minimum number of samples leading to acceptance in Problem 7?
10. Using the formulas, devise a sequential sampling plan having the following charac-
teristics: p1 = .01, p2 = .06, α = .05, and β = .10.

References
Cowden, D. J., 1957, Statistical Methods in Quality Control, Prentice-Hall, Englewood Cliffs, NJ.
Statistical Research Group, Columbia University, 1945, Sequential analysis of statistical data:
­applications, AMP Report 30.2R, Columbia University, New York.
Wald, A., 1947, Sequential Analysis, John Wiley & Sons, New York.
Wetherill, G. B., 1986, Sequential Methods in Statistics, 3rd ed., Chapman & Hall, London, U.K.
8
Variables Sampling for Process Parameter

Specifications are frequently written in terms of statistical parameters that describe the
product to be inspected. For attributes inspection, the parameter to be controlled is, of
course, the proportion nonconforming in the lot or process. When specifications are writ-
ten in terms of measurements, other parameters may be of importance, such as the aver-
age (mean) level of a certain characteristic of the process that produced the units to be
inspected, or in some instances its standard deviation. This implies Type B sampling.
Examples of such specifications are mean life of a lamp, average amount of discharge of an
impurity into a stream, average emission of carbon monoxide from cars of a certain make
and model, and the standard deviation of an electrical test on semiconductors for use in a
ballistic missile. Specifications of this type are in contrast to those on the individual mea-
surements themselves that relate to individual units of the product; variables sampling
plans for such specifications will be covered in a later chapter.
It is characteristic of specifications on a process parameter that certain levels are accept-
able and should be protected from rejection, while other levels are objectionable and
should be rejected by the plan. This was recognized by Freund (1957) when he distin-
guished between two critical levels:

θ1: Acceptable process level (APL), a process level that is acceptable and should be
accepted most of the time by the plan
θ2: Rejectable process level (RPL), a process level that is rejectable and should be
rejected most of the time by the plan

Here, θ is taken to be the parameter specified.

In a manner analogous to attributes plans, the probability of acceptance for each of these
levels is defined as follows:

α is the probability of rejection at the APL (producer’s risk).

1 − α is the probability of acceptance at the APL.
β is the probability of acceptance at the RPL (consumer’s risk).

Most variables acceptance sampling plans for a process parameter can be specified in
terms of these levels and risks.

Single Sampling for Process Parameter

Standard statistical tests of hypotheses form the basis for the methodology of single sam-
pling by variables for process parameters. In fact, such plans are simply tests of hypotheses.
Thus, the statistical tests shown in Table 8.1 may be employed as sampling plans in this con-
text. They are used as one- or two-sided tests depending on whether the parameter is to be

173
174 Acceptance Sampling in Quality Control

TABLE 8.1
Statistical Tests of Hypotheses
Parameter Specified Condition Test Statistic

X - m0 X - m0
Mean (μ0) μ0 specified, σ known Normal z-test z= =
sX s/ n

X - m0 X - m0
μ0 specified, σ unknown Student’s t-test t= =
sX s/ n
2
æ s ö
Standard deviation (σ0) s02 specified χ2-test c 2 = ( n - 1) ç ÷
è s0 ø

controlled against specifications on one or both sides. The operation of such tests is described
in standard statistical texts such as Bowker and Lieberman (1959). Sample sizes are critical
in acceptance sampling applications and may be determined from the appropriate power
or operating characteristic (OC) curves of the test. The power curve shows the probability
of rejection plotted against hypothetical values of the process parameter. Its complement is
the OC curve. These curves for variables plans for the process parameter are usually plotted
against the standardized displacement of the parameter from the APL (μ1) such as

d = ( m - m 1 ) /s

or

l = s2 /s12

Sample size for two-point plans can be determined from the standardized displacement of
the RPL (μ2) from the APL (μ1). For the tests mentioned, this is as follows:

Test Displacement

z-test, t-test m 2 - m1 RPL - APL

d0 = =
s s
χ2 test s2
2 RPL
l0 = 2
=
s1 APL

Figure 8.1 shows typical OC curves. For a set of curves with the specified α risk, the sample
size is found from the curve passing through (or nearest to) the intersection of d0 or λ0, plot-
ted on the horizontal axis, and β, plotted on the vertical axis. If no curve passes through
this point, crude interpolation may be necessary. Frequently, a 5% producer’s risk and a
10% consumer’s risk are employed. OC curves are given in the appendix tables for α = .05.
They include the following:

Appendix Figures Tests

F8.1 One-sided normal z-test

F8.2 Two-sided normal z-test
F8.3 One-sided Student’s t-test
F8.4 Two-sided Student’s t-test
F8.5 One-sided χ2-test
Variables Sampling for Process Parameter 175

1–a

Pa

n5 n4 n3 n2 n1

λ0 λ

FIGURE 8.1
Typical OC curves.

For example, suppose the mean life of a lamp was specified by the manufacturer as 1000 h
(APL), while the customer wished to be sure to reject shipments of lamps having a mean
life of 800 h (RPL). The standard deviation of life is not known but is expected to be in the
order of 200 h. A one-sided t-test is appropriate since the implied specification on mean life
is one-sided, that is, μ not less than 1000 h. The standardized displacement is

RPL - APL 800 - 1000

d0 = = =1
s 200

If risks are set at α = .05 and β = .10, the OC curve for the one-sided t-test shows that
a sample size of n = 10 is required. A sample of 10 would be selected from the lot and
the t-test applied, accepting or rejecting the lot as the null hypothesis is accepted or
rejected.

Acceptance Control Charts

A natural extension of a standard test of hypothesis on individual lots is to plot the results
for successive lots in control chart form. This serves to allow the trend and runs analysis on
a continuing series of lots and affords the acceptance control engineer all the advantages
176 Acceptance Sampling in Quality Control

Upper rejectable process level, RPL

Upper acceptance control limit, ACL
d
Upper acceptable process level, APL
X Nominal center line, NCL
Lower acceptable process level, APL
d
Lower acceptance control limit, ACL
Lower rejectable process level, RPL
Lot number

FIGURE 8.2
Acceptance control chart.

of the control chart technique. The critical value for the test serves as the acceptance con-
trol limit (ACL). Lots that plot inside the ACL are accepted. Those that plot outside are
rejected.
This idea was first proposed by Freund (1957) in a celebrated paper that later won
the Brumbaugh Award from the American Society for Quality Control as the best
technical contribution of the year. A two-sided chart appears as in Figure 8.2. A one-
sided chart would consist of the upper half or lower half of the chart shown depend-
ing upon the direction in which the mean is to be controlled. Of course, the APL and
RPL do not appear on a chart in application. Only the ACL and the nominal center
line (NCL) are shown in actual use. The NCL is, of course, halfway between the ACL
in the two-sided case.
The initial work on the acceptance control chart has been primarily with standard
deviation known. Freund (1957, p. 14) points out that “it is implied that α and β risks
will be selected for the APL and RPL values, respectively, and that σ will be known
from past experience or estimated in the usual control chart manner from the R or s
computed from about 20 samples.” Clearly, the measure of variability must have been
in control for 20 or more samples to assure the stability implicit in a known standard
deviation application. The acceptance control chart is easily set up using appropriate
formulas.
If ACL is the acceptance control limit, zα isthe standard normal deviate cutting off an
area of α in upper tail, zβ is the standard cutting off an area of β in upper tail, and d is the
distance ACL lies from APL in the direction of RPL, then

æ ( za + zb ) s ö
2

n = çç ÷÷
è RPL - APL ø
za
d= RPL - APL
za + zb

where the sign of |RPL − APL| is regarded as always positive. The distance, d, and its rela-
tion to the ACL, APL, and RPL are shown in Figure 8.2.
Variables Sampling for Process Parameter 177

Freund has derived special factors that facilitate the determination of the ACL. They
allow computation of the limits either from the APL as before or alternatively from the RPL
as a baseline. Using γ to represent the risk, the factors are the following:

Factor (A) Measure of Variability (V)

A0, γ σ known

( )
2
1 S X-X
A1, γ s= S for k lots with n samples
k n
1
A2, γ R= SR for k lots with n samples
k

( )
2
1 S X-X
A3, γ s= S for k lots with n samples
k n -1

Appendix Table T8.6 gives the Freund A factors for various values of α and β. To use the
A factors, the sample size must be calculated from the formula as above. For any of the
­factors Aγ with the corresponding measure of variability, V, we have the following:

Lower Limit Upper Limit

ACL = APL − AαV ACL = APL + AαV

ACL = RPL + AβV ACL = APL − AβV

It can be seen directly that d = AαV.

In a two-sided situation, the ACL may be so close to the nominal value (NCL) as to force
consideration of both tails of the distribution of sample mean values simultaneously. To
assess the need for special correction terms (CTs) in this situation, compute

n ( ACL - NCL )
D1 = , which is the deviation of upper ACL from NCL in terms of sX or,
s
alternatively

n ( APL - NCL )
D2 = , which is the deviation of upper APL from NCL in terms of sX ,
s

depending on whether the ACL themselves or the APL has already been specified. Apply
the CTs found in Appendix Table T8.7 if the value Δ1 or Δ2 calculated is less than the
value shown in Table 8.2. If a CT is necessary, it can be found in Appendix Table T8.7
corresponding to the value of Δ1 or Δ2 calculated. The A factor of Appendix Table T8.6 is

TABLE 8.2
Minimum Δ Values before Correction Terms Need Be Used
α Risk Minimum Δ1 Minimum Δ2 α Risk Minimum Δ1 Minimum Δ2
0.05 2.5 0.851 0.005 3.2 0.619
0.01 3.0 0.670 0.001 3.5 0.409
178 Acceptance Sampling in Quality Control

then multiplied by the CT to obtain a new A factor to be used in this two-sided situation.
Application of the new A factor proceeds as before. Alternatively, the CT can be used in the
formulas for n and d as follows:
2
æ é( CT ) za + zb ù s ö
n=çë û ÷
ç RPL - APL ÷
è ø

d=
( CT ) za RPL - APL
( CT ) za + zb

Freund (1957) presents the theory behind acceptance control charts as well as many excel-
lent examples. The following is an adaptation of one such example.
Bottles are filled with 10 cm3 of a solution. The amount of the solution is to be maintained
within ±0.5 cm3 with less than 0.1% of the bottles outside the specification. It is desired to
reject if more than 2.5% of the bottles are under- or overfilled. A sample is to be taken of
each half hour’s production to be plotted against an acceptance control chart having α = .05
and β = .10. The standard deviation has been estimated from control charts as σ = .10 and
fill is normally distributed.
Using normal distribution theory to estimate the APL and RPL, we obtain the upper
specification

RPL = USL - z.025s = 10.5 - 1.96 ( .10 ) = 10.30

APL = USL - z.001s = 10.5 - 3.09 ( .10 ) = 10.19

and the lower specification

APL = LSL + z.001s = 9.5 + 3.09 ( .10 ) = 9.81

RPL = LSL + z.025s = 9.5 + 1.96 ( .10 ) = 9.70

The sample size for α = .05 and β = .10 is

æ ( za + zb ) s ö
2

n = çç ÷÷
è RPL - APL ø
2
æ ( 1.645 + 1.282 ) .10 ö
=ç ÷ 7
è 10.30 - 10.19 ø

For a nominal value of 10,

7 ( 10.19 - 10.0 )
D2 = = 5.03
.10

and so no CT for double specification limits is needed. Then

za 1.645
d= RPL - APL = 9.70 - 9.81 = 0.062
za + zb 1.645 + 1.282
Variables Sampling for Process Parameter 179

Note that, for sample size n = 7, the Freund A factor from Appendix Table T8.6 is 0.622 and
d = A0,.05σ = .622 (.10) = .062 as it should be.
The acceptance control chart can be set up accordingly and mean values of samples of
size 7 are plotted against the acceptance limits to determine the acceptance of subsequent
lots. A control chart for variability should also be instituted to detect any change in stan-
dard deviation from the known value.

Sequential Plans for Process Parameter (σ Known)

When sample size must be kept to an absolute minimum, sequential plans also provide an
excellent approach in sampling against specified process parameters. These plans are used
on variables data in a manner analogous to sequential plans for attributes. A greater vari-
ety of sequential charts are available for variables, however, since the attributes sequential
test is usually limited to a one-sided test against an increase in proportion nonconforming
over that specified as the producer’s quality level. Separate sequential tests for variables
are available for various parameters against an upper specification limit, a lower specifica-
tion limit, or double specification limits. As in attributes testing, a cumulative statistic, Y,
is plotted against the number of samples taken.
In testing against an upper specification limit, the decision lines for the sequential plot
are of the same form as for attributes, namely, the rejection line, Y2 = h2 + sk, and the accep-
tance line, Y1 = −h1 + sk. Formulas for the acceptance constants h1, h2, and s for testing the
mean and variance against an upper specification limit are given in Table 8.3 together with
those for the average sample number (ASN) at APL, s, and RPL. Formulas for attributes
testing are also given in Table 8.3 for reference. In using the table, all computations should
be made consistently in common or natural logarithms with the factor L adjusted accord-
ingly. The form of the sequential chart for an upper specification limit is that of Figure 8.3.
Sommers (1979) has pointed out a simple relation between sequential plans and single-
sampling plans for the mean. Given a known standard deviation single-sampling plan
using a sample of n to test μ1 against an upper limit μ2 (i.e., μ2 > μ1) with risks α = .05 and
β = .10, the matching sequential plans have the following parameters:

h1 = .7693s n
h2 = .9877 s n
s
s = m1 + 1.4632
n

or

s
s = m 2 - 1.4632
n

ASN ( m1 ) = .4657 n
ASN ( s ) = .7598n
ASN ( m 2 ) = .5549n
 180

TABLE 8.3
Formulas for Single Upper Limit Sequential Plans for Process Parameter (For Common Logs, L = 2.3026; for Natural Logs, L = 1)
Test Plot (Y) h1 h2 s APL s RPL

Mean n m 2 + m1 h2
Lbs2 Las2 Pa = 1 - a Pa = b
Y= i
Pa =
APL = μ1 åX m 2 - m1 m 2 - m1 2 h1 + h2
RPL = μ2 i =1 ASN =
(1 - a ) h1 - ah2 ASN =
(1 - b ) h2 - bh1
s - m1 hh m2 - s
= Sum of observations ASN = 1 2 2
s
Variance n h2
2 2Lbs12 s22 2Las12 s22 æ s2 ö Pa = 1 - a Pa = b
Y= Pa =
APL = s12 å (X - m) i L log ç 22 ÷ s12 s22 h1 + h2
s22 - s12 s22 - s12 è s1 ø (1 - a ) h1 - ah2 (1 - b ) h2 - bh1
i =1 ASN = ASN =
RPL = s22 s - s12 hh s22 - s
If m unknown , plot s22 - s12 ASN = 1 22
2s
n
2
Y¢ = å(X - X ) i
i =1
against k ¢ = k - 1
n
h2
Proportion b a g2 Pa = 1 - a Pa = Pa = b
defective Y= di h1 + h2
å g1 + g 2 g1 + g 2 g1 + g 2 (1 - a ) h1 - ah2 (1 - b ) h2 - bh1
APL = p1 i =1 ASN = hh ASN =
s - p1 ASN = 1 2 p2 - s
RPL = p2 = Total defective s (1 - s )

Defects per n b a m 2 - m1 Pa = 1 - a h2 Pa = b
Y= Pa =
unit åd i ( log m2 - log m1 ) ( log m2 - log m1 ) L ( log m 2 - log m1 ) h1 + h2
APL = μ1 i =1 ASN =
(1 - a ) h1 - ah2 ASN =
(1 - b ) h2 - bh1
s - m1 hh m2 - s
RPL = μ2 = Total defective in ASN = 1 2
s
k units

æ 1-b ö æ 1- a ö æ p2 ö æ 1 - p2 ö Upper Lower

a = log ç ÷ b = log ç ÷ g1 = log çç ÷÷ g 2 = log çç ÷÷ specification specification
è a ø è b ø è p1 ø è 1 - p2 ø Rejection line: Rejection line:
Y2 = h2 + sk Y2¢= −h2 + sk
Acceptance line: Acceptance
Y1 = −h1 + sk line:
Y1¢= h1 + sk
Acceptance Sampling in Quality Control
Variables Sampling for Process Parameter 181

Y Y2

Reject

Y1
Continue

Accept
0 k

FIGURE 8.3
Sequential variables chart: the upper specification limit.

This provides for an immediate assessment of the efficiency of sequential sampling as

against single sampling for the mean.
In testing against a lower specification limit, the acceptance and rejection regions are, of
course, reversed. This gives rise to new decision lines, namely, the rejection line, Y2¢ = - h2 + sk ,
and the acceptance line, Y1¢ = h1 + sk, where the values of the acceptance constants are the
same as those given in Table 8.3 for the upper limit. The formulas given for ASN also
remain the same as for the upper limit but must be taken in absolute value. The results for
the decision lines can be seen to come about from interchanging μ1 with μ2 and α with β in
the formulas for the upper limit, thus reversing the roles of h1 and h2. These changes can
also be made in Sommers’ formulas for use with a lower limit (μ2 < μ1) by interchanging μ1
with μ2 in s and h1 with h2. The formulas for ASN remain the same. The form of the chart
for a lower limit is shown in Figure 8.4.
In general, when testing the mean against either an upper or a lower specification limit
using the formulas of Table 8.3 for h1, h2, and s, it can be shown that the following relations
hold for probability of acceptance and ASN for any given value of μ except μ = s

m 2 + m1 - 2m
w=
m 2 - m1

so that

w
éë( 1 - b )/a ùû - 1
Pa = w w
éë( 1 - b )/a ùû - éëb/( 1 - a ) ùû

182 Acceptance Sampling in Quality Control

Y Y΄1

Accept

Y΄2
Continue

Reject
0 k

FIGURE 8.4
Sequential variables chart: the lower specification limit.

and

é Ls2 log é( 1 - b ) /a ù ù
ê ë ûú
êë m 2 - m1 úû
ASN =
é 2m - m 2 - m1 ù
êë 2 úû
é w
éë( 1 - b )/a ùû - 1 ù é Ls2 log éb/( 1 - a ) ù - Ls2 log é( 1 - b )/a ù ù
ê úê ë û ë ûú
ê é( 1 - b )/a ù w - éb/( 1 - a ) ù w ú ê m 2 - m1 úû
ë û ë û ûë
+ ë
é 2m - m 2 - m1 ù
êë 2 úû

where L = 1 when natural logarithms are used and L = 2.3026 using common logarithms.
When μ = s
log éë( 1 - b )/a ùû
Pa =
log éë( 1 - b )/a ùû + log éë( 1 - a )/b ùû
L2s2 log éë( 1 - b )/a ùû log éë( 1 - a )/b ùû
ASN =
( m 2 - m1 )
2

These relations are useful in constructing the OC and ASN curves.
For example, for μ1 = 1 and μ2 = 1.2, with α = .025 and β = .10, we have the following
results for a value of the mean μ = 1 with σ = .1 using natural logarithms:
1.2 + 1 - 2 ( 1)
w= =1
1.2 - 1
[.9/.025] - 1 = .975
1

Pa =
[.9/.025] - [.1/.975]
1 1

Variables Sampling for Process Parameter 183

é 1 ( .01) log [.90/.025] ù

ê ú
1.2 - 1
ASN = ë û
é 2 - 1.2 - 1 ù
êë 2 úû
é [ .90/.025] - 1 ù é 1 ( .01) log [.10/.975] - 1 ( .01) log [.90/.025] ù
1

ê 1úê ú
êë [.90/.025] - [.10/.975] úû ë
1
1.2 - 1 û
+
é 2 - 1.2 - 1 ù
êë 2 úû
.1792 + [.975][ -.2930 ]
= = 1.065
-.1

Occasionally, it may be necessary to plot a linear function of the quality characteristic being
measured. This is often done to adjust the slope or scale of the chart to practical propor-
tions. Suppose

Y = å x

has been plotted and it is desired to plot

Y * = å ( ax + b )

Then the equations for the upper limit decision lines become

Y2* = ah2 + ( as + b ) k
Y1* = - ah1 + ( as + b ) k

so that, in effect,

h2* = ah2
h1* = ah1

s* = as + b

For example, if the slope of the chart is too steep, it may be adjusted by subtracting a con-
stant, C, from the points plotted to obtain

Y* = x - C
whereupon the equations for the decision lines become

Y2* = h2 + ( s - C ) k

Y1* = - h1 + ( s - C ) k

For a test of the mean, the constant C is often chosen equal to the APL. This places the APL
at Y = 0. When C = s, the sequential chart will have horizontal limits. When C is larger than
s, the chart will slope downward. That is, s* = s − C will be negative. Such an effect will be
184 Acceptance Sampling in Quality Control

observed when μ1, the APL, is subtracted from the individual observations to be plotted on
a chart testing against a lower specification limit on the mean. This is because
s* = s - m1
m + m2
= 1 - m1
2
m - m1
= 2
2
which will be negative in testing against a lower specification limit and positive in testing
against an upper specification limit. Note that, in this case, the decision lines for the lower
limit case can be found from the upper limit lines with the same α, β, and |μ1 − μ2| using
the relations rejection line: Y2¢ = h2 − sk and acceptance line: Y1¢ = −Y1 = h1 − sk. The form of
such a chart will be seen in Figure 8.5 for a lower specification limit where Y = x − μ1. The
form for an upper limit chart using this adjustment is that of Figure 8.3 when Y = x − μ1.
A chart for testing double specification limits on the mean can be constructed as the
superimposition of individual upper and lower specification limit charts. Such a chart is
illustrated in Figure 8.6.
A zero baseline is obtained for plotting the double specification limits chart by cumulating

Y = å ( x - m 1 )
This provides a common abscissa for the constituent upper and lower specification limit
charts. The APL, μ1, is taken halfway between the upper and lower RPL so that
upper RPL + lower RPL
m1 =
2
and the α risk is apportioned half to each plan so that each is set up using α/2. The formu-
las must be corrected for subtraction of the constant, μ1.

0 k
Accept

Continue
Y ΄1

Reject

Y ΄2

FIGURE 8.5
Adjusted sequential variables chart: the lower specification limit.
Variables Sampling for Process Parameter 185

Y Y2

Reject high

Y1
Continue

Accept
0 k
Accept

Continue
Y΄1

Reject low

Y΄2

FIGURE 8.6
Sequential variables chart: double specification limits.

As an example of application of such a double-limit sequential plan, suppose parts are

received that are to have a plastic coating of thickness 1 mm ± .2 mm. If it is expensive and
possibly destructive to measure the thickness of the coating, a sequential test is in order.
Here, there is a double specification limit involving the construction of two sequential
pans, for the upper and lower limits, respectively, which will then be combined into a sin-
gle chart for application of the plan. Taking the APL to be halfway between the two RPLs,
Lower RPL = m 0 = .8 mm
APL = m1 = 1 mm
Upper RPL = m 2 = 1.2 mm
Suppose the standard deviation is known to be σ = .1 mm. The plot must be adjusted by
subtracting the APL from each observation so that C = 1.
If conventional values of α = .05 and β = .10 are to be used, the α risk in each plan is
α = .025. The derivation of the upper limit plan using common logarithms is
æ 1-b ö æ .9 ö
a = log ç ÷ = log ç ÷ = 1.5563
è a ø è .025 ø
æ 1- a ö æ .975 ö
b = log ç ÷ = log ç ÷ = 0.9890
è b ø è .10 ø
2.3026 ( 0.9890 )( .1)
2
Lbs2
h1 = = = 0.1139
m 2 - m1 1.2 - 1.0
2.3026 ( 1.5563 )( .1)
2
Las2
h2 = = = 0.1792
m 2 - m1 1.2 - 1.0
m + m1 1.2 + 1.0
s= 2 = = 1.1
2 2
186 Acceptance Sampling in Quality Control

which leads to the limit lines

Y2 = .1792 + ( 1.1 - 1) k = .1792 + 0.1k

Y1 = -.1139 + ( 1.1 - 1) k = -1139 + 0.1k

For this plan, the operating properties are at APL (μ = 1 mm)

Pa = 1 - .025 = .975

ASN =
(1 - .025 )(.1139 ) - .025 (.1792 ) = 1.07
1.1 - 1.0

at s (μ = 1.1 mm)
.1792
Pa = = .6114
.1139 + .1792

ASN =
(.1139 )(.1792 ) = 2.04
( .1 )
2

at RPL (μ = 1.2 mm)

Pa = .10

ASN =
(1 - .1)(.1792 ) - .1 (.1139 ) = 1.50
1.2 - 1.1

If successive samples were 1.1, 1.15, 1.0, and 0.95, the sequential chart would appear as
in Figure 8.7 for a test against the upper limit only. Exploiting the symmetry, the relations

Y2¢ = -Y2 Y1¢ = -Y1

Y2 = 0.1792 + 0.1k

Reject

Y1 = –0.1139 + 0.1k

+0.1792
Accept
∑(x – 1)

0
1 2 3 4 5 6 k
–0 .1139

FIGURE 8.7
Upper limit chart, α =.025, β =.10.
Variables Sampling for Process Parameter 187

0.1139
∑(x – 1)

0
1 2 3 4 5 6 k

Accept
–0.1792

Y΄1 = 0.1139 – 0.1k

Reject

Y΄2 = –0.1792 – 0.1k

FIGURE 8.8
Lower limit chart, α =.025, β =.10.

can be used to obtain the decision lines for testing the lower limit. Using the same data as
before, the lower limit chart appears as in Figure 8.8 with

Y1¢ = .1139 - 0.1k ( accept ) Y2¢ = -.1792 - 0.1k ( reject )

Clearly, for this chart also, the ASN at the APL is still 1.07 while the ASN at the RPL is, by
symmetry, 1.50. For a test against the double specification limits 1.0 ± .2 mm, the charts are
superimposed as in Figure 8.9

Sequential Plans for Process Parameter (σ Unknown)

The preceding methods are for the case when the standard deviation is known. When this
is not the case, the methods of Barnard (1946) may be employed. Appendix Table T8.8
gives boundary values defining decision lines for a plot against successive values of k of
the statistic

S ( X - m1 )
Y=
S ( X - m1 )
2

188 Acceptance Sampling in Quality Control

Reject
0.5

0.4

0.3

0.2

0.1 Accept
∑(x – 1)

0
1 2 3 4 5 6 k
–0.1 Accept

–0.2

–0.3

–0.4

–0.5
Reject

FIGURE 8.9
Double specification limit chart, α = .05, β = .10.

This is a sequential version of the one-sample t-test against an upper specification limit,
that is, where the RPL is greater than the APL. In this case, we have μ2 > μ1 and μ2 = μ1 + Dσ
or D = (μ2 − μ1)/σ where D forms one of the arguments in the table of Barnard’s values. The
table also gives values of

k1 as the smallest number of values to reach a decision at APL

k2 as the smallest number of values to reach a decision at RPL
k1 as the ASN at APL
k 2 as the ASN at RPL

Values in brackets indicate that no decision is allowed at the given value of k.

For a test of a lower specification limit, that is, when the RPL is less than the APL, ­proceed
as follows:

1. Reverse signs of tabulated boundary values of decision lines Y1 and Y2.

2. Reverse acceptance and rejection regions.

Charts for lower and upper specification limits may be combined to test a double specifi-
cation limit in a manner analogous to that used when the standard deviation was known.
Such a chart is illustrated in Figure 8.10.
Variables Sampling for Process Parameter 189

+4
Reject
+3 Y2

+2 Y1

Continue
+1
Accept
0
k
Continue Accept
–1

–2 Y΄1

–3 Y΄2
Reject
Reject
–4

FIGURE 8.10
Barnard sequential chart.

As an example of application of the Barnard procedure, consider the previous coating

data. Suppose the standard deviation was not known and a chart testing against the upper
limit was to be prepared. Assume α = β = .05 is to be used where

μ1 = 1.0 mm  μ2 = 1.2 mm

The standard deviation is roughly approximated as σ = .1 so that

1.2 - 1.0
D= =2
.1

The necessary calculations are as follows:

(1) (2) (3) (4) (5) (6) (7) Decision Limits

k X X − μ1 Σ(X − μ1) (X − μ1)2 Σ(X − μ1)2 å( X - m 1 )2 Y = (4)/(7) Y1 Y2

1 1.10 .10 .10 .01 .01 .10000 1

2 1.15 .15 .25 .0225 .0325 .18028 1.387 0.37 (1.56)
3 1.00 0 .25 0 .0325 .18028 1.387
4 0.95 –.05 .20 .0025 .0350 .18708 1.069 1.03 1.82

The plot appears as in Figure 8.11. This test would lead to continued sampling on the
fourth sample without a decision.
190 Acceptance Sampling in Quality Control

Y
Reject Y2
2
Y1

1 Accept

Not for decision

0
1 2 3 4 5 6 7 8 k

–1

–2

FIGURE 8.11
Example of a Barnard sequential t-test with α =.05, β =.05.

Cumulative Sum Charts

It is sometime desirable to plot the results of sampling inspection in the form of cumulative
sum (CUSUM) charsts. Originated by Page (1954), their construction has been described
by Barnard (1959) and amplified by Johnson and Leone (1962). The charts consist of a
sequential plot to which a V-mask is applied point by point to assess the significance of
the plot against a specified value of the APL. Such a typical mask is shown in Figure 8.12.
Rejection occurs if any of the previous points plotted lie outside the angle defined by the
notch of the V-mask when the last point plotted is positioned a horizontal distance d from
the vertex of the angle of the notch. The mask is determined by two dimensions: the dis-
tance d and the angle of the notch 2θ.
Johnson (1961) has shown that the chart may be regarded “... as (roughly) equivalent to
the application of the sequential probability ratio test in reverse.” Thus, the notch of the
V-mask corresponds to the trapezoidal shape formed by the ordinate and the rejection
lines of a sequential graph. This is illustrated in Figure 8.12, which shows the two-sided
sequential graph corresponding to the V-mask.
Using this notion of equivalence, the following relationships may be employed to con-
vert the sequential parameters given in Table 8.3 to the dimensions of the V-mask:

h2
tan q = s d =
s

Furthermore, the “average run length” of points to a rejection is of interest in applica-

tion of the CUSUM chart. This is analogous to the ASN of the sequential procedure.
Variables Sampling for Process Parameter 191

h΄2
d
h2
θ
Y2 x

x
h2
θ x
x x Y1

d
Y΄1
–h΄2

Y΄2

FIGURE 8.12
Typical CUSUM chart.

An  approximation to the ARL, using sequential parameters, can also be developed
from the results of Johnson and Leone (1962). We find that at the APL
h2
ARL =
APL - s

Since a CUSUM chart cannot “accept” as such, the action rule is “not to reject” during
continuation. Rejection occurs only when the V-mask is violated. The CR is taken to the
zero for this approximation. Furthermore, the approximation should not be used when the
indicated ARL is 5 or less.
For example, in testing the mean,
h2
ARL =
m2 - s

so that with some algebra and the formulas of Table 8.3

2 log éë( 1 - b )/a ùû s2

ARL =
( m 2 - m1 )
2

and if β is taken to be zero

-2 log [ a ] s2
ARL =
( m 2 - m1 )
2

So that when d = (μ2 − μ1)/σ = .5 and α = .05, then ARL = 23.97, while the value calculated
by Johnson and Leone (1962) is 24.0.
192 Acceptance Sampling in Quality Control

These results assume the unit length of the vertical and horizontal scales is plotted 1:1.
For a scale using k units of length for the ordinate for one unit length of the abscissa,
d remains unchanged; however, the angle of the mask becomes θ = s/k. Naturally, the
sequential formulas for a linear transformation of the points plotted Σ(ax + b) may be used
to determine the sequential parameters when a CUSUM chart is plotted using a trans-
formed sum for scaling or other purposes.
Another variation on the cumulative sum involves a sequential plot with horizontal lim-
its. This may be obtained simply by subtracting the slope s from each point plotted before
it is added to the cumulative sum. The result is a horizontal sequential chart with limits h2
and −h1 and an NCL of zero. An interesting procedure for the use of such a chart in a one-
sided test has been given by Kemp (1962). A modification of his approach using a horizon-
tal one-sided sequential chart derived from the formulas given in Table 8.3 is as follows:

1. Set an upper limit at h2 and a lower boundary at 0.

2. Do not cumulate on the chart until an observation has been found to exceed s.
3. When an observation exceeds s, calculate and plot the cumulative sum Σ(X − s) on
the chart until
a. The cumulative sum Σ(X − s) exceeds h2. In this case, reject the process level as
significantly greater than the APL.
b. The cumulative sum Σ(X − s) returns to zero. In this case, discontinue accumu-
lation and return to step 2.

Such a chart is shown in Figure 8.13. The action rules for this chart are the same as those
for the CUSUM chart. Modifications of the procedure include the corresponding sequen-
tial test against a lower limit on the process level and use of the full chart with both the
acceptance and rejection regions defined. Kemp (1962) has also suggested that, for a plot
of Σ(X − s) without limits, a significant upward change in the process level is simply indi-
cated when the distance between the lowest point on the plot and the last point plotted is
greater than h2.
CUSUM charts and their variations offer many possibilities for the quality control engi-
neer in acceptance quality control as well as process quality control. The reader is referred

h2
∑(x – s)

x
x x
x x x
0 x x
1 2 3 4 5 6 7 8 9 10
k

FIGURE 8.13
Modified Kemp procedure.
Variables Sampling for Process Parameter 193

to Burr (1976), Johnson and Leone (1977), Duncan (1974), Wetherill (1977), and the l­ iterature
cited for more details on this interesting method. An excellent treatment of the philosophy
of application of the CUSUM chart will be found in (Craig 1969).
A detailed exposition of sequential methods for variables and attributes will be found in
(Wald 1947) and (Wetherill 1975).

Problems
1. In purchasing high-pressure cylinders, determine the sample size needed to assure
that the process level does not differ by more than 1 psi from nominal when the stan-
dard deviation is .5 psi. Use α = .05 and β = .10.
2. What test would be employed to determine if the variability in lengths of leads is at
the specified level of 6 cm? Suppose the standard deviation of a sample of 15 is 7 cm,
should the lot be rejected? Use α = .05.
3. An acceptance control chart is to be used in lot acceptance of a series of shipments
of glass tubes. The tubes are to an average not less than 90 cm in length. The feed
mechanism in the customer’s process would jam if the tubes are less than 87 cm. The
standard deviation is known to be 1.5 cm. Construct the appropriate chart. Successive
lot mean values are 90, 89, 88, 90, 87, 91, 92, and 89. Which lots should be rejected?
Use α = .05 and β = .10.
4. If tubes averaging more than 92 cm in length were also to be rejected in Problem 3,
what sample size would be required for an acceptance control chart with double
­limits having α = .005?
5. What is the meaning of seven successive points on the side of the NCL of an accep-
tance control chart?
6. The specification on the maximum average weight of a certain construction
material is 400 lb; however, if the average weight exceeds 408 lb the design must
be changed. The standard deviation is 8 lb. Set up a sequential chart to check the
weight of incoming lots of the material. Take α = .025 and β = .10. Should the lot
be accepted if the results from a lot are 397, 400, 385, 388, 404, 410, 411, 395, 394,
and 400?
7. Suppose it is decided that it is important for the material in Problem 6 also not to be
less than 392 lb. on the average. Construct a two-sided sequential chart with α = .05
and β = .10. Plot the sequential results.
8. It is suspected that the standard deviation in Problem 6 no longer equals 8. If it has
increased to 10, the lot should be rejected. Using α = .05 and β = .10, construct a
sequential chart for the variance. If successive values of (Xi − μ)2 are calculated from
the data of Problem 6, what is the decision after the last lot shown?
9. Using α = .05 and β = .05, what conclusion can be drawn from the data of Problem 6
when the standard deviation is unknown? Draw the sequential plot.
10. Convert the parameters of Problem 7 to those of a CUSUM V-mask.
194 Acceptance Sampling in Quality Control

References
Barnard, G. A., 1946, Sequential tests in industrial statistics, Journal of the Royal Statistical Society
(Series B), 8: 1–21.
Barnard, G. A., 1959, Control charts and stochastic processes, Journal of the Royal Statistical Society
(Series B), 21(2): 239–271.
Bowker, A. H. and G. J. Lieberman, 1959, Engineering Statistics, Prentice-Hall, Englewood Cliffs, NJ.
Burr, I. W., 1976, Statistical Quality Control Methods, Marcel Dekker, New York.
Craig, C. C., 1969, The X and R-chart and its competitors, Journal of Quality Technology, 1(2): 102–104.
Duncan, A. J., 1974, Quality Control and Industrial Statistics, 4th ed., Richard D Irwin, Homewood, IL.
Freund, R. A., 1957, Acceptance control charts, Industrial Quality Control, 14(4): 13–23.
Johnson, N. L., 1961, A simple theoretical approach to cumulative sum control charts, Journal of the
American Statistical Association, 56: 835–840.
Johnson, N. L. and F. C. Leone, 1962, Cumulative sum control charts-mathematical principles applied
to their construction and use, Industrial Quality Control, Part 1, 18(12): 15–21; Part 2, 19(1): 29–36;
Part 3, 19(2): 22–28.
Johnson, N. L. and F. C. Leone, 1977, Statistics and Experimental Design in Engineering and the Physical
Sciences, Vol. 1, 2nd ed., John Wiley & Sons, New York.
Kemp, K. W., 1962, The use of cumulative sums for sampling inspection schemes, Applied Statistics,
11: 16–31.
Page, E. S., 1954, Continuous inspection schemes, Biometrika, 41: 100–115.
Sommers, D. J., 1979, Personal communication with the author.
Wald, A., 1947, Sequential Analysis, John Wiley & Sons, New York.
Wetherill, G. B., 1975, Sequential Methods in Statistics, Chapman & Hall, London, U.K.
Wetherill, G. B., 1977, Sampling Inspection and Quality Control, 2nd ed., Chapman & Hall, London, U.K.
9
Bulk Sampling

Most of the literature on acceptance sampling relates to the inspection of discrete units
of product. For each unit, an associated quality characteristic is determined. This may be
either an attribute determination of acceptability (go, no-go) or a measurement of some
kind taken on each unit in the sample. However, another type of product, which con-
sists of material in bulk form, may be distinguished. The bulk sampling problem has been
described by Bicking (1967) as follows:

Bulk materials are essentially continuous and do not consist of populations of discrete,
constant, identifiable, unique units or items that may be drawn into the sample. Rather,
the ultimate sampling units must be created, at the time of sampling, by means of some
sampling device. The size and form of the units depend upon the particular device
employed, how it is used, the nature, the condition, and the structure of the material,
and other factors.

Bulk sampling may address issues such as the inspection of 100 ton of coal, sampling a
truck filled with gasoline, or the assessment of natural gas contained in a particular stor-
age tank. Sampling units might then be a shovel full of coal (whose size depends on the
shovel), a sampling bottle full of gasoline (the amount depending on the capacity of the
sampling bottle), or a sampling probe delivering gas to a container (of some size and at
some pressure). The important point is that the sample is constructed, not gathered up.
The objectives according to Bicking (1978, p. 304) are as follows:

1. Characterization of the material in place (as in a natural deposit) as to location,

amount, content, or value.
2. Characterization of a material as to grade, need for further processing, or
destination.
3. Control during processing.
4. Acceptance on a lot-to-lot basis.
5. Determination of weight or content for purposes of taxation or payment.
6. Determination of properties that must be known so that the end use will be
appropriate.
7. Experimentation and analysis to determine future sampling procedures or uses of
the material

Here, emphasis will be placed on the fourth objective, that of lot acceptance. As such, bulk
sampling may generally be regarded as a form of variables sampling for a process parameter.
Various devices have been developed to take samples of bulk materials. These have been
aptly described by Bicking (1968, 1978). Their proper use, however, depends upon knowl-
edge of any stratification in the material to be sampled. Sampling approaches in the pres-
ence of stratification have also been discussed by Bicking (1967). Consider, for example,

195
196 Acceptance Sampling in Quality Control

TABLE 9.1
Developing a Standard Sampling Method
1. Make clear the purpose of sampling.
a. What is the population from which the sample will be taken?
b. What information is required about the population; the mean, the variance, and the precision desired in
the estimate?
c. On what criterion will acceptance of the lot be based?
d. What action is to be taken to dispose of a rejected lot?
2. Specify the population and investigate the history of a lot.
a. Is the process that produced the lot in a state of control?
b. Is the definition of the lot size in conformity with the desires of the producer and the consumer?
c. Are the methods of handling and storage properly considered in determining the lot size?
3. Study the measurement error.
a. Separate the measurement error from the sampling error.
b. Compare the relative sizes of these two sources of error.
4. Estimate the several variances due to the process (within-lots and between-lots).
5. Prepare the sampling instruction, guarding against the following defects:
a. Lack of clarity in purpose of sampling.
b. Lack of specific enough instructions for taking increments. (“Take a representative sample” and “take a
random sample” are not specific enough.)
c. Unsuitable containers for the samples.
d. Failure to provide methods for checking sampling error, reliability, or measurement precision and bias.
6. Control the sampling operation.
a. Train the samplers.
b. Control the operation of the plan through check samples.
c. Periodically review the sampling instructions to provide for any changes in the process.
7. Periodically review the sampling instructions to provide for any changes in the process.

Source: Reproduced from Bicking, C.A., Mater. Res. Stand., 7(2), 103, 1967. With permission.

a  shipment of milk contained in a cylindrical tank car. Vertical samples may represent
strata disproportionally because of differences in horizontal dimension from top to bot-
tom. The cylinder is wider in the middle. Furthermore, without mixing, it may be very
important to be sure that the layer of cream is appropriately represented.
In all bulk sampling, the population sampled must be appropriately defined. Duncan
(1962) has discussed this in detail. It should be pointed out that it is most advantageous
to sample bulk material when it is moving, such as on a conveyor belt and in free fall.
Steps in developing a standard sampling method as given by Ishikawa (1958) are shown
in Table 9.1.

Construction of the Sample

The essential continuity of bulk materials allows parts of a sample to be blended or mixed
together to form a composite. The composite is then tested once, rather than individual tests
on its constituent parts. This is a physical way to average the composited samples. Suppose
three samples of coal were taken from a coal car on a siding as it was being unloaded. If ash
content was to be determined, three separate analyses could be performed. Alternatively,
the three samples might be mixed and blended into one composite sample. An analysis
of the composite should yield the same result as the average of the three distinct samples.
Lots, however, would be any measure of variation.
Bulk Sampling 197

Stratified sampling Compositing

Random sampling Lot of samples

Lot Zone Zone Zone

I II III Lot

Random
samples 1 2 3 4 1 2 3 4 5 6 1 2 3

Duplicate Composite
tests A BC D E FG H A B C DE F G H I JK L

Duplicate tests A B

FIGURE 9.1 
Types of sampling. (Reprinted from Bicking, C.A., Mater. Res. Stand., 7(2), 99, 1967. With permission.)

Often when the population is known to consist of several different subdivisions that may
give different results with respect to the quality characteristic measured, all the subdivi-
sions, or strata, are deliberately included in the sample. This is called “stratified sampling.”
When the subdivisions are sampled, we have “cluster” or “multistage sampling.” In any
such procedure, which deviates from “simple random sampling,” care must be taken to
properly weigh the sample results so that the effect of a result is proportional to its prob-
ability of occurring. This usually involves “proportional allocation.” It will be assumed
here that all samples are proportionally allocated. A comparison of types of sampling is
shown in Figure 9.1.
In bulk sampling, lots (or populations) of bulk material are regarded as being composed
of mutually exclusive subdivisions or segments. Sometimes obvious segments occur, when
the material comes in boxes or bags. Sometimes, however, the segments must be ­artificially
created by superimposing imaginary grids over the material or by other means of real
or synthetic division. Segments may be further subdivided into increments for s­ ampling
within a segment. In sampling theory, segments are often called primary units, while
­increments are called secondary units.
Segments are treated in a manner similar to the units in discrete sampling. Their
­average is considered as an estimate of the average of the lot and their variability as
a measure of variation on which to construct the standard error of the estimate of the
lot mean. With bulk material, however, the possibility for additional sampling within a
segment exists. Furthermore, the total variation observed may be broken into compo-
nents of variance that estimate the amount of variation that may be attributed to vari-
ous stages in the sampling process. Prior estimation of these components allows for the
determination of optimum sample size and for limits of error in situations in which the
sampling strategy precludes replicate observations. For example, the segments may
be sampled giving a variance between segments. Increments may be taken within seg-
ments to give a variance between increments, or sampling variance. The material from
each increment or from a composite of increments may then be reduced to the desired
particle size by crushing or grinding and the amount of material cut down by quarter-
ing to obtain one or more test units of a size just sufficient for a laboratory test of the
quality characteristic. This gives rise to the so-called “reduction variance.” The tests
198 Acceptance Sampling in Quality Control

themselves contribute a variance due to testing. A model* for the total variance in the
lot as broken into components is

sT2 = s12 + s22 + s23 + s24 ,

where
sT2 is the total variance in the lot
s12 is the variance between segments
s22 is the variance between increments within segments
s23 is the testing variance
s24 is the reduction variance

The last term in the model will be regarded as also containing variation from all sources
not explicitly shown in the model.
Frequently, samples may require no reduction or the reduction variance σ2 may be assumed
to be as small as essentially zero and omitted from the model. The variation due to reduc-
tion will then appear as part of the testing variance. These components of variance are often
assumed to be constant across the lot from segment to segment. As with all assumptions in
sampling, however, it is appropriate to check that it is true before setting up a new plan. This
is usually done by a control chart. Duncan (1962) has distinguished between two distinct
populations that may be conceptualized and tested through these procedures: populations
created by the act of sampling from what is called Type A bulk material with indistinguish-
able segments and populations having preexisting elements from what is called Type B bulk
material with distinguishable segments. An example of the former (Type A material) is a pile
of coal. An example of the latter (Type B material) is a lot consisting of 500 bags of fertilizer.
With Type A material, segments and increments must be artificially defined within the totality
of all the product submitted. With Type B material, the natural segments would be divided
into increments for sampling within segments. Bicking (1967) points out that “Type B materi-
als represent a transition between piece part sampling and sampling Type A materials.”

Estimation
Bulk sampling is primarily used to estimate the lot mean with a given degree of precision.
The resulting estimate may be sufficient in itself, or it may be used to determine lot accep-
tance. The magnitude of the standard error of the mean, and hence the precision of the
estimate, can, of course, be controlled by the number of samples taken. If, in the multistage
sampling of a lot of size N, the number of segments sampled is n1 and the number of incre-
ments taken within a segment is n2, while n3 tests are made on each increment, the variance
of the lot sample mean computed from all observations will be composed as follows:

s12 æ n1 ö s22 s23

s2X = ç 1 - ÷ + + ,
n1 è N ø n1n2 n1n2n3

* The conventional bulk sampling model reverses the roles of s23 and s24 ; ­however, the model given provides
consistency of enumeration when there is no compositing, that is, when s24 = 0.
Bulk Sampling 199

where
s12 is the variance between segments
s22 is the variance between increments within segments
s23 is the variance between tests within increments

Now this equation is applicable when increments or even segments are composited.
Compositing, however, may lead to an inability to estimate some, or all, of the components
of variance from the sample. It is a useful device when estimates of these variabilities are
not needed, when the components of variance are known.
Much is revealed by the partition of the variance of the sample mean. For example, in
sampling homogeneous liquids, s22 = 0 since the increments are all equal. Furthermore, if
n1 = N, as in stratified sampling, the first term goes to zero since the population of seg-
ments is exhausted. Furthermore, for given magnitudes of s12, s22  and s23, values of n1, n2,
and n3 may be determined by trial and error to find a combination that will reduce s2X to a
desired magnitude. Mean values of samples of this size will give the desired precision on
X whether composited or averaged over individual tests as long as initial estimates of the
magnitudes of the components of variance hold.
For example, suppose bags of argol are to be sampled by a split tube thief or trier, and it
is known that

s12 = .21 s12 = .31

If testing cost is high so that one test is to be made on each increment and if reduction vari-
ance is assumed negligible, we have

s12 s2
s2X = + 2 ,
n1 n1n2

where s22 will now include testing error since the latter will not be independently estimated.
To determine a plan that will give a 95% confidence of estimating the mean to within 0.4,
using the appropriate z value of 1.96 from the normal distribution, we must have

zsX = 0.4
0.4
sX = @ 0.2
1.96
s2X @ .04

A few possible combinations of n1 and n2 to give the desired sX are shown in Table 9.2,
where

.21 .31
s2X = +
n1 n1n2

Of the values shown n1 = 10 and n2 = 2 come closest to, but not greater than, the desired
s2X = .04 with the smallest number of segments. Note that any combination of n1 and n2
­giving s2X £ .04 is acceptable.
200 Acceptance Sampling in Quality Control

TABLE 9.2
Possible Combinations of n1 and n2

n1 n2 s 2X
9 1 0.058
2 0.041
10 1 0.052
2 0.036
11 1 0.047
2 0.033

It can be shown (Davies 1960, p. 111) that for a two-stage plan (n1 segments, n2 incre-
ments) costing c1 to sample a segment and c2 to sample an increment from a segment, an
economically optimum plan can be developed. For a lot with N segments, where the cost
of testing is the same for segments or increments, the most economical sample sizes with
which to estimate the lot mean to within ±E with 1 − α confidence are found as

c1s22
n2 =
c2s12

with

n1 =
(
N s22 + n2s12
,
)
Nn2 ( E/za/2 ) + n2s12
2

where zα/2 is the standard normal deviate associated with the confidence level to be incor-
porated in the two-sided estimate. For essentially infinite lot sizes, the previous formula
for n1 becomes

s22 + n2s12
n1 =
n2 ( E/za/2 )
2

Using the previous example on argol, if c1 was known to be $31 and c2 was $21, the sample
sizes required would be

.31 æ 21 ö
n2 = ç ÷ =1
.21 è 31 ø

and

.31 + 1( .21)
n1 =
1( .4/1.96 )
2

= 12.5 ~ 13
Bulk Sampling 201

The plan to minimize cost in this case is n1 = 13, n2 = 1. The total cost of this plan would be

c = $31 ( 13 ) + $21 ( 13 )( 1) = $676

and result in a standard error of estimate

.21 .31
sX = + = .04 = .2
13 13

The plan n1 = 10, n2 = 2 would cost

c = $31 ( 10 ) + $21 ( 10 )( 2 ) = $730

to produce a standard error of estimate

sX = .036 = .19

Both plans would meet the desired precision in estimating the lot mean.
Sampling of bulk material can be used most effectively when the components of vari-
ance are known. Knowledge of these values can allow estimation of the standard error of
the mean of the lot even when extensive compositing results in one test result on the lot.
Estimation of these components is straightforward. Let
X be the lot mean from n1 segments
X1 be the segment mean from n2 increments
X 2 be the increment mean from n3 tests
X 3 be the test result
then the mean squares used in constructing the estimates are

( )
2
å X1 - X
MS1 =
n1 - 1

( )
2
å X 2 - X1
MS 2 =
n1 ( n2 - 1)

( )
2
å X3 - X2
MS 3 =
n1n2 ( n3 - 1)

with degrees of freedom ν1, ν2, and ν3. Estimates of the components of variance can be
determined, regarding N as infinite, as follows:

Estimate of the testing component s23 is s32 = MS3 with ν3 = n1n2 (n3 − 1).
Estimate of the increment within segment component s22 is s22 = MS2 − s32/n3 with ( )
(s )
2
2
2
n2 =
(1/n ( n
1 2 - 1) ) ( MS 2 /1) + ( 1/n1n2 ( n3 - 1) ) ( MS 3 /n3 )
2 2
202 Acceptance Sampling in Quality Control

TABLE 9.3
Analysis of Variance Table for Nested Sampling
Degrees of Components of Variance
Source Sum of Squares Freedom Mean Square Estimated by Mean Square

Between segments n2n3(n1 – 1)MS1 n1 − 1 n2n3MS1 n2 n3 s12 + n3 s22 + s23

Increments within segments n1n3(n2 – 1)MS2 n1(n2 – 1) n3MS2 n3 s22 + s23
Tests within increments n1n2(n3 – 1)MS3 n1n2(n3 – 1) MS3 s23

Estimate of the between-segment component s12 is s12 = MS1 − s22 /n2 − s32 /n2n3 with ( ) ( )
(s )
2
2
1
n1 =
+ ( 1/n ) ( s /n ) ( )
2 2
(1/n1 - 1)( MS1/1) + ( 1/n 3 ) s32 /n1n2
2 2
2 2 2

These estimates of degrees of freedom are obtained using the Satterthwaite (1946) approxi-
mation and will usually be found to be conservative. They should be rounded down to
obtain integral values of degrees of freedom. Clearly, when the components of variance are
known, they may be regarded as having infinite degrees of freedom.
The mean squares MS1, MS2, and MS3 can be used to construct a nested analysis of vari-
ance table to display the variances involved as shown in Table 9.3. The multipliers shown
with the mean squares and the components of variance are necessary because analysis
of variance is usually performed on observation total rather than mean values as shown
here. For a discussion of analysis of variance performed using mean values, using ancillary
techniques, see Schilling (1973).
The standard error of the sample mean can be estimated as

s12 æ n1 ö s22 s23

sX = ç 1 - ÷ + +
n1 è N ø n1n2 n1n2n3

when the components of variance are known. When they are estimated, the formula for
estimation becomes

s12 æ n1 ö s22 s32

sX = ç1- ÷ + +
n1 è N ø n1n2 n1n2n3

with approximate degrees of freedom roughly

(s )
2
2
X
nX =
(1/n1 ) (1 - ( n1/N ) ) ( s12/n1 ) + (1/n 2 ) ( s22/n1n2 ) + (1/n 3 ) ( s32/n1n2n3 )
2 2 2 2

again obtained from the Satterthwaite (1946) approximation. This repeated use of the
approximation leads to a crude but often useful estimate of the degrees of freedom.
As pointed out by Duncan (1974b), the standard error of the mean can also be obtained
directly from the standard deviation of the segment results when they are available. The
estimate applies even if some of the segments have been composited to give the results, or
Bulk Sampling 203

in the face of other compositing or reduction within segments. The price of compositing
the segments is, of course, a reduction in degrees of freedom. This estimate is usually the
only one available when dealing with unique lots. The estimate is

( )
2
1 å X1 - X MS1
sX = =
n1 n1 - 1 n1

with degrees of freedom ν1 = n1 – 1.

It is sometimes necessary to estimate the variability within the lot. This is usually done
by taking a sample constructed using n1 segments, one increment from each segment, one
test per increment. It is recommended that at least 10 segments be selected as a sample. No
compositing is allowed. The estimate is

( )
2
å X1 - X
sX =
n1 - 1

with degrees of freedom ν = n1 − 1. The value of sX is then used to characterize the lot with
respect to variation in the lot and is also useful in determining the sample size. Note that
the components of variance associated with this measure are

s2X = s12 + s22 + s23

As an example of the application of these estimation techniques, consider the data of

Table  9.4 from a two-stage bulk sampling plan presented by Tanner and Lerner (1951),
which shows a sample taken from a shipment of argol.

TABLE 9.4
Percent Potassium Bitartrate in Shipment of Argol
Trierful
Bag 1 2 Mean Standard Deviation
1 86.37 86.46 86.42 0.0636
2 87.50 86.36 86.93 0.8061
3 85.75 86.05 85.90 0.2121
4 87.09 87.38 87.24 0.2051
5 87.31 86.78 87.04 0.3748
6 85.85 85.75 85.80 0.0707
7 86.46 85.44 85.95 0.7212
8 84.62 86.16 85.39 1.0889
9 86.41 86.26 86.34 0.1061
10 85.44 86.46 85.95 0.7212
Mean 86.28 86.31 86.296
Standard deviation 0.8938 0.5350 0.6080
Source: Reproduced from Tanner, L. and Lerner, M., Economic accumula-
tion of variance data in connection with bulk sampling, ASTM STP
1145, American Society for Testing and Materials, Philadelphia,
PA, 1951, 9. With permission.
204 Acceptance Sampling in Quality Control

Here N is large relative to n1 so that the finite population correction is not necessary.
Moreover, there is only one test per sample, so that s22  includes the variability of testing
and the model for the variance of the lot sample mean is

s12 s22
s2X = +
10 10 ( 2 )

where s23 and s24 are not shown in the model since they cannot be estimated from this
sample design.
Now

( 86.42 - 86.296 ) + ( 86.93 - 86.296 ) +  + ( 85.95 - 86.296 )

2 2 2

MS1 =
10 - 1
= .36967

and

( 86.37 - 83.42 ) + ( 86.46 - 86.42 ) +  + ( 86.46 - 86.95 )

2 2 2

MS 2 =
10 ( 2 - 1)
= .3124

so that

s2 = .3124 = .5589

with

n 2 = 10 ( 2 - 1) = 10

and

3124
s1 = .36967 -
2
= .21347 = .4620

(.21347 )
2

n1 = 2 2
1 æ .36967 ö 1 æ .3124 ö
ç ÷ + ç ÷
9 è 1 ø 10 è 2 ø
= 2.58 ~ 2

The standard error of the mean may then be estimated as

s12 s2 .2135 .3124

sX = + 2 = +
10 10 ( 2 ) 10 20

= .03697 = .1923
Bulk Sampling 205

with degrees of freedom

(.03697 )
2

nX = 2 2 = 5.4
1 æ .21347 ö 1 æ .3124 ö
ç ÷ + ç ÷
2 è 10 ø 10 è 20 ø

This is a conservative approximation since

s12 s22
sX = +
10 20
1 æ MS 2 ö MS 2
= ç MS1 - ÷+
10 è 2 ø 20
MS1
=
10

and MS1 has exactly 9 degrees of freedom as can be seen from the formula for its calcula-
tion. This estimate can also be obtained directly from the bag (segment) mean as

( 86.42 - 86.296 ) + ( 86.93 - 86.296 ) +  + ( 85.95 - 86.296 )

2 2 2
1
sX =
10 10 - 1
1
= MS1 = .1923
10

with, of course, 9 degrees of freedom. An estimate of this sort would be the only available
method for determining the standard error of the mean from a unique lot and is obviously
useful regardless of compositing within the segments.
A 95% confidence interval for the mean would be

X ± tsX
86.3 ± 2.26 ( .1923 )
86.296 ± .43

An estimate of the variability in the lot can be obtained using the results of, say, the trier-
ful 1. This gives

( 86.37 - 86.28 ) + ( 87.50 - 86.28 ) +  + ( 85.44 - 86.28 )

2 2 2

sx =
10 - 1
= .8938

with

ν = 9
206 Acceptance Sampling in Quality Control

Any estimate of this sort contains bag variation, trier variation within bags, any trier reduc-
tion variation, and the testing error. In an experiment such as this, a number of alternatives
would be available for compositing. Some of them are as follows:

1. No compositing. In the case of a unique lot or for a pilot study to determine the
components of variance to be used in continuing series of lots, this option provides
the most information about the variability involved. This estimate of the mean has
standard error

s12 s22
sX = +
10 20

The standard error can be estimated from the sample using the method given in
this example.
2. Composite the trier samples. Here, 10 analyses would be required each having a
variance

s22
s2 = s12 +
2

but the resulting standard error of the mean would be

1 æ 2 s22 ö s12 s22

s2X = ç s1 + ÷= +
10 è 2 ø 10 20

as before. This standard error could be checked using

10 S ( x1 - x )
2

sX =
10 10 - 1

This estimate is useful with unique lots. In that case, this estimate of the standard
error of the mean would provide 9 degrees of freedom.
3. Composite the odd and even segments (bags), respectively, into two samples. This
would result in two values that would be averaged to produce the estimated lot
mean. Each of these values would have a variance

s12 s22
s2 = +
5 10

but the resulting average of the two results would still have

1 æ s12 s22 ö s12 s22

sX = ç + ÷= +
2 è 5 10 ø 10 20

Bulk Sampling 207

However, now the standard error could be checked from the segment mean values
with 1 degree of freedom as

1 S ( xi - x )
2
R
sX = =
2 2-1 2

where R is the range of the two readings. This is quite useful on a continuing series
of lots since it provides a check that the variability has not changed from that pre-
dicted from the components of variance.
4. Composite the entire sample. With one analysis, this would show just one value—
the estimated mean of the lot. No estimate of standard error would be available
from the sample but known components of variance could be used to estimate the
standard error if available and if there was confidence that they had not changed
since they were obtained. The standard error of the mean would be determined as

s12 s22
sX = +
10 20

This procedure can sometimes be used on a continuing series of lots.

Thus, various strategies are available for compositing depending upon the structure of the
sample, cost and feasibility constraints, the desired precision of the estimate, the available
information, and the ingenuity of the individual designing the procedures (for further dis-
cussion, see Davies 1954).

Sampling Plans
The sampling plans that have been suggested for use with bulk materials are essentially
variable plans for a process parameter. Indeed, if segments are of equal size, the results
for the segments can be used with the plans given in Table 8.1 as if the segments were
individual units of product. Bulk sampling is, however, somewhat more complicated and
is distinguished by exploiting the essential continuity of the basic material in the lot in the
development of more complex and informative sampling plans. The method of sampling
and compositing must be considered in assessing the overall results. For this reason, tables
of bulk sampling plans are not available since the plans must be tailored to the individual
sampling situation and the analytical methods used.
Bicking (1970) has enumerated the following steps in setting up a sampling plan:

1. State the problem for which an estimate is desired.

2. Collect information on the relevant properties of the material (average properties,
components of variance in the properties).
3. Consider various approaches, taking into account cost, precision, and difficulties.
208 Acceptance Sampling in Quality Control

4. Evaluate these plans in terms of cost of sampling and testing, delay, supervisory
time, and convenience.
5. Select a plan.
6. Reconsider the preceding steps.

Consider a test of the mean of a lot against some specified value. The statistic t = (U − X)/sX
would be used for an upper limit where U plays the role of μ0 in Table 8.1. Similarly, L acts
as μ0 when a lower limit is involved. Once the mean of the lot is estimated by X and its stan-
dard error sX determined, the resulting value of t is compared to the relevant upper tail criti-
cal value from the t-distribution to determine the disposition of the lot (lower tail for lower
specification limits). Note that the problem of setting special test specification limits, which
takes into account measurement error, has been addressed by Grubbs and Coon (1954).
In the earlier example, suppose a lower specification limit on the average percent potas-
sium bitartrate in the lot was L = 87% with a producer’s risk of α = .05. Using the sample
results

X-L
t=
sX
86.3 - 87
t= = -3.68
.19

Comparison to the critical value of t = –1.83 with 9 degrees of freedom shows −3.68 is
less than −1.83 and the lot should be rejected. This test was made on the lower tail of the
t-distribution since a lower specification limit was involved. A diagram ­showing the
application of the test is the test given in Figure 9.2, where X = L at t = 0. In practice,
the consumer’s risk involved in such an assessment would be incorporated in deter-
mining the sample size. Note that the upper tail of the t-distribution could have been
used if the statistic were calculated as t = (X - L)/sX .
In discrete sampling, measurements are taken directly on well-defined units of product;
however, in bulk sampling, the continuous nature of the bulk within a segment allows for
considerable flexibility for sampling within a segment in an attempt to characterize it with
respect to the quality characteristic. A wide range of sampling techniques have been, and
may be, employed. For example, stratified sampling (see Bennett and Franklin 1954, p. 482),
multistage sampling (see Deming 1950, p. 160), ratio estimation (see Deming 1950, p. 183),
systematic sampling (see Bicking 1967), and interpenetrating subsamples from a stream
of product (see Duncan 1974b, p. 25A-9) are a few among others. It should be pointed out
that, in the literature of sampling, segments are referred to as primary units, increments

–3.68 –1.83 0 t

FIGURE 9.2 
The t-test for purity of liquid.
Bulk Sampling 209

as secondary units, and tests often as tertiary units. Procedures are also available for the
assessment of bulk quality characteristics in terms of proportions as well as measurements.
It is important to caution that chemical measurements are frequently expressed in units of
proportion or percent but should be analyzed as measurement data, for example, percent
carbon monoxide in the exhaust of a car. The statistical analysis of proportions refers to an
actual count of a discrete characteristic within a sample of a given size, for example, the
number of black grains in a sample of 100 grains of sand.
The various sampling techniques available in the literature of sampling theory (see, e.g.,
Deming 1950; Cochran 1953; Williams 1978) lead to an estimate of the lot or population
mean (or proportion) with its standard error, together with its associated degrees of free-
dom. These are readily available together with formulas for confidence interval estimation.
For use in acceptance sampling, the estimated parameter and its standard error of estimate
may be substituted in the criteria of Table 8.1 and used to test conformance of specifica-
tions. Furthermore, the procedures of sequential sampling and acceptance control charts
given in Chapter 8 can also be used in straightforward fashion once the estimate of the lot
mean and its standard error and degrees of freedom are determined. Sequential plans can
be used on the segment mean values to arrive at an early decision on the lot.
An adaptation and modification of the basic procedure of ASTM Standard E-300
(American Society for Testing and Materials 2004) will be given to illustrate the nature
and application of specific bulk sampling plans. While intended for sampling of indus-
trial chemicals, the procedure is easily generalized to other bulk sampling situations. For
a complete discussion of the method of ASTM E-300, refer the standard; also see Bicking
(1970). Note that this procedure involves sampling for a process parameter with a risk of
rejection when the process is at the specification limit, which acts as μ1.

Simple Random Sampling of a Unique Lot

(Components of Variance Unknown)
Unique lots present a problem in bulk sampling because some or all of the components of
variance associated with the inspection of the lot will, in general, be unknown. Furthermore,
the variability of the lot will also be unknown, requiring an initial preliminary estimate
before sample size can be determined.
Assume a unique lot is to be sampled for lot acceptance against a lower specification
limit L. The producer’s risk is to be α = .05, while the consumer’s risk is to be β = .10. Values
of θ1 (acceptable quality level) and θ2 (rejectable quality level) are given. The procedure is
as follows:

1. Take a preliminary sample of n1* segments (n1* ≥ 10) at random from the lot. Use
one increment per segment with one test per increment. In other words, use one
test unit per segment sampled.
2. Compute

å
n1*
Xi
X* = i =1

n1*
210 Acceptance Sampling in Quality Control

and

å ( )
2
n1*
Xi - X *
i =1
s* =
n1* - 1

3. Calculate
L - q2
d=
s*

and determine sample size n1 required from the operating characteristic (OC)
curve for the t-test.
4. Randomly select additional n1 − n1* units from the lot, and then pool them with
those of the previous sample. Compute

å
n1
Xi
X= i =1

n1

and

å (X - X )
n1 2
i
s= i =1

n1 - 1

so that

s
sX =
n1

5. Check the adequacy of the sample size selected by recomputing

L - q2
d=
s

and rereading the OC curve to obtain a new estimate of sample size. If this esti-
mate exceeds n1 by more than 20%, obtain additional units as necessary to reach
the indicated sample size. Use the increased sample size as n1 and return to step 4,
otherwise proceed to step 6.
6. Using the final estimates of X and sX , calculate

L-X
t=
sX

and compare this statistic to the upper .05 critical value of the t-distribution with
n1 − 1 degrees of freedom. If the calculated value exceeds the critical value, reject
the lot; otherwise accept.
Bulk Sampling 211

7. In dealing with an upper specification limit, proceed following the previous steps
using the formulas

q2 - U
d=
s
X -U
t=
sX

8. For double specification limits, check to be sure that

U - L > 6sX

If not, take additional samples to raise the sample size to

2
æ 6s ö
n1 ³ ç ÷
èU -Lø

Then test the upper and lower limits separately following the previous step, reject-
ing if either test rejects the lot.

As an example of the application of this procedure, let us return to the evaluation of the
percent potassium bitartrate given earlier. Suppose in that case θ1 = L = 87% and θ2 = 86%.
Using the sample results from trierful 1 for an initial sample of 10 segments

X * = 86.28 s* = .8938

with 9 degrees of freedom. Then

87 - 86
d= = 1.12
.8938

The OC curve shows that a sample size of 10 is required; hence, no further samples are
required, and

.8938
sX = = .2826
10

so

87 - 86.28
t= = 2.55
.2826

tested against a critical value of 1.83 with 9 degrees of freedom. Since 2.55 > 1.83, the lot
would be rejected.
212 Acceptance Sampling in Quality Control

Sampling from Stream of Lots

As in discrete sampling, inspection frequently takes place on a steady stream of product
produced by the same supplier. Assuming the process to be in control, a pilot study on the
initial product can be used to estimate the components of variance. From these estimates,
appropriate bulk sampling plans can be developed. Of course, these estimates must be
checked during the application of the procedure to be sure that they continue to hold.
A procedure for a pilot study to estimate the relevant components of variance and to
assess their stability is suggested in ASTM E-300. A modification and adaptation of the
procedure is given here to outline the detailed analysis necessary before a bulk sampling
procedure can be set up on a stream of lots. The procedure estimates that
s12 is the variance between segments (batch variability)
s22 is the variance between increments within segments (sampling variability within batches)
s23 is the testing variance (variability between tests)
s24 is the reduction variance (variability introduced by reduction of a gross sample to a
test unit size)

Estimation of Testing and Reduction Variances

1. Take 20 increments from each of five segments. Make 20 composites from the sets
of five of the first, second, third, …, twentieth increments across segments. Make
two tests on each composite.
2. Prepare two control charts using standard procedures available in any textbook
on statistical quality control, such as Burr (1976), Duncan (1974a), or Grant and
Leavenworth (1972), as follows:
a. Chart I: Range chart on the differences of the two tests on each increment to
test the stability of the testing variance
b. Chart II: Moving range chart on the means of the 20 composites to test the sta-
bility of the reduction variance
3. If both of these charts exhibit a state of control, estimate the testing and reduction
variances. Let X 4 be the composite mean, X3 be the test result, and X be the mean
of all measurements, then

( )
2
å å X3 - X4
MS 3 =
20 ( 2 - 1)

( )
2
å X4 - X
MS 4 =
20 - 1

so that
s 23 = MS 3
MS 3
s42 = MS 4 -
2
These estimates are used in the estimation of the variances between segments and
between increments.
Bulk Sampling 213

Estimation of Segment and Increment Variances

1. Take two increments from each of 25 successive segments produced by the process.
2. Make a single test on each of the 50 increments under uniform conditions (same
time, equipment, operator, etc.).
3. Prepare three control charts using standard procedures as follows:
a. Chart III: Range chart on the difference of results from two increments from
each segment to check the stability of within-segment variance
b. Chart IV: X chart for segment mean values to check the stability of a process
from segment to segment with respect to trend, runs, etc.
c. Chart V: Moving range chart for segment mean values to check the stability of
variance of segments
4. If the charts all show evidence of control, without exception, the components of
variance may be determined. As before, let X be the grand mean, X1 be the seg-
ment mean, and X 2 be the increment mean, then

( )
2
å X1 - X
MS1 =
25 - 1
( )
2
å X 2 - X1
MS 2 =
25 ( 2 - 1)

so that

s22 = MS 2 - s32
MS 2
s12 = MS1 -
2

Note that s42 is not subtracted from s22 since there is no reduction in the sense of
compositing in this part of the procedure.

The stability and magnitude of the components of variance having now been determined, it
is possible to apply the acceptance procedure to the stream of lots. The procedure suggested
by ASTM E-300 is based on the results of two composite samples obtained from each lot. The
lot is taken to be composed of N = n1 segments, all of which are of equal size and sampled to
produce a stratified sample of the lot. A sample of n2 increments is taken from each segment
where n2 is chosen to be an even number. If Type A bulk material is to be sampled, n1n2 ran-
dom increments are taken directly from the lot. The odd and even increments from within
segments are separately composited to form two composites A and B. Two tests are made on
each composite. The components of the variance model for the variance of the mean from
this procedure is
s12 s2 s2 s2
s2X = + 2 + 3+ 4
n1 n1n2 4 2

In testing against a lower specification limit, L, on the lot mean, the following procedure
is employed given the values of the acceptable process level θ1 = L, the rejectable process
level θ2, producer’s risk α = .05, and consumer’s risk β = .10.
214 Acceptance Sampling in Quality Control

Application of Plan to Stream of Lots

1. Given that n1 = N, n3 = 2, and n4 = 2, determine n2 as

s22
n2 = ,

n1 ( (( L - q2 ) /8.567 ) - ( s12/n1 ) - ( s32/4 )( s42/2)
2
)
where 8.567 = (zα + zβ)2. Round up to an even integer. For a test of an upper speci-
fication limit, substitute θ2 − U for L − θ2 in the previous formula.
2. Perform a check on the validity of the components of variance using two control charts:
a. Chart VI: Range chart of differences between the two tests made on each com-
posite. Use
UCL: 3.686s3
CL: 1.128s3
LCL: 0
which employ standard control chart factors for the range. This is a continu-
ation of chart I to check if the testing variance is stable at the estimated level.
Proceed if both points plot within the limits and the chart exhibits a state of
control, otherwise revert to the methods for a unique lot.
b. Chart VII: Range chart of the difference between the mean values of the A and
B composites. The chart checks the stability of the other components of vari-
ance. Its limits are

s12 2s22 s32

UCL : 3.686 + + + s42
n1 n1n2 2
s12 2s22 s32
CL : 1.128 + + + s42
n1 n1n2 2
LCL : 0
Proceed only if the point plots within the limits and the chart exhibits a state of
control, otherwise revert to the methods for a unique lot.
3. Estimate the standard error of the mean by

s12 s2 s2 s2
sX = + 2 + 3+ 4
n1 n1n2 4 2

4. Accept for single specification limits if (a) for lower specification limit, X ≥ L − 1.645 sX ,
or (b) for upper specification limit, X ≥ U + 1.645 sX . For double specification limits,
the acceptance procedure is as follows. If

s12 s2 s2 s2
(U - L ) £ 6 + 2 + 3+ 4
n1 n1n2 4 2

discontinue inspection since specification limits are too close to be assessed at this
sample size; otherwise proceed to test both upper and lower specification limits
separately. Reject if either test fails. Accept if both pass.
Bulk Sampling 215

To illustrate the application of this procedure, consider the concentration of an ingredient

in shipments of a certain chemical compound. The average level of the ingredient is not to
exceed 10%. Shipments consist of six bags each containing 50 lb of the material. The cus-
tomer does not wish to accept any material if the process average from which the sample
was taken exceeds 12%. It is known that
s12 = 1.0 = variance of bags
s22 = 1.9 = variance of samples within bags
s23 = 0.8 = variance of testing
s24 = 0 = reduction variance (assumed zero)
The plan is applied as follows:

1. Given that n1 = 6 and n3 = 2 with the reduction variance assumed negligible, then,
the number of increments needed from each bag will be

1.9
n2 =
æ ( 12 - 10 )2 1.0 0.8 ö
6ç - - ÷
ç 8.567 6 4 ÷
è ø
= 3.16 ~ 4

From a shipment of six bags, four increments are taken from each. The first and
third increments from each of the bags are composited into composite A while the
second and fourth increments from each of the bags are composited into compos-
ite B. Two tests are made on composite A and two tests on composite B. The results
are as follows:

Composite A Composite B

Test 1 8.3 8.8

Test 2 8.2 8.7

with an overall mean X = 8.5.

2. The differences

RA = 8.3 - 8.2 = .1

RB = 8.8 - 8.7 = .1

are plotted on a range chart to check the stability of the testing variance. Similarly,
the mean values of the two composites

X A = 8.25, XB = 8.75

are used to obtain the range of composite means

RX = 8.75 - 8.25 = 0.5

216 Acceptance Sampling in Quality Control

which is plotted on a range chart to check the stability of the other components of
variance using an upper control limit of

1.0 2 ( 1.9 ) 0.8

3.686 + + + 0 = 3.14
6 6 ( 4) 2

a center line of

1.0 2 ( 1.9 ) 0.8

1.128 + + + 0 = 0.96
6 6 ( 4) 2

and a lower control limit of 0. Both tests are in control.

3. The standard error of the mean is

1.0 1.9 0.8

sX = + + + 0 = .67
6 6 ( 4) 4

4. Since

X < 10 + 1.645 (.67 )

8.5 < 11.1

the lot is accepted.

ISO 10725
According to the Scope of the International Standard ISO 10725, “Acceptance sampling
plans and procedures for the inspection of bulk materials,” this standard specifies accep-
tance sampling plans by the determination of variables and use of acceptance inspection
procedures for bulk materials. These sampling plans comply with specific OC curves at
reasonable cost.
This International Standard is applicable to the inspection where the lot mean of a single
quality characteristic is the principal factor in the determination of lot acceptability, but it
also gives special procedures for multiple quality characteristics. The standard is appli-
cable to the cases where the values of standard deviations at individual stages of sampling
are known or are imprecise.
ISO 10725 is applicable to various kinds of bulk materials but is not always applicable to
minerals such as iron ores, coals, and crude petroleum, where the accurate estimation of
the lot mean is more important than the determination of lot acceptability.
For special cases when standard procedures are not always adequate and the measure-
ment standard deviation is dominant, this standard specifies special acceptance sampling
plans and procedures, such as in the case for liquids.
Bulk Sampling 217

Sampling increments (nI = 6)

Composite samples No. 1 No. 2

Test samples
(nT = 2)
Test portions
(measurements) (nM = 2)

Key
Odd numbered sampling increments
Even numbered sampling increments
Used
Not used

FIGURE 9.3 
Schematic model of bulk acceptance sampling procedures.

Section 5 of the standard discusses procedures for the inspection of an individual lot.
These procedures include increment sampling, constitution of composite samples, prepa-
ration of test samples, and measurements. Figure 9.3 illustrates a schematic model of bulk
acceptance sampling procedures as described in this standard.
Section 5 also covers descriptions of standard deviation estimates for between sam-
pling increments, between test samples, and the measurement standard deviation. The
costs of components, taking a sampling increment, preparing a test sample, and the cost
of a measurement are also discussed in this section. In addition, Section 5 compares
the quality measures, the acceptance quality limit (mA), and the nonacceptance quality
limit (mR), which are specified in accordance with procedures for computing: the inter-
val between mR and the specification limit, mR.L – LSL or USL – mR.U; the discrimina-
tion interval (D) between mA and mR for each specification limit, DL = mA.L – mR.L and
DU = mR.U – mA.U; and the interval between the acceptance quality limits (Δ) for two-sided
specifications that should be equal to or greater than the limiting interval δ × D, such that
Δ = mA.U – mA.L ≥ δ × D. Table 1 of ISO 10725 gives values of δ for two-sided specification
limits when the standard deviations are imprecise. When the standard deviations are
known, use δ = 0.636. This section concludes with a description of the functions of the
responsible authority and its contractual relationship in advance of acceptance sampling
being performed.
In Section 6, the inspection procedures are given for the assessment of the standard devi-
ations, determination of sample sizes, selection and preparation of samples, determination
of the acceptance value, and determination of lot acceptability.
Section 7 provides examples of the use of the standard in the cases of

• Imprecise standard deviation with one-sided specification limit (7.1)

• Imprecise standard deviation with two-sided specification limits (7.2)
• Optional procedure for known standard deviation with one-sided specification
limit (7.3)
218 Acceptance Sampling in Quality Control

• Known standard deviation with one-sided specification limit (7.4)

• Known standard deviations with two-sided specification limits (7.5)
• Revision of discrimination interval (7.6)
• Results from one lot (7.7)
• Results from consecutive lots (7.8)

Problems
A shipment of crushed raw material is received in five special railroad cars, each with two
compartments believed filled separately, which dump from the bottom. It is to be tested for
an impurity that is specified to be less than 5%. Levels of 7% or more cannot be tolerated by
the customer’s progress. Risks of α = .05 and β = .10 are deemed reasonable.

1. If the components of variance were unknown, how might the preliminary sample be
taken?
2. If, in the preliminary sample, X = 5.0% and s = 3%, what additional sample size is
necessary? How should these be taken?
3. Final estimates are X = 5.5% and s = 2%. Should the shipment be accepted?
4. Additional information was gathered on 20 increments from the first five segments
(compartments) in an effort to estimate the testing and reduction variances. The
results were MS3 = .7 and MS4 = .45. What are the estimates of the testing and reduc-
tion variances?
5. Successive shipments are made. After 25 compartments have each been sampled
twice, control charts confirmed the stability of the data. The segment and increment
mean squares were MS1 = 4.75 and MS2 = 2.2. Estimate the segment and increment
variances.
6. Present the mean squares given in Problem 5 in the form of an analysis of variance
table.
7. A shipment of eight railroad cars is received. On the basis of the results from Problems
4 and 5, how many increments should be taken from each compartment if odd and
even increments from each compartment are composited and two tests are made on
each composite?
8. If the grand mean of the results from the eight cars was X = 5.9, should the shipment
be accepted?
9. What would be the standard deviation of a single observation from the ship-
ment  from Problem 8? Construct a 95% confidence interval for the lot mean in
Problem 8.
10. If the lot mean is to be estimated within ±1% in Problem 7, when the cost of sam-
pling a segment is equal to that of an increment, what are the most cost-effec-
tive segment and increment sample sizes disregarding any testing or reduction
variance?
Bulk Sampling 219

References
American Society for Testing and Materials, 2004, Standard Recommended Practice for Sampling
Industrial Chemicals, ASTM Standards E-300, Vol. 15.05, West Conshohocken, PA.
Bennett, C. A. and N. L. Franklin, 1954, Statistical Analysis in Chemistry and the Chemical Industry, John
Wiley & Sons, New York.
Bicking, C. A., 1967, The sampling of bulk materials, Materials Research and Standards, 7(2): 95–116.
Bicking, C. A., 1968, Sampling, Kirk-Othmer Encyclopedia of Chemical Technology, 2nd ed., Vol. 17, John
Wiley & Sons, New York, pp. 744–762.
Bicking, C. A., 1970, ASTM E-105-58 and ASTM E-300-69 standards for the sampling of bulk ­materials,
Journal of Quality Technology, 2(3): 165–173.
Bicking, C. A., 1978, Principles and methods of sampling, in Treatise on Analytical Chemistry, 2nd ed.
(I. M. Kolthoff and P. J. Elving, Eds.), Vol. 1, John Wiley & Sons, New York, Part I, Sec. B,
Chapter 6, pp. 299–359.
Burr, I. W., 1976, Statistical Quality Control Methods, Marcel Dekker, New York.
Cochran, W. G., 1953, Sampling Techniques, John Wiley & Sons, New York.
Davies, O. L. (Ed.), 1954, Statistical Methods in Research and Production, Hafner Publishing Co.,
New York.
Davies, O. L. (Ed.), 1960, Design and Analysis of Industrial Experiments, Hafner Publishing Co.,
New York.
Deming, W. E., 1950, Some Theory of Sampling, John Wiley & Sons, New York.
Duncan, A. J., 1962, Bulk sampling problems and lines of attack, Technometrics, 4(3): 319–344.
Duncan, A. J., 1974a, Quality Control and Industrial Statistics, 4th ed., Richard D. Irwin, Homewood, IL.
Duncan, A. J., 1974b, Bulk sampling, in Quality Control Handbook, 3rd ed. (J. M. Juran, Ed.), McGraw-
Hill, New York, Sec. 25A, pp. 25A-1–25A-14.
Grant, E. L. and Leavenworth, R. S., 1972, Statistical Quality Control, 4th ed., McGraw-Hill, New York.
Grubbs, F. E. and H. J. Coon, 1954, On setting test limits relative to specification limits, Industrial
Quality Control, 10(5): 15–20.
International Organization for Standardization, 2000, Acceptance Sampling Plans and Procedures for the
Inspection of Bulk Materials, (ISO 10725), International Organization for Standardization (ISO),
Geneva, Switzerland.
Ishikawa, K., 1958, How to rationalize the physical material sampling in plants, Reports of statistical
applied research, Japanese Union of Scientists and Engineers, 5(2): 15.
Satterthwaite, F. E., 1946, An approximate distribution of estimates of variance components,
Biometrika Bulletin, 2: 110–114.
Schilling, E. G., 1973, A systematic approach to the analysis of means, Part I. Analysis of treatment
effects, Journal of Quality Technology, 5(3): 93–108.
Tanner, L. and M. Lerner, 1951, Economic Accumulation of Variance Data in Connection with Bulk Sampling,
ASTM STP 114, American Society for Testing and Materials, Philadelphia, PA, pp. 8–12.
Williams, W. H., 1978, A Sampler on Sampling, John Wiley & Sons, New York.
10
Sampling by Variables for Proportion Nonconforming

The distinction between discrete and continuous variables involves good grammar as
well as good statistics. We state how many we have of a discrete variable and how much
when the variable is continuous. We may be interested in how many cans of soup were
underweight by as much as a milligram, or how many rivets were off center by as much
as 0.5 mm. These statements imply that continuous (measurement) variables can be sub-
jected to an attribute (go/no-go) type of test simply by counting the number of items in
a sample beyond some limit. Thus, attributes sampling plans could be applied in these
two cases.
Alternatively, if the shape of the underlying distribution of individual ­measurements
were known, acceptance sampling could be performed directly on the measure-
ments  themselves. Such procedures form the basis for variables sampling plans for
­proportion nonconforming and, when applicable, provide a considerable savings in
sample size.
The basic idea of variables sampling for proportion nonconforming is to show that the
sample results are sufficiently far within the specification limit(s) to assure the acceptabil-
ity of the lot with reasonable probability.
Variables plans involve comparing a statistic, such as the mean X, with an acceptance
limit A in much the same way that the number nonconforming, d, is compared to an accep-
tance number, c, in attributes plans. A comparison of the procedures involved in variables
and attributes sampling plans is shown in Figure 10.1.
Some of the advantages of variables sampling are as follows:

1. The same protection with smaller sample size than attributes

2. Feedback of data on the process that produced the units
3. More data available in waiver situations
4. Extent of conformity of each unit given weight in application of the plan
5. Increased likelihood of errors in measurement being detected

Some of the disadvantages of variables sampling are as follows:

1. Dependence of the results on correctness of the assumption of the shape of the

underlying distribution of measurements
2. Applicable to one (only) characteristic at a time
3. Higher inspection cost per unit
4. Higher clerical cost per unit
5. Possibility of no nonconforming unit found to show the producer after rejection

221
222 Acceptance Sampling in Quality Control

Take sample Take sample

Compare Compare
d with c X with A

d≤c d>c X≤A X>A

Accept Reject Accept Reject

(a) (b)

FIGURE 10.1
Comparison of attributes and variables sampling. (a) Attributes, single sampling; (b) variables, single sampling
(upper specification limit). (Reprinted from Schilling, E.G., Qual. Progr., 7(5), 16, 1974b. With permission.)

The principal advantage of variables plans over attributes is reduction in sample size. For
example, in comparing average sample sizes for plans matched to the single-sampling
attributes plan n = 50, c = 2, we have, for a single specification limit:

Plan Average Sample Number (ASN)

Single attributes 50
Double attributes 43
Multiple attributes 35
Sequential attributes 33.5
Variables (σ unknown) 27
Variables (σ known) 12

Specification Limits
Specification limits can be of two types. A single specification limit implies only one
boundary value for acceptability, either upper U or lower L. Thus, a measurement does not
­conform to the specification limit if

X > U

for an upper specification limit, or if

X < L

for a lower specification limit. Double specification limits place both upper and lower
boundary values on the acceptability of a measurement. That is, the measurement X is
acceptable if and only if

L < X < U
Sampling by Variables for Proportion Nonconforming 223

Assumptions and Theory

Probably, the most important consideration in applying variables sampling plans is the
requirement that the shape of the underlying distribution of measurements to which
the plan is to be applied must be known and stable. This means that probability plots or
statistical tests on past data must show that the distribution of measurements involved
actually is that assumed by the plan. Control chart evidence also is desirable to indicate
its stability. For a known distribution, it is the underlying distribution of measurements
that relates the proportion of units outside the specification limit to a fixed position of the
population mean of the measurements. Variables plans for a process parameter may then
be used to confirm or deny that the population mean is in the proper position. In a crude
way, this is how variables sampling works. In fact, some plans are devised in just this way.
It must be emphasized that the underlying distribution must be known to be that assumed
by the plan for variables sampling to be properly applied.
The basic theoretical nature of the variables acceptance sampling plans is illustrated
in Figure 10.2, which involves an upper specification limit and assumes the underlying
distribution of individual measurements to be normal. If the procedure of Figure 10.1 is
applied, the mean X of a sample of n measurements is compared to an acceptance limit A
and the lot accepted if X is not greater than A. Figure 10.2 shows that if the distribution
of individual measurements is as shown, with σ known, a proportion p of the product
above the upper specification limit U implies the mean of the distribution must be fixed at
the position indicated by μ. Sample means of size n then would be distributed about μ as
shown and so the probability of obtaining an X not greater than A is as indicated by the
shaded area in the diagram. Published plans for known standard deviation often are given
in terms of sample size and k, the distance in units of the standard deviation, between the
(upper) specification limit U and the acceptance limit A. From Figure 10.2, we see

U-A
k= = zU - z A
s

A U
Probability of
acceptance on
kσ Distribution of
variables sample
sample means

Distribution of
individuals
p

Original units
µ A U x of measurement
kσ
Transformed
O zA zU z z units
k

FIGURE 10.2
Distributions in variables sampling. (Reprinted from Schilling, E.G., Qual. Progr., 7(5), 16, 1974b. With permission.)
224 Acceptance Sampling in Quality Control

1.0

0.9

0.8

0.7

0.6

0.5 V A
Pa

0.4

0.3

0.2

0.1

0
0 5 10 15 20 25 30
p

FIGURE 10.3
Comparison of OC curves. For variables, V, n = 7, and k = 1.44. For attributes, A, n = 20, and c = 1.

for the distribution of individual measurements positioned as shown where the z values
are taken from the standard normal table. The situation is analogous, but reversed, for a
lower specification limit.
Using Figure 10.2 and normal probability theory, the probability of acceptance Pa can
be calculated for various possible values of p, the proportion nonconforming. Figure 10.3
shows the operating characteristic (OC) curve of the variables plan n = 7, k = 1.44 for known
standard deviation compared to that of the attributes plan n = 20, c = 1. OC curves of vari-
ables plans are generally considered to be Type B.
It can be seen that the two OC curves are well matched, that is they give about the same
protection. The variables plan, however, uses only about a third as large a sample size as
the attributes plan. Thus, the variables plan appears superior. It must be remembered,
however, that the superiority of the variables plan rests on the assumption of the normality
of the underlying distribution of the measurements. If this assumption cannot be justified,
the variables plans may give unreliable results and recourse must be either to an attributes
or to a mixed variables–attributes plan.
The danger involved in using a variables plan that assumes normality when, in fact,
the underlying distribution of individual measurements is actually nonnormal is illus-
trated in Figure 10.4. This shows the proportion of the product beyond z standard devi-
ation units from the mean to be heavily dependent on the shape of the distribution.
A variety of distributions is represented by various shape parameters for the family of
Weibull distributions. Note that the tail area beyond three standard deviations is over
2% for a Weibull distribution with shape parameter 0.5, while it is 0.13% for a normal
distribution.
Sampling by Variables for Proportion Nonconforming 225

Normal
β = 5.0
β = 3.5

β = 2.0

β = 1.5
0.1 β = 1.0
Percent tail area

0.5 β = 0.5
1.0

2.5

5.0

1.0 2.0 3.0 4.0 5.0

Standard deviation units from mean (z)

FIGURE 10.4
Curves of upper tail areas of several Weibull distributions. (Reprinted from Schilling, E.G., Qual. Progr., 7(5), 16,
1974b. With permission.)

Operation
X Method
The simplest application of variables plans for proportion nonconforming is when a single
specification limit is involved and the standard deviation is known. In this case, a straight-
forward procedure, which we shall call the X method, may be employed. It requires that
the sample size and an acceptance constant k be specified and that σ be known. An accep-
tance limit A for X is set a distance kσ within the specification limit. The procedure, then,
is as shown in Table 10.1.

k Method
The X method is actually a special case of what is called the k method. The proce-
dure involved in the k method is shown in Table 10.2. The more general k method may

TABLE 10.1
X Method
Lower Specification Limit Upper Specification Limit

1. Set A = L + ks 1. Set A = U − ks
2. Select a random sample of size n 2. Select a random sample of size n
3. Compute X 3. Compute X
4. If X ≥ A, accept the lot; if X < A, reject the lot 4. If X ≤ A, accept the lot; if X > A, reject the lot
226 Acceptance Sampling in Quality Control

TABLE 10.2
k Method (Given n, k)
Lower Specification Limit Upper Specification Limit

1. Select a random sample of size n 1. Select a random sample of size n

( )
2. Compute z = X - L /s, for σ known, or z = X - L /s ,( ) ( )
2. Compute z = U - X /s, for σ known, or
for σ unknown
( )
z = U - X /s, for σ unknown
3. If z ≥ k, accept the lot; if z < k, reject the lot 3. If z ≥ k, accept the lot; if z < k, reject the lot
4. Equivalently, if X − kσ ≥ L, accept the lot; if X – kσ < L, 4. Equivalently, if X + kσ ≤ U, accept the lot; if
reject the lot. If σ is unknown, use appropriate values X + kσ > U, reject the lot. If σ is unknown, use
of n and k with X − ks as above appropriate values of n and k with X + ks as above

be used  when the standard deviation is not known simply by substituting the sample
­standard deviation

å(x - x)
2

s=
n -1

for σ in the known standard deviation procedure and choosing an appropriate value of
k and sample size n for the unknown standard deviation case. It may be applied in two
alternative but equivalent ways as shown in Table 10.2.
It can be seen that the X method is a special case of the k method, since for a lower speci-
fication limit acceptance would occur if

X - ks ³ L
X ³ L + ks
X³A

Furthermore, note that A is a fixed constant only if σ is known and so the k method is the
only real alternative for the case of unknown standard deviation. The X method, however,
offers a simpler approach for lot acceptance when it is applicable. It can be directly pre-
sented diagrammatically as in Figure 10.1. A diagrammatic representation of the relation-
ship between the X and k methods, when σ is known as presented by Schilling in Juran
(1999) Quality Control Handbook, is given in Figure 10.5.

Double Specification Limits

When double specification limits are involved, the procedure of implementing variables
sampling plans becomes somewhat more complicated. This is because, when variability
is large relative to the distance between the lower and upper specification limits, it is pos-
sible to have a significant proportion of product outside both specification limits at the
same time. Clearly, if the specification limits are sufficiently far apart, two separate single
specification limit plans may be used since the occurrence of the product outside either of
the limits will be mutually exclusive. That is, the product may be outside one or the other
of the specification limits, but not both.
When the standard deviation is known, a simple procedure, modified from that sug-
gested by Duncan (1974), may be used to determine if two separate single specification
Sampling by Variables for Proportion Nonconforming 227

Distribution of
Probability sample means
of acceptance

Distribution of
individual
measurements
p

µ (U – kσ) U

(a) kσ

Probability of
acceptance Distribution
of (X + kσ)
Distribution of
individual
measurements

µ (µ + kσ) U

(b) kσ

FIGURE 10.5
X and k methods are compared. (a) U − ks method. (b) X + kσ method. (Reprinted from Juran, J.M., Ed., Quality
Control Handbook, 5th ed., McGraw-Hill, New York, 1999; Schilling, E.G., Sampling by variables, Section 25.
With permission.)

limit plans may be used. Suppose a plan is to be instituted with producer’s quality level,
PQL = p1, and consumer’s quality level, CQL = p2. The method is as follows:

1. Compute zp = (U − L)/2σ.
2. Find p* from the normal table as the upper tail area corresponding to zp. This is the
minimum proportion nonconforming outside one of the specification limits.
3. Criteria
a. If 2p* ≤ p1/2, use two single-limit plans.
b. If p1/2 < 2p* ≤ p1, the specifications may be too close to prevent nonconformi-
ties on both sides when the distribution is centered. Using normal probability
theory, determine the split of proportion nonconforming outside the upper
limit pU and outside the lower limit pL, which will sum to p1 as the distribution
is moved between the specifications. Use the larger of these two proportions as
p1 in two single-limit plans together with specified p2.
c. If p1 < 2p* < p2, the specifications of the plans must be reconsidered.
d. If 2p* ≥ p2, the product should be rejected outright.
228 Acceptance Sampling in Quality Control

For example, suppose a plan is desired to check on the resistance of a certain electrical
device. The specifications are U = 100 Ω and L = 90 Ω with p1 = .01 and p2 = .05. The stan-
dard deviation is known to be 1.5 Ω. Then

1. zp = (100 − 90)/2(1.5) = 3.33.

2. From the normal table, p* = .0004.
3. Since .0008 < p1/2 = .005, two single-sided plans are appropriate.

When the standard deviation is unknown, the double specification limit problem becomes
still more difficult since there are two random quantities to be considered in the acceptance
decision: the mean X and the standard deviation s. In such a situation, it is customary to
check the sample standard deviation against the maximum value before proceeding to
check against two separate single-limit plans. The so-called maximum standard deviation
(MSD) becomes part of the acceptance procedure. It may be approximated as follows from
a procedure suggested by Wallis (1950):

1. Find the upper tail normal area p** corresponding to z*p = k.

2. Find z**
p corresponding to a normal upper tail area of p**/2.

3. Maximum standard deviation is MSD ≅ (U − L)/2z** p .

The acceptance criteria for double specification limits then add the following initial check
to the procedure for two single-limit plans.
Check(s) against MSD:

1. If s ≤ MSD, use two separate single specification limit plans.

2. If s > MSD, reject the lot since the standard deviation is too large to be consistent
with the acceptance criteria.

The idea can be expressed graphically as shown in Figure 10.6. The sample standard devia-
tion is plotted against the sample mean. The x-axis, the MSD, and the two single-sided
acceptance sampling criteria (X + ks = U and X − ks = L) form an acceptance polygon. If the
point (X, s) plots within the polygon, the lot should be accepted. If not, the lot is rejected.
Actually, the polygon shown is an approximation of a more accurate acceptance region,
the development of which was attributed by Wallis (1950) to Kenneth J. Arnold. The region
is defined by the points (X¢, Y¢ ) such that for any two proportions p¢0 and p0² summing
to p** corresponding to z*p = k :

Uz p0¢ + Lz p²0
X¢ =
z p0¢ + z p²0
U-L
s¢ =
z p0¢ + z p²0

( )
Given a point X¢, s¢ on one side of the “polygon,” of course, the symmetric point for a
given s′ is

X¢¢ = U + L - X¢
Sampling by Variables for Proportion Nonconforming 229

Reject
MSD

Reject Accept Reject

s

sL = 1 x – L sU = 1 x + U
k k k k

L U
x

FIGURE 10.6
Acceptance polygon.

Such a polygon is shown in Figure 10.6 in dotted lines. The dotted curve will intersect the
straight sides of the original polygon at approximately

U-L
s=
3+k

It can be seen that the first polygon approximation to the acceptance region is slightly loose
in that it overstates the acceptance region. The solution for s′ at X¢ = (U + L )/2 results in the
approximation for MSD given earlier.
Wallis (1950) outlined the method for determining the more accurate acceptance region
as follows:

1. Determine n and k from the usual one-sided procedures (given in the following
section on Selection, Formulas subsection).
2. Find the indifference quality p**, which is the probability that a standard normal
deviate will exceed z*p = k .
3. Divide p** into two parts p1¢ and p¢2 such that p1¢ + p¢2 = p **. Each pair p1¢ and p¢2 leads
to a point on the acceptance region boundary.
4. Find z1 and z2 as normal deviates corresponding to the upper tail areas p1¢ and p¢,
2
respectively.
5. Substitute each pair, z1 and z1, into the equation

Uz1 + Lz2
X¢ =
z1 + z2
U -L
s¢ =
z1 + z2

6. Plot enough points to define the acceptance region.

230 Acceptance Sampling in Quality Control

3 MSD
2.81
Reject Reject
s

2
sL = 0.694x – 62.5 sU = 0.694x + 69.4

Accept
1

90 95 100
x

FIGURE 10.7
Acceptance polygon: Example.

For example, consider the unknown standard deviation plan

n = 13 k = 1.44

to be applied against the previous specifications for resistance, U = 100 Ω, L = 90 Ω. The

acceptance polygon would appear as in Figure 10.7.
The polygon is constructed as follows:

1. The two acceptance lines are

1 L
sL = X - = .694X - 62.5
k k
1 U
sU = - X + = -.694X + 69.4
k k

2. The MSD is determined as follows:

a. z p* = 1.44, so p** = .0749.
p **
b. = .0375, so z** p = 1.78.
2
c. MSD = (U − L)/2z** p = 10/[2(1.78)] = 2.809.

3. The two single specification lines will intersect when

U + L 100 + 90
X= = = 95
2 2
Sampling by Variables for Proportion Nonconforming 231

TABLE 10.3
Polygon Derived by the Wallis Method
(1) (2) (3) (4) (5) (6) (7) (8) X = éë(U + L )/2ùû MSD
p¢1 p¢2 z1 z2 (3) + (4) (3) − (4) (6)/(5) 2/(5) + [(U − L)/2] * (7) s = [(U − L)/2] * (8)

.0375 .0375 1.78 1.78 3.56 0.00 .0000 .5618 95.00 2.809
.0325 .0425 1.85 1.72 3.57 0.13 .0364 .5602 95.18 2.801
.0275 .0475 1.92 1.67 3.59 0.25 .0696 .5571 95.35 2.786
.0225 .0525 2.00 1.62 3.62 0.38 .1050 .5525 95.52 2.762
.0175 .0575 2.11 1.58 3.69 0.53 .1436 .5420 95.72 2.710
.0125 .0625 2.24 1.53 3.77 0.71 .1883 .5305 95.94 2.652
.0075 .0675 2.43 1.49 3.92 0.94 .2398 .5102 96.20 2.551
.0025 .0725 2.81 1.46 4.27 1.35 .3162 .4684 96.58 2.342
.0001 .0749 3.89 1.44 5.33 2.45 .4597 .3752 97.30 1.876

which corresponds to a height (s) of

s = -.694 ( 95 ) + 69.4 = 3.47

4. The lines are drawn to obtain the polygon.

Using the Wallis method, the more accurate acceptance region can be obtained using the
tabulation shown in Table 10.3, given p** = .075.

Selection
Tables
Extensive tables of variables plans for proportion nonconforming and defective will be
found in the well-known military standard MIL-STD-414 (U.S. Department of Defense
1957). The OC curves presented therein can be used to select a plan appropriate to the
sampling situation. MIL-STD-414 and its derivatives will be discussed in a later chapter.
Procedures for variables plans assuring lot tolerance percent defective (LTPD) or average
outgoing quality limit (AOQL) protection have been developed by Romig (1939) and are
presented in his PhD dissertation.
Appendix Table T10.2, computed by Sommers (1981), gives acceptance criteria for
single-sampling variables plans as well as matched double-sampling variables plans
(discussed later). Sample sizes are shown for standard deviation known (nσ) and
unknown (ns) for a given acceptance constant (k). The table is indexed by PQL and CQL
with α = .05 and β = .10. Plans were derived using the computational formulas given
in the following. The Wallis approximation was used for standard deviation unknown
plans.
The selection of p1 and p2 values used by Sommers was made to be the same as those
used by the Statistical Research Group, Columbia University (1947) in a similar tabu-
lation to facilitate a comparison with sequential plans. As an example of the use of
Appendix Table T10.2, it will be seen that for p1 = .01 and p2 = .05 with α = .05 and β = .10,
232 Acceptance Sampling in Quality Control

the following plans are given:

Known standard deviation: n = 19, k = 1.94

Unknown standard deviation: n = 54, k = 1.94

Matching binomial attributes and narrow limit plans have been tabulated by Schilling and
Sommers (1981) for the same selection of p1 and p2 values and appear with the single-
sampling variables plans in Appendix Table T13.3.

Formulas
The acceptance criteria for variables plans may be readily determined from computational
formulas for n and k. In these formulas, the standard normal deviates, z, representing

zp1 as the Area of p1 in upper tail

zp2 as the Area of p2 in upper tail
zα as the Area of α in upper tail
zβ as the Area of β in upper tail

The values of k and n are obtained from the following formulas. It will be seen that the
formula for n depends upon the state of knowledge of the standard deviation. The results
should always be rounded up:

z p2 za + z p1 zb
k=
za + zb

for σ known
2
æ z +z ö
n = çç a b ÷÷
è z p1 - z p 2 ø

for σ unknown
2
æ z +z ö æ k2 ö
n = çç a b ÷÷ ç 1 + ÷
è z p1 - z p2 ø è 2 ø

The latter formula is due to Wallis (1947) and corrects the sample size found for σ known
by the factor (1 + k2/2) obtained from the noncentral t-distribution. Although this is an
approximation, it is surprisingly accurate and certainly adequate for practical purposes. It
can be shown to be extremely accurate when compared to the exact values obtained using
the noncentral t-distribution. This can be seen from Appendix Table T10.3 prepared by the
Columbia Statistical Research Group, Columbia University (1947, p. 65) for unknown stan-
dard deviation plans where both approximate and exact PQL and CQL are given.
Suppose p1 = .018 and p2 = .18. For α = .05 and β = .10, the formulas give

0.92 ( 1.64 ) + 2.10 ( 1.28 )

k= = 1.44
1.64 + 1.28
Sampling by Variables for Proportion Nonconforming 233

for σ known
2
æ 1.64 + 1.28 ö
n=ç ÷ = 6.12 ~ 7
è 2.10 - 0.92 ø

and for σ unknown

2
æ 1.64 + 1.28 ö æ 1.44 2 ö
n=ç ÷ ç1+ ÷
è 2.10 - 0.92 ø è 2 ø

= 6.12 ( 2.04 ) = 12.5 ~ 13

Jacobson Nomograph for Plan Selection

A nomograph also exists for determining variables plans for proportion nonconform-
ing. Due to Jacobson (1949), it can be used in a manner similar to that of Larson (1966)
for attributes plans. It is based on the Wallis formula. It is shown in Figure 10.8. To use

tA
0.001
3
100
05
00
200 Sam 0.01
tp ple
0.001 100 siz 0.02
3 e, n 2
80
0.002 40 0.05
30
20
15 0.10
10
Probability of acceptance, PA (1 – α)

0.005 1
5
3.5 0.20
0.01 3.0
Fraction defective, p (normal law distribution)

0.30
2.5 0.40
+
0.02 2.0 0.50 0
2 –
0.60
1.0 0.70
0.05 k 0.80
10 1
0.90
0.10 5
0.95

1 0 0.98 2
5 0.99
0.20
10
15
0.30 20 n p pA 3
30 ize, n β 0.999
60 ple s p1
0 Sam k
0.40 10
0
20 p2
00 1–α
0
0.50 01

FIGURE 10.8
Jacobson nomograph for variables plans for proportion defective. (After Jacobson, L.J., Ind. Qual. Control, 63, 23, 1949.)
234 Acceptance Sampling in Quality Control

p pa

13 0.10

0.018

1.44

0.18
0.95

7
13

FIGURE 10.9
Application of Jacobson nomograph.

the nomograph to derive a plan, given p1, p2, α, and β, proceed as follows for the case of
σ unknown:

1. Connect p1 on the left fraction defective axis with (1 − α) on the right probability of
acceptance axis.
2. Connect p2 on the left fraction defective axis with β on the right probability of
acceptance axis.
3. From the point of intersection of the two lines, read the sample size n and the
acceptance constant k.

When the standard deviation is known, the nomograph can be employed to derive a plan
as follows:

1. Draw the two lines and obtain the point of intersection following the previous
procedure and read the value of k.
2. Draw a line through the point of intersection parallel to the left and right vertical
axes and read the value of n at the intersection of this line with the bottom sample
size scales on the chart (i.e., where k = 0). This follows since for k = 0, the Wallis
formula reduces to that of the known standard deviation plan.

Figure 10.9 shows the derivation of the σ known plan k = 1.44, n = 7 when p1 = .018, p2 = .18,
α = .05, and β = .10. The dotted line shows the location of the sample size of 13 for the plan
when the standard deviation is unknown.

Measures
Jacobson Nomograph for Operating Characteristics
The Jacobson nomograph shown in Figure 10.8 may be used to derive the OC curve for a
variables plan for proportion nonconforming. The procedure differs slightly between the
Sampling by Variables for Proportion Nonconforming 235

case of standard deviation known and that of standard deviation unknown. The method is
as follows for a plan specified by n and k.
Standard deviation unknown

1. Locate the point (n, k).

2. Locate the proportion nonconforming p on the left fraction defective axis.
3. Intersection of a line through the two points with the right axis gives the probabil-
ity of acceptance.
4. The process may be reversed to give the proportion nonconforming associated
with a given probability of acceptance.

Standard deviation known

1. Locate the sample size, on the bottom sample size axis (i.e., where k = 0).
2. Draw a line through the sample size parallel with the right and left vertical axes.
3. Intersection of the line drawn with the appropriate curve for k gives the point (n, k)
for the known standard deviation case.
4. Follow the unknown standard deviation procedure from step 2.

It will be observed that the two lines drawn in Figure 10.9 may be regarded as represen-
tative of the procedure for the unknown standard deviation plan n = 13, k = 1.44 or for
the known standard deviation plan n = 7, k = 1.44, respectively. They show the 10th and
95th percentage points of the OC curve.

Calculation: σ Known
When the standard deviation is known, the calculation of OC curves for variables plans
may be performed using the normal distribution. Referring to Figure 10.2, which describes
the X method, we see that for any given proportion nonconforming p, the population mean
μ must be a fixed distance zUσ from the upper specification limit. Furthermore, the distance
from the population mean to the acceptance limit A is zAσ, where

zU = standard normal deviate for the distribution of individual measurements

­corresponding to proportion nonconforming in the upper tail
zA = standard normal deviate for the distribution of individual measurements
­corresponding to the acceptance limit A

As noted previously, k = zU − zA; so zA = zU − k.

To find the probability of acceptance, the probability of obtaining a mean below A must
be determined. But means are distributed according to the distribution of sample means,
which is also shown in Figure 10.2. To get the probability of acceptance, the distance
between A and μ must be found in terms of the standard deviation of the distribution of
sample means. Recall

s
sX =
n
236 Acceptance Sampling in Quality Control

so

nsX = s

and the distance zA between μ and A in terms of the standard deviation of individuals may
be used to obtain the distance in terms of the standard deviation of means since

s
z As = nz A = zAsX
n

where the conventional bar denoting average is used to indicate that zA is from the distri-
bution of sample means. We then have

zA = nz A or zA = n ( zU - k )

Thus, for any given value of p, the probability of acceptance can be determined as follows
for an upper specification limit:

1. Determine zU from p.
2. Obtain zA = zU − k.
3. Convert zA to the distribution of sample means as zA = nz A .
4. The probability of a normal variate exceeding zA is the probability of rejection. Its
complement, the probability of a result less than zA , is the probability of acceptance.

For a lower specification limit, this becomes

1. Determine zL from p.
2. Obtain zA = zL + k.
3. Convert zA to the distribution of sample means as zA = nz A .
4. The probability of a normal variate equal to or exceeding zA is the probability of
acceptance. Its complement, the probability of a result less than zA , is the probabil-
ity of rejection.

For example, consider the plan n = 7, k = 1.44; Table 10.4 shows the computation of the
probability of acceptance and compares the results to the attributes plan n = 20, c = 1. A plot
of both OC curves was given in Figure 10.3.

Calculation: σ Unknown
As suggested by Wallis (1950), the OC curve for an unknown standard deviation plan,
specified by k and ns, can be approximated by using the known standard deviation proce-
dure earlier with

ns
n=
1 + k 2/2
Sampling by Variables for Proportion Nonconforming 237

TABLE 10.4
Calculation of Probability of Acceptance: n = 7 and k = 1.44
Proportion Nonconforming Probability of Acceptance

P zU z A = zU − k z A = nz A Pr = 1 − Pa Pa = 1 – Pr n = 20, c = 1

.0075 2.43 0.99 2.62 .0044 .9956 .99

.018 2.10 0.66 1.75 .0401 .9599 .95
.027 1.93 0.49 1.30 .0968 .9032 .90
.048 1.66 0.22 0.58 .2810 .7190 .75
.083 1.39 −0.05 −0.13 .5517 .4483 .50
.129 1.13 −0.31 −0.82 .7939 .2061 .25
.181 0.91 −0.53 −1.40 .9192 .0808 .10
.216 0.78 −0.66 −1.75 .9599 .0401 .05
.289 0.56 −0.88 −2.33 .9901 .0099 .01

or using a slightly more accurate form of the Wallis (1947) approximation to relate the
sample sizes of known (nσ) and unknown (ns) standard deviation plans
ns
ns =
(
1 + k ns/2 ( ns - 1)
2
)
So that if z1 − Pa denotes the upper tail standard normal deviate corresponding to the prob-
ability of rejection (1 − Pa), Wallis (1947) gives the following relation:

1 k2
z A = zU - k = z1- Pa +
ns 2 ( ns - 1)

Note that the standard normal deviate corresponding to the probability of acceptance is

zPa = - z1- Pa

When the standard deviation is unknown, the exact calculation of the OC curve becomes
less straightforward. The statistic (shown here for an upper specification limit)

U -X
t=
s

has a Student’s t-distribution only for 50% nonconforming. For all other values of propor-
tion nonconforming, the statistic is distributed by the noncentral t-distribution, the dis-
tribution involved in calculating the OC curve when the standard deviation is unknown.
For a variate t that can be expressed as
z+d
t= ,
w

where
z is the distributed standard normal (μ = 0, σ = 1)
w is the distributed χ2/f with f degrees of freedom independent of z
δ is a constant
238 Acceptance Sampling in Quality Control

The noncentral t-probability distribution function has been expressed by Resnikoff and
Lieberman (1957) as

( f + 1)/2
) æç ö
t0 ¥ 2
f! ( )(
- ( 1/2 ) f d2 / f + t 2 f v f -(1/2)æçè v - dt/ f + t 2 ö÷

P ( f , d, t0 ) = ( f -1)/2
2 G ( f /2 ) pf òe
-¥
(
÷
ç f + t2 ÷
è ø ) ò
0
f!
e ø
dv dt

where
f is the degrees of freedom in t
δ is the noncentrality parameter
t is the random variate

Resnikoff and Lieberman (1957) have extensively tabulated the noncentral t-distribution.
A sample page is shown as Figure 10.10. For a noncentrality parameter d = f + 1K p , they
give values of the distribution function

(
Pr t/ f £ x )
tabulated by p and x, where Kp = standard normal deviate exceeded with probability p.
To use the noncentral t-distribution to obtain the probability of acceptance when the
standard deviation is unknown, the acceptance criterion may be expressed as

U-X
³k
s
æU-X ö
nç ÷ ³ nk
è s ø

and

æ n (U - m )
ç -
n X -m ( ) ö÷ s ³ nk
ç s s ÷s
è ø

The left-hand side of the inequality is distributed noncentral t with

f = n -1
n (U - m )
d= = nzU
s

so that the probability of acceptance is simply

Pa = Pr t ³ nk( )
Sampling by Variables for Proportion Nonconforming 239

Degrees of freedom 12
P
x 0.2500 0.1500 0.1000 0.0650 0.0400 0.0250 0.0100 0.0040 0.0025 0.0010
−0.50 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.45 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.40 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.35 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.30 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.25 0.0006 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.20 0.0010 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.15 0.0017 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.10 0.0029 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
−0.05 0.0047 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.00 0.0075 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.05 0.0119 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.10 0.0184 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.15 0.0279 0.0007 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.20 0.0413 0.0012 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.25 0.0594 0.0022 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.30 0.0832 0.0039 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.35 0.1134 0.0067 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.40 0.1503 0.0110 0.0009 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.45 0.1939 0.0177 0.0016 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.50 0.2435 0.0273 0.0030 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.55 0.2983 0.0407 0.0052 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.60 0.3567 0.0587 0.0088 0.0009 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.65 0.4173 0.0820 0.0143 0.0016 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
0.70 0.4784 0.1109 0.0223 0.0030 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000
0.75 0.5383 0.1456 0.0336 0.0052 0.0005 0.0000 0.0000 0.0000 0.0000 0.0000
0.80 0.5958 0.1858 0.0488 0.0087 0.0010 0.0001 0.0000 0.0000 0.0000 0.0000
0.85 0.6496 0.2310 0.0685 0.0141 0.0018 0.0002 0.0000 0.0000 0.0000 0.0000
0.90 0.6992 0.2803 0.0930 0.0217 0.0032 0.0004 0.0000 0.0000 0.0000 0.0000
0.95 0.7439 0.3326 0.1227 0.0323 0.0055 0.0008 0.0000 0.0000 0.0000 0.0000
1.00 0.7837 0.3866 0.1572 0.0464 0.0091 0.0015 0.0000 0.0000 0.0000 0.0000
1.05 0.8185 0.4412 0.1963 0.0644 0.0143 0.0028 0.0001 0.0000 0.0000 0.0000
1.10 0.8487 0.4952 0.2394 0.0866 0.0217 0.0047 0.0002 0.0000 0.0000 0.0000
1.15 0.8745 0.5476 0.2856 0.1133 0.0317 0.0077 0.0003 0.0000 0.0000 0.0000
1.20 0.8964 0.5976 0.3341 0.1442 0.0447 0.0120 0.0006 0.0000 0.0000 0.0000
1.25 0.9148 0.6445 0.3839 0.1792 0.0612 0.0182 0.0012 0.0001 0.0000 0.0000
1.30 0.9302 0.6879 0.4341 0.2177 0.0813 0.0265 0.0021 0.0001 0.0000 0.0000
1.35 0.9429 0.7276 0.4836 0.2593 0.1052 0.0373 0.0035 0.0002 0.0001 0.0000
1.40 0.9534 0.7635 0.5319 0.3031 0.1329 0.0510 0.0056 0.0004 0.0001 0.0000
1.45 0.9621 0.7956 0.5781 0.3484 0.1641 0.0679 0.0087 0.0008 0.0002 0.0000
1.50 0.9691 0.8240 0.6220 0.3944 0.1984 0.0881 0.0131 0.0014 0.0004 0.0000
1.55 0.9749 0.8491 0.6629 0.4405 0.2356 0.1115 0.0190 0.0024 0.0007 0.0001
1.60 0.9796 0.8709 0.7009 0.4859 0.2749 0.1382 0.0267 0.0038 0.0013 0.0001
1.65 0.9834 0.8900 0.7357 0.5300 0.3159 0.1679 0.0365 0.0059 0.0021 0.0003
1.70 0.9865 0.9064 0.7674 0.5725 0.3579 0.2004 0.0486 0.0089 0.0034 0.0005
1.75 0.9890 0.9205 0.7959 0.6129 0.4003 0.2352 0.0633 0.0129 0.0052 0.0008
1.80 0.9911 0.9326 0.8215 0.6509 0.4426 0.2719 0.0805 0.0182 0.0078 0.0013
1.85 0.9927 0.9429 0.8443 0.6864 0.4842 0.3099 0.1005 0.0250 0.0113 0.0021
1.90 0.9941 0.9517 0.8645 0.7192 0.5248 0.3489 0.1230 0.0335 0.0159 0.0033
1.95 0.9952 0.9592 0.8824 0.7495 0.5639 0.3884 0.1480 0.0440 0.0219 0.0050
2.00 0.9961 0.9655 0.8980 0.7771 0.6013 0.4278 0.1753 0.0564 0.0293 0.0073
2.05 0.9968 0.9709 0.9117 0.8021 0.6367 0.4667 0.2047 0.0711 0.0384 0.0104
2.10 0.9974 0.9754 0.9237 0.8248 0.6701 0.5049 0.2359 0.0879 0.0493 0.0144
2.15 0.9978 0.9792 0.9341 0.8452 0.7013 0.5419 0.2685 0.1069 0.0622 0.0194
2.20 0.9982 0.9825 0.9431 0.8634 0.7302 0.5776 0.3022 0.1280 0.0770 0.0258
2.25 0.9985 0.9852 0.9510 0.8797 0.7569 0.6116 0.3367 0.1511 0.0939 0.0335
2.30 0.9988 0.9875 0.9577 0.8942 0.7815 0.6439 0.3716 0.1761 0.1127 0.0427
2.35 0.9990 0.9894 0.9636 0.9071 0.8040 0.6744 0.4066 0.2028 0.1335 0.0535
2.40 0.9992 0.9910 0.9386 0.9184 0.8245 0.7030 0.4414 0.2310 0.1561 0.0661
2.45 0.9993 0.9924 0.9730 0.9285 0.8431 0.7297 0.4756 0.2603 0.1804 0.0803

FIGURE 10.10
Resnikoff–Lieberman table. (Reprinted from Resnikoff, G.J. and Lieberman, G.J., Tables of the Non-Central
t-­Distribution, Standard University Press, Stanford, CA, 1957, p. 327. With permission.)
240 Acceptance Sampling in Quality Control

and since tables are for t/ f , we have*

æ t n ö
Pa = 1 - Pr çç £ k÷
è n -1 n - 1 ÷ø
æ n ö
= 1 - Pr çç n - 1, nzU , k÷
è n - 1 ÷ø

Hence, using the Resnikoff–Lieberman tables, proceed as follows to evaluate the OCs of an
upper specification limit plan specified by n and k for proportion nonconforming p.

1. Degrees of freedom are f = n − 1; select the table of the probability integral with f
degrees of freedom.
2. Compute n/( n - 1) k ; this is x in the table.
3. For the value of p given and x calculated, obtain the probability of rejection P(R)
from the table.
4. The complement is the probability of acceptance Pa = 1 − P(R).

For example, ­consider the plan n = 13, k = 1.44 to be evaluated at p = .025.

1. f = 13 − 1 = 12.
13

2. x =

(1.44 ) = 1.50.
12
3. P(R) = .0881.
4. Pa = .9119.

This value may be obtained from Figure 10.10.

For reasons of symmetry, the evaluation of the OC curve for a lower specification limit
plan is analogous. It will be seen that the specification limit (upper or lower) does not
appear in the steps for determining the probability of acceptance. The procedure will work
for either.

Double Specification Limits

Evaluation of the probability of acceptance in the two-sided specification limit case is
­analogous to the single specification limit procedure when the standard deviation is
known. This amounts to an evaluation of the probability of rejection and the proportion
nonconforming outside each of the specification limits over various fixed positions of
the population mean. Their sum gives values of p and Pa, which may be plotted as the
OC curve.
Consider the earlier double specification limit example in which U = 100 Ω, L = 90 Ω,
and suppose σ = 2.0. The plan n = 7, k = 1.44 is to be applied to both specification limits.
The double specification limit analog to Table 10.3 would appear as does Table 10.5, where

* Note that this relation allows the Jacobson nomograph to be used in reverse to obtain approximate values for
the noncentral t-distribution by using n = f + 1; k = t n - 1/n ; tp = d/ f + 1 ; 1 − P( f, δ, t) = Pa.
TABLE 10.5
Calculation of Probability of Acceptance: Double Specification Limits (Known Standard Deviation σ = 2; n = 7, k = 1.44)
Upper Specification Limit Lower Specification Limit

μ zU = (U – μ)/σ pU z A = zU – k z = nz A P(R) zL = (μ − L)/σ pL z A = zL – k z = nz A P(R) p P(R) P(A)

90 5.0 .0000 3.56 9.42 .0000 0.0 .5000 –1.44 –3.81 .9999 .5000 .9999 .0001
91 4.5 .0000 3.06 8.10 .0000 0.5 .3085 –0.94 –2.49 .9936 .3085 .9936 .0064
92 4.0 .0000 2.56 6.77 .0000 1.0 .1587 –0.44 –1.16 .8770 .1587 .8770 .1230
93 3.5 .0002 2.06 5.45 .0000 1.5 .0668 0.06 0.16 .4364 .0670 .4761 .5239
Sampling by Variables for Proportion Nonconforming

94 3.0 .0013 1.56 4.13 .0000 2.0 .0228 0.56 1.48 .0694 .0241 .0694 .9306
95 2.5 .0062 1.06 2.80 .0026 2.5 .0062 1.06 2.80 .0026 .0124 .0052 .9948
96 2.0 .0228 0.56 1.48 .0694 3.0 .0013 1.56 4.13 .0000 .0241 .0694 .9306
97 1.5 .0668 0.06 0.16 .4364 3.5 .0002 2.06 5.45 .0000 .0670 .4761 .5239
98 1.0 .1587 –0.44 –1.16 .8770 4.0 .0000 2.56 6.77 .0000 .1587 .8770 .1230
99 0.5 .3085 –0.94 –2.49 .9936 4.5 .0000 3.06 8.10 .0000 .3085 .9936 .0064
100 0.0 .5000 –1.44 –3.81 .9999 5.0 .0000 3.56 9.42 .0000 .5000 .9999 .0001
241
242 Acceptance Sampling in Quality Control

1.0

0.9

0.8

0.7

0.6

0.5
Pa

0.4

0.3

0.2

0.1

0
90 91 92 93 94 95 96 97 98 99 100 µ

0.50 0.31 0.16 0.07 0.02 0.01 0.02 0.07 0.16 0.31 0.50
p

FIGURE 10.11
OC curve: double specification limits, σ known.

pU and pL simply represent the proportion nonconforming outside U and L, respectively.

The OC curve would appear as Figure 10.11.
When the standard deviation is unknown, complications in deriving the OC curve
arise. The curve becomes a narrowband of possible values for probability of accep-
tance. A special procedure has been developed, however, utilizing a minimum variance
unbiased estimate of the proportion nonconforming in the lot. Called the M method,
it is the only procedure recommended by MIL-STD-414 for use with double specifica-
tion  limits. The procedure and the corresponding OCs will be discussed later in this
chapter.

Measures of Performance
In addition to the probability of acceptance, there are other measures of performance of
variables plans for proportion nonconforming, such as average outgoing quality (AOQ)
and average total inspection (ATI), since these measures are functions based on the OC
curve, the formulas for AOQ and ATI remain the same as in attributes inspection as shown
in Table 5.1:

AOQ  pPa
Sampling by Variables for Proportion Nonconforming 243

and

ATI = nPa + N ( 1 – Pa )

AOQL must, however, be evaluated from the AOQ curve. A crude approach to find the
AOQL of a variables plan would be to use that of a matching attributes single-sampling
plan. The match must necessarily be very good in the region of the AOQL. Because of the
difference in the inherent shape of the OC curves of variables and attributes plans, how-
ever, such an approach would have to be regarded as only a very rough approximation.
As an example, consider the known standard deviation plan n = 7, k = 1.44. For p = .18,
we have Pa = .08. Hence, for lots of size N = 120

AOQ @ pPa = .18 ( .08 ) = .014

while

ATI = nPa + N ( 1 - Pa )
= 7 ( .08 ) + 120 ( .92 ) = 110.96

Furthermore, since the plan is matched to the attributes plan n = 20, c = 1, a crude measure
of the AOQL would be that of the attributes plan PM = .036. Actually, calculation of AOQ
over the range of Table 10.4 would indicate that this is not far from the actual value.

M Method
Occasionally, it is desirable to base lot acceptance on estimates of the proportion
­nonconforming in the lot. This provides those administering the inspection with ancillary
information, which is meaningful to those not familiar with statistical methods. An esti-
mate of this sort can be made under attributes inspection simply by dividing the number
of nonconformances or defectives d found by the sample size n to obtain an unbiased esti-
mate of the proportion nonconforming in the lot. This estimate would then be compared
to the constant c/n to determine lot acceptance. When variables procedures are employed,
more sophisticated methods of estimation must be used. Such a procedure has been devel-
oped by Lieberman and Resnikoff (1955), which involves the use of a uniform minimum
variance unbiased estimate of the lot proportion nonconforming p in the acceptance sam-
pling criteria. Using X and s (or σ), a standard normal deviate Q is obtained that is then
adjusted and employed to estimate p. A comparison of the equivalent k and M methods for
variables and attributes is shown in Figure 10.12.
The procedure for application of the M method when the standard deviation is known
is given in Table 10.6.
When the standard deviation is known, M may be found as
¥
1 - t 2 /2
M=
ò(
n/ n -1)
2p
e dt
k

244 Acceptance Sampling in Quality Control

Attributes Variables

U–X
k Method d≤c z= ≥k
σ
d c U–X
M Method p= ≤ =M Q= p≤M
n n σ

FIGURE 10.12
Equivalent criteria for acceptance (sample of n). (From Schilling, E.G., Qual. Progr., 7(5), 19, 1974b. With permission.)

TABLE 10.6
M Method Standard Deviation Known (Given n, M)
Lower Specification Limit Upper Specification Limit

1. Select a random sample of size n 1. Select a random sample of size n

2. Compute X 2. Compute X
X-L n U-X n
3. Compute QL = and obtain 3. Compute QU = and obtain
s n-1 s n-1
¥ t2 ¥ t2
1 -2 1 -2
p̂L =
QL ò2p
e dt from tables. This gives a minimum p̂U =
QU ò
2p
e dt from tables. This gives a

variance unbiased estimate of p. minimum variance unbiased estimate of p.

4. If p̂L < M, accept the lot; if p̂L > M, reject the lot. 4. If p̂U < M, accept the lot; if p̂U > M, reject the lot.

so that M is simply the upper normal tail for z M = k n/( n - 1) . Note also that the estimated
proportion nonconforming is simply the upper normal tail area corresponding to QU or QL.
When the standard deviation is unknown, the noncentral t-distribution is involved in
the estimation procedure, which leads to incorporation of values from the incomplete beta
function in the estimate of p. The incomplete beta function is defined as

x ì0 £ x £ 1
G (a + b) ï
I x ( a, b ) =
ò
v ( 1 - v ) dv í a > 0
a -1 b -1

G ( a) G (b) ïb > 0
0
î

and may be evaluated using the computer or from special tables, such as those by Pearson
(1968) or, in special cases, by its relation to the binomial distribution:

I x ( a, b ) = 1 - Fbin ( a - 1|p = x , n = a + b - 1)

or

Fbin ( y|p, n ) = I x = p ( a = y + 1, b = n - y )

A page from the Pearson (1968) tables for the cumulative function Ix(p, q) is shown in
Figure 10.13. For example, we have from the table

I.31 ( 14, 14 ) = .0191640

Sampling by Variables for Proportion Nonconforming 245

x = 0.10–0.70 q = 14 p = 14–19

p= 14 15 16 17 18 19
B(p, q) = 0.3561 0481 × 1/106 0.1780 5241 × 1/106 0.9209 6072 × 1/105 0.4911 7905 × 1/105 0.2693 5625 × 1/105 0.1515 1289 × 1/105
x
0.10 .0000 001
0.11 .0000 002
0.12 .0000 006 .0000 001
0.13 .0000 015 .0000 004 .0000 001
0.14 .0000 036 .0000 009 .0000 002 .0000 001
0.15 .0000 083 .0000 023 .0000 006 .0000 002
0.16 .0000 179 .0000 053 .0000 015 .0000 004 .0000 001
0.17 .0000 362 .0000 114 .0000 035 .0000 010 .0000 003 .0000 001
0.18 .0000 699 .0000 232 .0000 075 .0000 023 .0000 007 .0000 002
0.19 .0001 239 .0000 451 .0000 153 .0000 051 .0000 016 .0000 005
0.20 .0002 285 .0000 840 .0000 300 .0000 105 .0000 036 .0000 012
0.21 .0003 905 .0001 505 .0000 565 .0000 207 .0000 074 .0000 026
0.22 .0006 454 .0002 603 .0001 022 .0000 391 .0000 146 .0000 054
0.23 .0010 346 .0004 357 .0001 786 .0000 714 .0000 279 .0000 107
0.24 .0016 129 .0007 079 .0003 024 .0001 261 .0000 514 .0000 205
0.25 .0024 500 .0011 186 .0004 972 .0002 157 .0000 915 .0000 380
0.26 .0036 333 .0017 227 .0007 954 .0003 585 .0001 580 .0000 682
0.27 .0052 692 .0025 906 .0012 406 .0005 799 .0002 651 .0001 188
0.28 .0074 840 .0038 098 .0018 895 .0009 149 .0004 333 .0002 011
0.29 .0104 244 .0054 872 .0028 145 .0014 097 .0006 907 .0002 317
0.30 .0142 565 .0077 498 .0041 060 .0021 247 .0010 758 .0005 338
0.31 .0191 640 .0107 453 .0058 736 .0031 364 .0016 390 .0008 395
0.32 .0253 448 .0146 415 .0082 480 .0045 398 .0024 458 .0012 918
0.33 .0330 071 .0196 246 .0113 810 .0064 503 .0035 789 .0019 470
0.34 .0423 632 .0258 962 .0154 452 .0090 047 .0051 404 .0028 777
0.35 .0536 230 .0336 688 .0206 321 .0123 619 .0072 539 .0041 748
0.36 .0669 863 .0431 604 .0271 494 .0167 023 .0100 652 .0059 503
0.37 .0826 346 .0545 876 .0352 165 .0222 258 .0137 436 .0083 385
0.38 .1007 226 .0681 578 .0450 585 .0291 489 .0184 800 .0114 979
0.39 .1213 695 .0840 603 .0568 991 .0376 996 .0244 858 .0156 106
0.40 .1446 518 .1024 577 .0709 528 .0481 117 .0319 886 .0208 816
0.41 .1705 958 .1234 768 .0874 151 .0606 167 .0412 272 .0275 361
0.42 .1991 730 .1472 002 .1064 535 .0754 351 .0524 450 .0358 154
0.43 .2302 954 .1736 584 .1281 977 .0927 668 .0658 810 .0459 706
0.44 .2638 151 .2028 244 .1527 306 .1127 809 .0817 611 .0582 549
0.45 .2995 240 .2346 087 .1800 799 .1356 048 .1002 865 .0729 146
0.46 .3371 573 .2688 580 .2102 116 .1613 152 .1216 229 .0901 777
0.47 .3763 986 .3053 548 .2430 256 .1899 290 .1458 901 .1102 430
0.48 .4168 872 .3438 207 .2783 531 .2213 963 .1731 506 .1332 674
0.49 .4582 276 .3839 219 .3159 570 .2555 957 .2034 010 .1593 544
0.50 .5000 000 .4252 770 .3555 356 .2923 324 .2365 648 .1885 428
0.51 .5417 724 .4674 668 .3967 279 .3313 385 .2724 881 .2207 979
0.52 .5831 128 .5100 463 .4391 231 .3722 780 .3109 378 .2560 042
0.53 .6236 014 .5525 576 .4822 716 .4147 531 .3516 034 .2939 618
0.54 .6628 427 .5945 434 .5256 976 .4583 149 .3941 031 .3343 861
0.55 .7004 760 .6355 607 .5689 143 .5024 763 .4379 922 .3769 115
0.56 .7361 849 .6751 941 .6114 385 .5467 265 .4827 758 .4210 989
0.57 .7697 046 .7130 675 .6528 057 .5905 478 .5279 236 .4664 475
0.58 .8008 270 .7488 543 .6925 850 .6334 320 .5728 872 .5124 096
0.59 .8294 042 .7822 851 .7303 914 .6748 975 .6171 186 .5584 088
0.60 .8553 482 .8131 542 .7658 968 .7145 044 .6600 889 .6038 596
0.61 .8786 305 .8413 213 .7988 385 .7518 685 .7013 066 .6481 886
0.62 .8992 774 .8667 127 .8290 244 .7866 721 .7403 338 .6908 548
0.63 .9173 654 .8893 184 .8563 352 .8186 725 .7768 004 .7313 691
0.64 .9330 137 .9091 878 .8807 239 .8477 057 .8104 145 .7693 114
0.65 .9463 770 .9264 229 .9022 118 .8736 881 .8409 698 .8043 435
0.66 .9576 368 .9411 698 .9208 825 .8966 138 .8693 479 .8362 190
0.67 .9669 929 .9536 104 .9368 734 .9165 484 .8925 170 .8647 875
0.68 .9746 552 .9639 519 .9503 658 .9336 209 .9135 271 .8899 950
0.69 .9808 360 .9724 173 .9615 740 .9480 132 .9315 008 .9118 786
0.70 .9857 435 .9792 367 .9707 346 .9599 475 .9466 222 .9305 579

FIGURE 10.13
Tables of the incomplete β function. (Reproduced from Pearson, E.S., Tables of the Incomplete Beta-Function, 2nd ed.,
Cambridge University Press, London, U.K., 1968, p. 296. With permission.)
246 Acceptance Sampling in Quality Control

When the standard deviation is unknown, the basic procedure for the M method is the
same as that shown for σ known in Table 10.6 using s in the denominator of Q so that

U-X X-L
QU = QL =
s s

The estimation procedure then becomes

p̂U = I X ( a, b )

where

ïì 1 1 n ïü
x = max í0, - QU ý
ïî 2 2 n - 1 ïþ
n
a = b = -1
2

and

p̂L = I X ( a,b )

where

ïì 1 1 n ïü
x = max í0, - QL ý
îï 2 2 n - 1 þï
n
a = b = -1
2

For single specification limits, the M and k methods can be shown to be equivalent for a
given sample size with

k=
( n - 1) (1 - 2BM )
n
1æ n ö
BM = çç 1 - k ÷
2è n - 1 ÷ø

where BM is defined such that

æ n-2 n-2ö M
I BM ç , ÷=
è 2 2 ø 100

for M expressed in percent.

The M method is unique for double specification limits since the sum of the estimates of
p above and below specification limits give the total estimated proportion nonconforming,
that is

pˆ L + pˆ U = pˆ
Sampling by Variables for Proportion Nonconforming 247

This total estimate p̂ can be compared to M to determine acceptance of the lot. Under
the uniform minimum variance unbiased estimation technique, borders for an accep-
tance polygon can be found by finding a k equivalent to M and using the formulas for the
k method. The acceptance region itself will be found to be slightly different than that given
( )
by Wallis and is defined by the points X¢, s¢ , resulting from the simultaneous solution of

X=
( ) (
U 1 - 2Bp ¢ - L 2Bp²0 - 1
0
)
(
2 1 - Bp ¢ - Bp²0
0
)
U -L æ n ö
s= çç ÷÷
(
2 1 - Bp ¢ - Bp²0
0
) è n -1ø

where

M = p ** = p¢0 + p0²

When the k method is to be used with double specification limits without the benefit of
a polygon, a more refined estimate of the MSD may be obtained using the method of
Lieberman and Resnikoff (1955). We have, for a given value of M

n
MSD = (U - L )
2 ( 1 - 2BM/2 ) ( n - 1)

To obtain BM/2, a value of x from the incomplete β distribution must be found such that

æ n-2 n-2ö M æ 1 ö
IX ç , ÷= ç ÷
è 2 2 ø 2 è 100 ø

Then BM/2 = x and the previous formula for MSD can be evaluated. This is the method used
in MIL-STD-414.
As an example to show the relation of the k and M methods, consider the plan n = 30,
k = 2.00, which is listed in MIL-STD-414 as the Code J, 0.65 AQL normal plan. Using the
relation given in the earlier text, M is such that

1æ n ö
BM = çç 1 - k ÷
2è n - 1 ÷ø
1æ 30 ö
= çç 1 - ( 2.00 ) ÷ = .311
2è 29 ÷ø

It is then necessary to evaluate the incomplete β distribution to obtain the value of M since

æ n-2 n-2ö M
I BM ç , ÷=
è 2 2 ø 100
I.311 ( 14, 14 ) = .0197
248 Acceptance Sampling in Quality Control

by interpolation from Figure 10.13. Using the binomial relation

I.311 ( 14,14 ) = 1 - Fbin ( 13|p = .311, n = 27 )

= 1 - .9803 = .0197

by linear interpolation in the Department of Commerce (1950) binomial tables. So

M
= .0197 M = 1.97
100

MIL-STD-414 shows M = 1.98. The difference is due to rounding.

Furthermore, the MSD associated with this plan is

n
MSD = (U - L )
2 ( 1 - 2BM/2 ) ( n - 1)

For M = 1.98, it is necessary to find IX(14,14). By linear interpolation from Figure 10.13, or
using the binomial relation,

I.288 ( 14, 14 ) @ .0099

Hence

30
MSD = (U - L ) = .223 (U - L )
2 ( 1 - 2 ( .288 ) ) ( 29 )

and MIL-STD-414 gives this value for the MSD of this plan.

Plans Based on Sample Range

In practice, it is often desirable to substitute the sample range for the sample standard
deviation in applying variables plans for proportion nonconforming or process param-
eter. The range has several desirable properties and is used extensively in quality control.
Among them are

1. Commonly understood
2. Quick to compute
3. Easy to explain
4. Easy to verify
5. Easy and inexpensive to compute

The chief disadvantage of the range is the loss of efficiency resulting from its use. This
can be compensated by increasing the sample size. Use of the average range of m random
Sampling by Variables for Proportion Nonconforming 249

subgroups of size nR taken from the original sample also improves the efficiency s­ omewhat.
The d2* factor developed by Duncan (1955) can be used to estimate the standard deviation as

R
sˆ =
d2*

This estimate has the same bias as s. Accordingly, the following procedure can be used
if the average range is to be substituted for s in a given variables sampling plan, which
requires a sample size of ns. Normality of the underlying observations are assumed:

1. Select the subgroup size to be used (subgroups of size 5 are often used, although 8
is considered to be optimum).
2. Determine the equivalent sample size for the range plan. If the original plan using
s had sample size ns, the number of subgroups m of size nR in the range plan will
be approximately
ns - 1
m@
.9 ( nR - 1)

This value of m is conservative when rounded upward.

3. Substitute R/d2* for s in the original sampling plan, where

æ 0.2778 ö
d2* @ d2 ç 1 + ÷
ç m ( nR - 1) ÷ø
è
and d2 is the standard control chart factor for subgroups of size nR.
4. Use the original sampling plan as given with the decision criteria, using the statis-
tic and sample size as modified earlier.

This approximation to Duncan’s d2* factor from Schilling (1973) has been found to be quite
sufficient for practical purposes. The formula for m uses Ott’s (1955) approximation

n R = .9m ( nR - 1)

for degrees of freedom of the range, νR. More accurate values for d2* and the degrees of free-
dom associated with the average range estimate of s as given by Nelson (1975) will be found
in Appendix Table T10-1. The values of the constant difference (c.d.) found at the bottom of
Nelson’s table can be used to determine degrees of freedom for numbers of samples not listed as

n¢ = n + ( c.d.) ( k¢ - k )

where degrees of freedom ν′ are required for k′ samples of n, but only ν for k samples of n
is listed. Thus, using k = 20, the degrees of freedom for 25 samples of 5 is

n¢ = 72.7 + 3.62 ( 25 - 20 ) = 90.8 ~ 91.

Ott’s approximation gives

n R = .9 ( 25 )( 5 - 1) = 90
250 Acceptance Sampling in Quality Control

Example
Consider the following example taken from MIL-STD-414.
The specifications for electrical resistance of a certain electrical component is 650.0 ±
30 Ω. A lot of 100 items is submitted for inspection … with AQL = 2.5% …. Suppose the
values of sample resistance in the order reading from left to right are as follows:

643, 651, 619, 627, 658, 670, 673, 641, 638, 650

For these data,

X = 647

( )
2
S X-X
s= = 17.22
n -1

Let us assume that standard deviation was unknown and it was desired to institute a plan
based on average range. The first step in determining an average range plan would be to
develop the appropriate plan for variability unknown using the sample standard devia-
tion. For this example, take p2 = .215. Then, for the standard deviation plan

p1 = .025, a = .05
p2 = .215, b = .10

so

z p2 za + z p1 zb 0.79 ( 1.645 ) + 1.96 ( 1.282 )

k= = = 1.30
za + zb 1.645 + 1.282
2 2
æ z +z ö æ k 2 ö æ 1.645 + 1.282 ö æ 1.30 2 ö
ns = çç a b ÷÷ ç1+ ÷ = ç ÷ ç1+ ÷ = 11.5 ~ 12
è z p1 - z p2 ø è 2 ø è 1.96 - 0.79 ø è 2 ø

And to find MSD, follow these steps:

p** corresponding to z*p = k = 1.30 is p** = .0968.

a.
z**
b. p corresponding to p**/2 = .0484 is z **
p = 1.66.
c. So MSD  (U - L )/2z**
p = ( 680 - 620 )/( 2 ( 1.66 ) ) = 18.07 .

In applying this plan, for s = 17.22 and X = 647 as indicated in the example consider the
following:

s = 17.22 ≤ MSD = 18.07.

1.
( )
TU = U - X /s = ( 680 - 647 )/17.22 = 1.92 > k = 1.30.
2.
( )
TL = X - L /s = ( 647 - 620 )/17.22 = 1.57 > k = 1.30.
3.
Sampling by Variables for Proportion Nonconforming 251

And since all three acceptance criteria are met, the lot would be accepted. To convert the
previously mentioned plan for use of the average range, follow these steps:

1. Use subgroup size nR = 5 (arbitrarily)

2. Number of subgroups m:
ns - 1 10 - 1 9
m= = = = 2.5 ~ 3
.9 ( nR - 1) .9 ( 5 - 1) 3.6

3. Substitute R/d2* for s in the previously mentioned acceptance criteria, where

æ .2778 ö æ .2778 ö
d2* @ d2 ç 1 + ÷÷ = 2.326 çç 1 + ÷ = 2.38
ç m ( nR - 1) ø 3 ( 5 - 1) ÷ø
è è

Note Appendix Table T10-1 gives d2* = 2.38 with 11.1 degrees of freedom in contrast
to the 9 degrees of freedom that would have been obtained under the standard
deviation plan with five fewer observations.

The average range plan then requires a sample of three r­ andom subgroups of size 5 each.
Assume the following values are obtained:

643 670 651

651 673 627
619 641 670
627 638 641
658 650 650

Then
X = 647.27
39 + 35 + 43
R= = 39
3
so the acceptance criteria become

R
1. £ MSD
d2*
R £ d2* MSD = MAR
39 £ 2.38 ( 18.07 ) = 43.0

where MAR is the maximum allowable range that serves the same purpose as
the MSD.
U - X 680 - 647.27
TU =
2. = = 2.00 > k = 1.30
R/d2* 39/2.38
X - L 647.27 - 620
TL =
3. = = 1.66 > k = 1.30
R/d2* 39/2.38
and the lot is accepted.
252 Acceptance Sampling in Quality Control

Many production situations demand the simplicity and utility of the use of the range.
Even hand calculators that can calculate s directly will not supplant the intuitive under-
standing and familiarity, which operators and inspectors have for the range as a measure
of spread. It is important that users of acceptance sampling techniques have an under-
standing of the basic approach. The range can contribute much in this regard.

Double Sampling by Variables

As in the case of attributes sampling plans, savings in average sample size can be achieved
by using double-sampling plans. Indications are that a reduction in ASN to about 80% of
a single-sampling unknown standard deviation plan is possible for single sample sizes
greater than 25. Such plans are presented by Bowker and Goode (1952) and follow the
same procedure as in attributes double sampling with variables acceptance criteria in place
of the familiar first and second sample acceptance and rejection numbers.
A double-sampling variables plan is specified for the case of standard deviation
unknown by

n1 as the first sample size

n2 as the second sample size
ka as the first sample acceptance constant
kr as the first sample rejection constant
kt as the second sample acceptance constant

Primes are added to the acceptance constant to give k a¢ , k r¢ , kt¢ for use when the standard
deviation is known.
To apply a double-sampling plan to a single specification limit, perform the following:

1. Draw the first sample of n1 and calculate X1 and s1 from the data.
2. Compute

U - X1
TU1 =
s1

for an upper specification limit and

X1 - L
TL1 =
s1

for a lower specification limit

3. Test the first sample

TU1 ³ k a or TL1 ³ k a accept

TU1 £ k r or TL1 £ k r reject
k r < TU1 < k a or k r < TL1 < k a resample
Sampling by Variables for Proportion Nonconforming 253

4. Draw the second sample of n2 from the lot. Combine the data with that from the
first sample to obtain the total sample. Calculate

å X1 + å X 2
Xt =
n1 + n2
( n1 + n2 ) ( å X12 + å X22 ) - ( å X1 + å X2 )
2

st =
( n1 + n2 ) ( n1 + n2 - 1)

5. Compute

U - Xt
TUt =
st

for an upper specification limit and

Xt - L
TLt =
st

for a lower specification limit

6. Test combined total sample

TUt ³ kt or TLt ³ kt accept

TUt < kt or TLt < kt reject

When the standard deviation is known, the procedure is the same with s1 and st replaced
by σ. In the case of double specification limits, test against both specification limits follow-
ing the previously mentioned step rejecting if indicated by any of the tests. In addition,
calculate MSD values from kr and kt, respectively, using the method of Wallis to obtain
MSD1 and MSDt. Reject if s1 > MSD1 or st > MSDt.
Bowker and Goode (1952) have tabulated double-sampling plans for standard deviation
known and unknown. They also give information on the OCs and AOQL of the plans.
Note that in their tabulation, Bowker and Goode (1952) define AQL to be the 95th percen-
tile of the OC curve.
The following is an example of application given by Bowker and Goode (1952) in pre-
senting their double-sampling plans. A manufacturer purchases stud bolts that are to have
a minimum tensile strength of 125,000 lb/in.2. The plan

n1 = 8 n2 = 16

k a = 3.041 k r = 1.344 kt = 2.245

is to be used with standard deviation unknown. The measurements for the first and second
samples are shown in Table 10.7. Applying the plan, consider the following:

X1 = 129, 062 s1 = 1, 522.

1.
129, 062 - 125, 000
2. TL1 = = 2.67 .
1, 522
254 Acceptance Sampling in Quality Control

TABLE 10.7
Tensile Strength of Stud Bolts
Item Ultimate Strength (lb/in. 2) Item Ultimate Strength (lb/in. 2)

First sample Second sample

1 129,500 9 129,500
2 131,000 10 131,000
3 128,500 11 129,500
4 126,500 12 128,000
5 130,000 13 129,000
6 130,500 14 127,000
7 127,500 15 132,500
8 129,000 16 130,500
17 129,000
18 130,000
19 133,000
20 129,000
21 131,500
22 130,000
23 132,000
24 128,500

Source: Reproduced from Bowker, A.H. and Goode, H.P., Sampling Inspection by Variables,
McGraw-Hill, New York, 1952, 95. With permission.

3. 2.67 < ka = 3.041 cannot accept.

2.67 > kr = 1.344 cannot reject.

So a second sample is needed

Xt = 129, 688 st = 1, 647.

4.
129, 688 - 125, 000
5. TLt = = 2.85.
1, 647
6. 2.85 > kt = 2.245 accept.

Sommers (1981) has obtained two-point double-sampling variables plans that provide
minimum ASN when the proportion nonconforming is at the PQL. The plans are given in
Appendix Table T10-2 and cover the values of p1 and p2 for α = .05 and β = .10, which were
tabulated by the Statistical Research Group (1947). The plans presented are for n1 = n2 and
kt = kr; hence, only n1, ka, and kr are shown. Sample sizes are given as nσ and ns for known and
unknown standard deviation, respectively. ASNs are represented in a similar manner. For
known standard deviation plans, k¢a = k a and kt¢ = k¢r = k r . Given these constraints, Sommers
used an adaptation of the Wallis approximation together with an iterative procedure to
minimize ASN at the specified p1. Appendix Table T10-2 also presents a set of matched
single-sampling plans for each p1 and p2. For example, when p1 = .01, p2 = .05, α = .05, β = .10

Known standard deviation: n1 = 13, n2 = 13, k¢a = 2.09, k¢r = kt¢ = 1.87

Unknown standard deviation: n1 = 36, n2 = 36, k a = 2.09, k r = kt = 1.87

Sampling by Variables for Proportion Nonconforming 255

The ASNs for these plans are 14.9 and 41.5 for known and unknown standard deviation,
respectively. The matched single-sampling plans have sample sizes 19 and 54. This is
indicative of the type of average sample size reduction possible through double-sampling
variables plans.

Tolerance Intervals and Variables Plans for Percent Nonconforming

The use of tolerance intervals in attributes sampling was discussed earlier. Tolerance inter-
vals constructed from variables data can be used in a similar way as a medium for lot
acceptance and reliability assessment.
The form in which reliability specifications are written, that is, requirements of π reli-
ability with γ confidence, fosters the use of tolerance intervals in this regard. This is natural
since a tolerance interval guarantees that at least a stated proportion π of the population is
contained within the limits of the interval with λ confidence. Therefore, a tolerance inter-
val, constructed from the data, which is entirely within the specification limits shows that
the requirements have been met with the corresponding confidence. A tolerance interval
that overlaps the specification limits is evidence that the requirements have not been met,
provided

a. The π and γ used are reasonably exact (not gross inequalities)

b. Sample size has been chosen to take the producer’s risk (PR) into account

Unless these conditions are checked, judgment should be withheld as to whether the lot
should be rejected.
As shown earlier, specifications in terms of reliability can be converted into the usual
quality control notation through the relations

p = 1 - p g = 1 - Pa

Thus, the tolerance interval approach has found application in acceptance sampling as
well as reliability.
We shall be concerned here with tolerance intervals on measurements. An underlying
normal distribution is assumed. It should be pointed out that when all population param-
eters are known, a tolerance interval having γ = 1 can be obtained from the normal dis-
tribution itself. For example, a resistor having μ = 10 Ω and σ = 1 Ω will have 95% of the
population within

m ± 1.96 s

or

8.04 to 11.96 W

This tolerance interval is constructed with 100% confidence, so that if the specification
limits are 8.0–12.0 Ω and a reliability of 95% is required, the product should be accepted.
256 Acceptance Sampling in Quality Control

This is true even if less confidence, say, γ = 90%, was originally specified. Specification of a
confidence value always means at least the stated amount for lot acceptance.
When population parameters are unknown, 100% confidence can rarely be achieved
short of 100% inspection. Even then we cannot often be 100% confident of the inspection
procedure. Estimates must be substituted for population parameters, a set of confidence
levels, and more sophisticated procedures, often based on the noncentral t-distribution,
employed. When parameters are unknown, a typical one-sided variables tolerance interval
is of the form

X + ks

for an upper tolerance limit, or

X - ks

for a lower tolerance limit. A two-sided interval may be expressed as

X ± ks

It will be recalled that the acceptance criteria for the k method in the variables procedure
were precisely

X + ks < U

or

X - ks > L

for one-sided plans, with corresponding criteria for the two-sided case.
An extensive set of tables of tolerance limit factors and associated criteria has been pub-
lished by Odeh and Owen (1980). The resulting tabulation is useful in acceptance sampling
and reliability applications. The contents of the tables are described in Table 10.8.
Odeh–Owen Tables 1, 3, and 7 are primarily intended for tolerance and confidence inter-
val estimation. Their Table 7 provides confidence limits for the tail areas of the normal
distribution using the procedure of Owen and Hua (1977). The Odeh–Owen Tables 8, 9,
and 10 are useful in implementing a screening strategy in which the proportion of variable
Y above a lower specification limit L is improved by screening on a related variable X. By
selecting out a proportion β of the population in which X > μx − zβσx, the proportion of Y
above L is raised from γ to δ. Of course, the effectiveness of the procedure depends on the
strength of the correlation ρ between X and Y. However, it presents a useful alternative for
screening when tests are destructive as, for example, in life tests. This type of screening
strategy is described in Owen et al. (1975).
Tables 2, 5, and 6 are intended to be used directly in acceptance sampling. Table 2 pres-
ents one-sided k factors and sample sizes for specified α, β, p1, and p2. Table 5, reproduced
here as Appendix Table T10.4, shows two-sided equal-tailed k values for specified P = p2
and β = .10. To use the table with specified PR (say at α = .05), it is necessary to use the Wallis
formula to determine the sample size and then improve the approximation by selecting k
from the Odeh–Owen table.
Sampling by Variables for Proportion Nonconforming 257

TABLE 10.8
Content of Odeh–Owen Tables of Tolerance Limits
Tables Content Application

1 Factors for one-sided tolerance limits One-sided variables plans with one risk specified
(k by γ, n, P = π) One-sided tolerance intervals for reliability estimation
2 Sample size for one-sided sampling plans One-sided variables plans with both risks specified
(n, k by α, β, p1, p2)
3 Two-sided (central) tolerance limits to Two-sided tolerance intervals for estimation with equal
control both tails equally (k by γ, n, P = π) tails
4 Two-sided (noncentral) tolerance limits to Two-sided tolerance intervals for estimation with
control tails separately (k by γ, n, P = π) nonequal tails
5 Two-sided sampling plan factors to Two-sided equal tails variables plans with β = .10 (only)
control equal tails (k by γ = .90, n, p) specified
May be used with two risks by approximating n with
Wallis formula (use α/2). Then find k from Table 5.
6 Two-sided sampling plan factors to Two-sided unequal tails variables plans with β = .10
control tails separately (k by γ = .90, n, p) (only) specified
May be used with two risks by approximating n with
Wallis formula (use α/2). Then find kL and kU for lower
and upper limits separately from Table 6.
7 Confidence limits for proportion in tail of Lower confidence limit on proportion of product above
normal distribution lower specification limit. Shows proportion above L
tabulated by confidence = η, n, and K = (X − L)/s
8 Screening proportion-population Proportion measurement Y above L is γ. Y to be
parameters known (β by δ, γ, ρ) screened on concomitant variable X. μx, μy, σx, σy, ρ
known. Proportion Y above L may be raised from γ to δ
by selecting all X above μx – zβσx where β is proportion
of population which all will be selected
9 Screening proportion-population Same as Table 8 for γ, δ, ρ known and X, sx used as estimates
parameters unknown (β by f, δ, γ, ρ) from preliminary sample with f degrees of freedom
10 Confidence limits on the correlation Confidence limits for r. Shows upper, lower, and two-sided
coefficient confidence limits given n, sample R = r, and risk α
Used with Table 9 and ρ is unknown
Source: Jacobson, L.J., Ind. Qual. Control, 63, 23, 1949. With permission.

For example, suppose a plan is desired having p1 = .005, p2 = .10, α = .05, and β = .10. It
is to be used to check the length of Kovar leads, which are specified to be 9 cm ± .05 mm.
Using the Wallis formula, we get the following:

za/2 z2 + zb z1
k=
za/2 + zb
1.96 ( 1.28 ) + 1.28 ( 2.58 )
= = 1.79
1.96 + 1.28
2
æz +z ö æ k2 ö
n = ç a/2 b ÷ ç 1 + ÷
è z1 - z2 ø è 2 ø
2
æ 1.96 + 1.28 ö æ 1.792 ö
=ç ÷ ç1+ ÷
è 2.58 - 1.28 ø è 2 ø
= 16.2 ~ 17
258 Acceptance Sampling in Quality Control

The Odeh–Owen Table 5 shows that, for p2 = .10 and sample size 17, a more accurate value
of k would be 1.95. This will hold the CQL of .10. The sampling plan is as follows:

Sample size: n = 17.

Accept if: 8.95 £ X - 1.95s and X + 1.95s £ 9.05.
Reject if: X - 1.95s < 8.95 or 9.05 < X + 1.95s.

Sometimes the CQL is specified to be different for the lower and upper tails. In this case,
the Odeh–Owen Table 6 reproduced here as Appendix Table T10.5 is used for specified pL
and pU in the lower and upper tails, respectively; here also β = .10. More extensive values
are given in Odeh–Owen Table 1. Again the Wallis formula is used to obtain the approxi-
mate sample size associated with a specified producer’s risk α.
Suppose longer leads could be tolerated by the process better than shorter leads in the
previous example so that the CQL was broken into two parts, pL = .025 against the lower
specification and pU = .05 against the upper specification, still with the PQL, p1 = .005
with risks α = .05 and β = .10. Using the Wallis formula with zL and zU corresponding to
pL and pU, k and n are calculated for the lower and upper specification limits as shown in
Table 10.9.
Since only one sample size can be taken, it will be necessary to take a sample size
of approximately 94. This might be rounded to 100 for administrative purposes. The
Odeh–Owen Table 6 shows for n = 100; kL = 2.203, and kU = 1.861. The sampling plan is
as follows:

Sample size: n = 100.

Accept if: 8.95 £ X - 2.20s and X + 1.86s £ 9.05.
Reject if: X - 2.20s < 8.95 or 9.05 < X + 1.86s .

This procedure may be used as a substitute for the M method and is somewhat simpler
for inspectors to understand and use. Of course, the Odeh–Owen Table 2 can be used for
one-sided specification limits.

TABLE 10.9
Calculation of k and n for Lower and Upper Specification Limits
Lower Specification Limit Upper Specification Limit

za/2 zL + zb z1 za/2 zU + zb z1
k= k=
za/2 + zb za/2 + zb
1.96 ( 1.96 ) + 1.28 ( 2.58 ) 1.96 ( 1.64 ) + 1.28 ( 2.58 )
= =
1.96 + 1.28 1.96 + 1.28
= 2.20 = 2.01
2 2
æ k2 öæ z + z ö æ k2 öæ z + z ö
n = ç 1 + ÷ ç a/2 b ÷ n = ç 1 + ÷ ç a/2 b ÷
è 2 ø è z1 - zL ø è 2 ø è z1 - zU ø
2 2
æ 2.22 ö æ 1.96 + 1.28 ö æ 2.012 ö æ 1.96 + 1.28 ö
= ç1 + ÷ç ÷ = ç1 + ÷ç ÷
è 2 ø è 2.58 - 1.96 ø è 2 ø è 2.558 - 1.64 ø
= 93.4 = 35.9
Sampling by Variables for Proportion Nonconforming 259

The theory of the use of tolerance limits in sampling inspection has been described in
detail by Owen (1964, 1967), Owen and Frawley (1971), and Owen et al. (1972). Earlier
tables of tolerance factors in the quality control literature include Lieberman (1958) and
Zobel (1958). Tables of tolerance limits based on the range have been given by Bingham
(1962) and Owen et al. (1971). Nelson (1977) has discussed tolerance intervals in which the
mean and standard deviation are estimated by separate samples.

Sequential Plans for Proportion Nonconforming

Availability and ease of calculation have made sequential sampling plans a natural choice
when expensive or difficult inspection is to be performed. An operator may be alerted by
various signaling devices and displays of the sequential plan called up as needed. Sample
sizes can be substantially reduced when variables sequential plans for proportion noncon-
forming are utilized.
When the standard deviation is known, a simplified relationship between sequential
plans for the mean and single-sampling variables plans, given by Sommers (1979, personal
communication with the author), can be used to easily obtain sequential variables plans for
proportion nonconforming with α = .05 and β = .10 for specified p1 and p2. Once a known
standard deviation variables plan for proportion nonconforming has been obtained to
these specifications, its parameters ns, k can be converted to those of a sequential plan for
an upper specification limit by using the relations*:

h1 = .7693s n s
h2 = .9877 s n s
s
s = U - k s - .1818
n

with

ASN ( p1 ) = .4657 ns ASN ( p2 ) = .5549ns

Thus, the sequential plan matching the known standard deviation plan nσ = 7, k = 1.44 has
parameters

h1 = 2.04s h2 = 2.61s s = U - 1.51s

with

ASN ( p1 = .018 ) = 3.26 ASN ( p2 = .180 ) = 3.88

( )
* The formula for s is obtained from that given by Sommers (1979) by using the relation m1 = U - z1s = U - k + za/ n s.
260 Acceptance Sampling in Quality Control

This approach may be applied to a lower satisfaction limit using the procedures outlined
in Chapter 8, where s can be calculated from the lower specification limit as

s
s = L + k s + .1818
n

Further Considerations
Derivation of n, k Formulas
When the standard deviation is known, the formulas that give n and k as a function
of fixed p1, p2, α, and β are easily derived. Figure 10.14 will be used as motivation for
the algebra involved. It supposes an upper specification limit U and is based on the
X method.

A U

µ1 α p1

x
µ1 + z1–α σ X µ1 + z1 σ

β µ2 p2

x
µ2 + zβ σ X µ2 + z2 σ

FIGURE 10.14
Derivation of variables plan.
Sampling by Variables for Proportion Nonconforming 261

To obtain n, if we regard the z values as representing upper tail probability points

U - A = k s

= m1 + z1s - ( m1 + zasX ) (10.1)

= m 2 + z2s - ( m 2 - zbs X ) (10.2)

Since Equations 10.1 and 10.2 equal to each other
s s
z1s - za = z2s + zb
n n
1
( za + zb ) = ( z1 - z2 )
n
2
æz +z ö
n=ç a b ÷
è z1 - z2 ø

To obtain k,

æ s ö
k s = m1 + z1s - ç m1 + za ÷
è nø
z z
k = z1 - a or k = z2 + b
n n

So carrying further

k n z1 n
= -1
za za

and

k n z2 n
= +1
zb zb

so adding

k n k n z1 n z2 n
+ = +
za zb za zb
æz +z ö z z +z z
kç a b ÷ = 1 b 2 a
è za zb ø za zb
z z +z z
k= 1 b 2 a
za + zb

The formula may be developed equally well using a lower specification limit.
The derivation of the formulas when the standard deviation is not known is more com-
plicated and is given by Wallis (1947) who developed the formulas. An excellent discussion
of approximations of the type proposed by Wallis, their application, and their efficiency in
relation to indifference quality has been given by Hamaker (1979).
262 Acceptance Sampling in Quality Control

Need for Normality

The plans discussed in this chapter are based on the assumption of normality of the
underlying measurements. While Jennett and Welch (1939) have argued that they may
give roughly correct results when applied to distributions other than normal, extreme care
must be taken in such circumstances. Certainly, when dealing with small values of propor-
tion nonconforming, that is, the tails of the distribution, the validity of the assumption of
­normality is critical.
Plans may be derived for underlying distributions other than normal. Owen (1969) has
presented an excellent paper in this regard.

Software Applications
Minitab
Minitab can create a variables single-sampling plan using Stat > Quality Tools > Acceptance
Sampling by Variables > Create/Compare. Inputs to generate the plan include the AQL,
rejectable quality level (RQL) (or LTPD), α, β, lower and/or upper specification limit,
­historical standard deviation (optional), and the lot size (if known).

Example*
A large mail order firm receives men’s dress shirts at its distribution center. A quality
requirement is that the length from the armscye (opening for the sleeve to be attached) to
the hem is the proper length. Rather than inspecting 100% of the lot of 10,000 shirts, the
manager of the distribution center wants to use an acceptance sampling plan for incoming
inspection. The vendor has contractually agreed to an AQL of 1% and an RQL of 2%, and
a β of 10%. The manager decides to use 5% for α. Thus, we have p1 = .01 or 1%, p2 = .02 or
2%, α = .05, β = .10, U = 0.5, L = −0.5, and σ = 0.1. The dialog box should be completed for
this example as shown in Figure 10.15.

Acceptance Sampling by Variables: Create/Compare

Lot quality in percent defective
Lower Specification Limit (LSL) -0.5
Upper Specification Limit (USL) 0.5
Historical Standard Deviation 0.1
Lot Size 10000
Acceptable Quality Level (AQL) 1
Producer’s Risk (α) 0.05
Rejectable Quality Level (RQL or LTPD) 2
Consumer’s Risk (β) 0.1
Generated Plan(s)

* This example is used by Minitab in their training on acceptance sampling and has been provided to the author
for preparation of this edition.
Sampling by Variables for Proportion Nonconforming 263

FIGURE 10.15
Minitab dialog box to generate a variables plan.

Sample Size 116

Critical Distance (k Value) 2.17318
Z.LSL = (mean - lower spec)/historical standard deviation
Z.USL = (upper spec - mean)/historical standard deviation
Accept lot if Z.LSL ≥ k and Z.USL ≥ k; otherwise reject.
Percent Probability Probability
Defective Accepting Rejecting AOQ ATI
1 0.950 0.050 0.939 605.3
2 0.099 0.901 0.196 9019.9
Average outgoing quality limit (AOQL) = 0.978 at 1.137 percent defective.

The generated variables plan is n = 116 and k = 2.173. Minitab also generates the OC,
AOQ, and ATI curves (Figure 10.16). Note that the Minitab output mentioned previously
indicates that the AOQL occurs at 1.137 percent defective with a value of 0.978 outgo-
ing percent defective in lots at this quality level. The ATI curve demonstrates that as
the lot percent defective approaches 3%, these lots will require 100% inspection of the
10,000 items.
Now, let’s collect sample data on n = 116 shirts randomly from all shirt sizes. In order to
evaluate all shirt sizes in the lot, we will calculate the deviations from the nominal sizes.
Open the Minitab worksheet Armscye.mpj using File > Open and load the worksheet into
Minitab. In this worksheet is a conversion table in columns 6 and 7. Once the conversion
is made, the deviation from nominal can be computed. Note that the measurements of
the shirts are read to the nearest 1/16”. We need to store our design lengths, so click in the
 264 Acceptance Sampling in Quality Control

Operating characteristic (OC) curve Average outgoing quality (AOQ) curve

1.00

AOQ (percent defective)

1.0
0.75

0.50
0.8
0.25
Probability of acceptance

0.6 0.00
0 1 2 3 4
Incoming lot percent defective
0.4 Average total inspection (ATI) curve

Average total inspection

9000
0.2 7000
5000

0.0 3000
1000
0 1 2 3 4
Lot percent defective 0 1 2 3 4
Sample size = 116, Critical distance = 2.17318 Lot percent defective

FIGURE 10.16
Minitab OC, AOQ, and ATI curves for the generated variables plan.

column name cell for column 3 and type Design Length. Next, choose Data > Recode > Use
Conversion Table and complete the dialog box as shown in Figure 10.17. This produces a
summary table of the converted values.

Recode Individual Values

Original Recoded Number
Value Value of Rows
S 18 23
M 19 15
L 20 26
XL 21 29
XXL 22 23

We must now create a column to store our deviations, so click on the column name cell for
column 4 and type Delta. Then, choose Calc > Calculator and in the Store result in vari-
able input cell enter Delta. In the Expression window, enter ‘Measured’—‘Design Length’
(Figure 10.18) and click on OK to generate the column of deviations in column 4. A portion
of the completed worksheet is shown in Figure 10.19.
With the sample data in hand, we can now use our generated variables sampling plan to
determine whether to accept or reject the lot from which these sample data came. Choose
Stat > Quality Tools > Acceptance Sampling by Variables > Accept/Reject Lot and ­complete
the dialog box as given in Figure 10.20 and click OK.
Sampling by Variables for Proportion Nonconforming 265

FIGURE 10.17
Minitab worksheet Armscye.mpj showing a dialog box to convert shirt size values.

FIGURE 10.18
Minitab worksheet Armscye.mpj showing a dialog box to create delta values.
266 Acceptance Sampling in Quality Control

FIGURE 10.19
Minitab worksheet Armscye.mpj showing completed columns prior to analysis.

FIGURE 10.20
Minitab dialog box to analyze worksheet Armscye.mpj to determine lot acceptance or rejection.
Sampling by Variables for Proportion Nonconforming 267

Acceptance Sampling by Variables: Accept/Reject Lot

Make Accept or Reject Decision Using Delta
Sample Size 116
Mean -0.295797
Historical Standard Deviation 0.1
Lower Specification Limit (LSL) -0.5
Upper Specification Limit (USL) 0.5
Z.LSL 2.04203
Z.USL 7.95797
Critical Distance (k Value) 2.173
Decision: Reject lot.

So what does this mean? The lot of 10,000 shirts is rejected based on the data. The distribu-
tion center may want to send this lot back to the supplier or to sort the lot at the supplier’s
expense. Note that the mean is 0.2958 short of the nominal length. This may suggest that
the supplier is trying to reduce their material costs. The standard deviation of the sample
data is 0.098, which is very close to the historical value of 0.1. Thus, the process variation
has remained constant as the mean has shifted off center.
Now, suppose that we didn’t know the standard deviation, that is, an historical estimate
is unavailable. What would be the impact on the sample size n? The k value wouldn’t be
expected to change, but we can generate this variables sampling plan with Stat > Quality
Tools > Acceptance Sampling by Variables > Create/Compare… and complete the dialog
box as shown in Figure 10.21. Minitab now provides the MSD and produces an acceptance
region plot.

FIGURE 10.21
Minitab dialog box to generate a variables plan, for example, when standard deviation is unknown.
 268 Acceptance Sampling in Quality Control

Acceptance Sampling by Variables: Create/Compare

Lot quality in percent defective
Lower Specification Limit (LSL) –0.5
Upper Specification Limit (USL) 0.5
Lot Size 10000
Acceptable Quality Level (AQL) 1
Producer’s Risk (α) 0.05
Rejectable Quality Level (RQL or LTPD) 2
Consumer’s Risk (β) 0.1
Generated Plan(s)
Sample Size 388
Critical Distance (k Value) 2.17313
Maximum Standard Deviation (MSD) 0.205326
Z.LSL = (mean - lower spec)/standard deviation
Z.USL = (upper spec - mean)/standard deviation
Accept lot if standard deviation ≤ MSD, Z.LSL ≥ k, and Z.USL ≥ k;
otherwise reject.
Percent Probability Probability
Defective Accepting Rejecting AOQ ATI
1 0.951 0.049 0.914 859.6
2 0.102 0.898 0.197 9017.1
Average outgoing quality limit (AOQL) = 0.953 at 1.142 percent defective.

The unknown standard deviation variables plan for this example shows n = 388 (vs. 116 for
known standard deviation) with k = 2.173 (unchanged). The acceptance region plot is now
included with the OC, AOQ, and ATI curves as given in Figure 10.22.

Operating characteristic (OC) curve Acceptance region plot

1.00 0.2053
Probability of acceptance

0.20
0.75
Sample stDev

0.15

0.50 0.10

0.25 0.05

0.00 0.00
0 1 2 3 4 –0.50 –0.25 0.00 0.25 0.50
(a) Lot percent defective (b) Sample mean

Average outgoing quality (AOQ) curve Average total inspection (ATI) curve
1.00
AOQ (percent defective)

Average total inspection

9000
0.75
7000
0.50 5000
0.25 3000
1000
0.00
0 1 2 3 4 0 1 2 3 4
(c) Incoming lot percent defective (d) Lot percent defective
Sample size = 388, Critical distance = 2.17313
FIGURE 10.22
(a) Minitab OC, (b) acceptance region plot, (c) AOQ, and (d) ATI curves for variables plan when standard
­deviation is unknown.
Sampling by Variables for Proportion Nonconforming 269

Statgraphics
Statgraphics has the ability to produce variable sampling plans. Choose SPC >
Acceptance Sampling > Variables > OC/AOQL/LTPD Plans. Using the same example
as evaluated by Minitab, complete the dialog box as shown in Figure 10.23 and click
OK. The Tables and Graphs dialog box appears. Select all graph types as shown in
Figure 10.24 and click OK to generate the graphs. The output from Statgraphics agrees
with that from Minitab.

FIGURE 10.23
Statgraphics dialog box to generate a variables plan.

FIGURE 10.24
Statgraphics dialog box to select graph types.
270 Acceptance Sampling in Quality Control

Acceptance Sampling for Variables

Lot size: 10,000
Desired Features
Producer’s risk (alpha): 5.0%
Consumer’s risk (beta): 10.0%
Generated Plan
Sample size (n) = 116
Critical distance (k) = 2.17274
Plan Attributes
Acceptable quality level (AQL): 1.0%
Producer’s risk (alpha) = 4.90198%
LTPD: 2.0%
Consumer’s risk (beta) = 9.99992%
AOQL = 0.97888% at 1.13851% defective
ATI =
600.512 units per lot at the AQL
1402.11 units per lot at the AOQL
9011.61 units per lot at the LTPD

The StatAdvisor
This procedure creates sampling plans for lot inspection. In this case, it has generated a
sampling plan based upon desired producer's and consumer's risks. The plan states that
116 items should be sampled from each lot of 10,000, each of the items measured, and the
lot accepted without further inspection if the sample mean is at least 2.17274 standard
deviations away from the specification limit(s). Using such a plan, a lot containing 1.0%
defective items will be rejected only 4.90198% of the time, while a lot containing 2.0%
defective items will be accepted only 9.99992% of the time. If rejected lots are subjected
to 100% inspection and all bad items are replaced with good ones, the average percent of
defective items shipped will be no greater than 0.97888% (the AOQL).

The Statgraphics OC, AOQ, and ATI curves are shown in Figure 10.25. They agree with
those produced by Minitab in Figure 10.16.
We can generate the variables plan when the standard deviation is unknown by choos-
ing SPC > Acceptance Sampling > Variables > OC/AOQL/LTPD Plans and selecting the
Use sample estimate option for the process sigma as seen in Figure 10.26. Unfortunately,
Statgraphics does not produce an acceptance region plot as Minitab does, but the sampling
plans do agree.

Acceptance Sampling for Variables

Lot size: 10,000
Desired Features
Producer's risk (alpha): 5.0%
Consumer's risk (beta): 10.0%
Generated Plan
Sample size (n) = 388
Critical distance (k) = 2.17313
Sampling by Variables for Proportion Nonconforming 271

Operating characteristic (OC) curve

n = 116, k = 2.17274
1

0.8
Prob. of acceptance

0.6

0.4

0.2

0
0 1 2 3 4
(a) True percent defective

Average outgoing quality (AOQ) curve

n = 116, k = 2.17274
1

0.8
Percent defective

0.6

0.4

0.2

0
0 1 2 3 4
(b) True percent defective

Average total inspection (ATI) curve

(×1000) n = 116, k = 2.17274
10

8
Units inspected

0
0 1 2 3 4
(c) True percent defective

FIGURE 10.25
Statgraphics (a) OC, (b) AOQ, and (c) ATI curves for the generated variables plan.
272 Acceptance Sampling in Quality Control

FIGURE 10.26
Statgraphics dialog box to generate a variables plan, for example, when standard deviation is unknown.

Plan Attributes
Acceptable quality level (AQL): 1.0%
Producer's risk (alpha) = 4.98609%
LTPD: 2.0%
Consumer's risk (beta) = 9.98154%
AOQL = 0.950333% at 1.13712% defective
ATI =
867.263 units per lot at the AQL
1642.65 units per lot at the AOQL
9040.57 units per lot at the LTPD

The StatAdvisor
This procedure creates sampling plans for lot inspection. In this case, it has generated
a sampling plan based upon desired producer's and consumer's risks. The plan states
that 388 items should be sampled from each lot of 10000, each of the items measured,
and the lot accepted without further inspection if the sample mean is at least 2.17313
standard deviations away from the specification limit(s). Using such a plan, a lot con-
taining 1.0%  defective items will be rejected only 4.98609% of the time, while a lot
containing 2.0% defective items will be accepted only 9.98154% of the time. If rejected
Sampling by Variables for Proportion Nonconforming 273

lots are subjected to 100% inspection and all bad items are replaced with good ones,
the average percent of defective items shipped will be no greater than 0.950333%
(the AOQL).

Select all the graph types as shown in Figure 10.24 and click OK. Statgraphics produces
the same output and graphs as before.

Problems
1. Given the following specifications

p1 =  . 006  1 − α =  . 95

p2 =  . 057  β =  . 10

derive the following variables plans for percent nonconforming

a. X method – σ known, σ = .08
b. k method – σ unknown
by formula and from the Jacobson nomograph.
2. The upper specification limit for the response time for a certain emergency warn-
ing device is 7 s. They are known to be normally distributed. A sample of 10 such
devices yields the following results: 6.82, 6.85, 6.70, 6.73, 6.89, 6.95, 6.96, 6.80, 6.79,
and 6.85. Apply the plan obtained in Problem 1 to these data.
3. Convert the results of Problem 1a to the M method and apply the plan to the
results of Problem 2.
4. Determine the value of M to be used with Problem 1b. What is the estimated pro-
portion nonconforming if U = 1000, X = 973, s = 9, and n = 30? Should the lot be
accepted?
5. Convert the plan of Problem 1b to a plan using the range with subgroups of 5.
6. Maximum and minimum specifications for hardness of a certain material are 69
and 60 when read on a certain scale. Construct an unknown standard deviation
plan to the following specifications:

p1 = .01 1 - a = .95

p2 = .14 b = .10

7. If the results of sampling a lot from Problem 6 yielded 63.0, 64.5, 64.0, 62.5, 63.0,
70.0, 71.0, 61.0, 60.0, 67.5, 66.5, 64.0, and 68.0. Should the lot be accepted?
8. Using the tolerance interval approach, assess the results of Problem 7.
9. Would the double-sampling plan n1 = 10, n2 = 10, ka = 2.51, kr = 1.58, and kt = 2.05
lead to a second sample on the basis of Problem 2 results?
10. Draw the OC curve for the sampling plan given in Problem 1a. What is the indif-
ference quality level?
274 Acceptance Sampling in Quality Control

References
Bingham, R. S., 1962, Tolerance limits and process capability studies, Industrial Quality Control, 19(1):
36–39.
Bowker, A. H. and H. P. Goode, 1952, Sampling Inspection by Variables, McGraw-Hill, New York.
Duncan, A. J., 1955, The use of ranges in comparing variabilities, Industrial Quality Control, 11(5):
18–22.
Duncan, A. J., 1974, Quality Control and Industrial Statistics, 4th ed., Richard D Irwin, Homewood, IL.
Hamaker, H. C., 1979, Acceptance sampling for percent defective by variables and by attributes,
Journal of Quality Technology, 11(3): 139–148.
Jacobson, L. J., 1949, Nomograph for determination of variables inspection plan for fraction defec-
tive, Industrial Quality Control, 6(3): 23–25.
Jennett, W. J. and B. L. Welch, 1939, The control of proportion defective as judged by a single qual-
ity characteristic varying on a continuous scale, Supplement to the Journal of the Royal Statistical
Society (Series B), 6: 80–88.
Juran, J. M., Ed., 1999, Quality Control Handbook, 5th ed., McGraw-Hill, New York.
Larson, H.R., 1966, A nomograph of the cumulative binomial distribution, Industrial Quality Control,
23(6): 270–278.
Lieberman, G. J., 1958, Tables for one-sided statistical tolerance limits, Industrial Quality Control,
14(10): 7–9.
Lieberman, G. J. and G. J. Resnikoff, 1955, Sampling plans for inspection by variables, Journal of the
American Statistical Association, 50: 457–516.
Nelson, L. S., 1975, Use of the range to estimate variability, Journal of Quality Technology, 7(1): 46–48.
Nelson, L. S., 1977, Tolerance factors for normal distributions, Journal of Quality Technology, 9(4):
198–199.
Odeh, E. and D. B. Owen, 1980, Tables for Normal Tolerance Limits, Sampling Plans, and Screening, Marcel
Dekker, New York.
Ott, E. R., 1967, Analysis of means—A graphical procedure, Industrial Quality Control, 24(2): 101–109.
Owen, D. B., 1964, Control of percentages in both tails of the normal distribution, Technometrics, 6(4):
377–387 [Errata, 8(3): 570 (1966)].
Owen, D. B., 1967, Variables sampling plans based on the normal distribution, Technometrics, 9(3):
417–423.
Owen, D. B., 1969, Summary of recent work on variables acceptance sampling with emphasis on non-
normality, Technometrics, 11(4): 631–637.
Owen, D. B. and W. H. Frawley, 1971, Factors for tolerance limits which control both tails of the nor-
mal distribution, Journal of Quality Technology, 3(2): 69–79.
Owen, D. B., W. H. Frawley, C. H. Kapadia, and J. N. K. Rao, 1971, Tolerance limits based on range
and mean range, Technometrics, 13(3): 651–656.
Owen, D. B. and T. A. Hua, 1977, Tables of confidence limits on the tail area of a normal distribution,
Communications in Statistics, B6: 285–311.
Owen, D. B., D. D. McIntire, and E. Seymore, 1975, Tables using one or two screening variables to
increase acceptable product under one-sided specifications, Journal of Quality Technology, 7(3):
127–138.
Owen, D. B., J. N. K. Rao, and K. Subrahamaniam, 1972, Effect of non-normality on tolerance limits
which control percentages in both tails of the normal distribution, Technometrics, 14(3): 571–575.
Pearson, E. S., 1968, Tables of the Incomplete Beta-Function, 2nd ed., Cambridge University Press,
London, U.K.
Resnikoff, G. J. and G. J. Lieberman, 1957, Tables of the Non-Central t-Distribution, Standard University
Press, Stanford, CA.
Romig, H. G., 1939, Allowable average in sampling inspection, PhD dissertation, Columbia University,
New York.
Sampling by Variables for Proportion Nonconforming 275

Schilling, E. G., 1973, A systematic approach to the analysis of means, Part I. Analysis of treatment
effects, Journal of Quality Technology, 5(3): 93–108.
Schilling, E. G., 1974a, Sampling by variables, in Quality Control Handbook, 3rd ed. (J.M. Juran, Ed.),
Section 25, McGraw-Hill, New York, pp. 25.1–25.41.
Schilling, E. G., 1974b, Variables sampling and MIL-STD-414, Quality Progress, 7(5): 16–20.
Schilling, E. G. and D. J. Sommers, 1981, Two-point optimal narrow limit plans with applications to
MIL-STD-105D, Journal of Quality Technology, 13(2): 83–92.
Sommers, D. J., 1979, Personal communication with the author.
Sommers, D. J., 1981, Two-point double variables sampling plans, Journal of Quality Technology, 13(1):
25–30.
Statistical Research Group, Columbia University, 1947, Techniques of Statistical Analysis, McGraw-Hill,
New York.
United States Department of Commerce, 1950, Tables of the Binomial Probability Distribution, National
Bureau of Standards, Applied Mathematics Series No. 6, U.S. Government Printing Office,
Washington, DC.
United States Department of Defense, 1957, Military Standard, Sampling Procedures and Tables for
Inspection by Variables for Percent Defective (MIL-STD-414), U.S. Government Printing Office,
Washington, DC.
Wallis, W. A., 1947, Use of variables in acceptance inspection for percent defective, in Techniques of
Statistical Analysis (C. Eisenhart, M. Hastay, and W.A. Wallis, Eds.), Statistical Research Group,
Columbia University, McGraw-Hill, New York, Chapter 1, pp. 3–93.
Wallis, W. A., 1950, Lot quality measured by proportion defective, in Acceptance Sampling—A
Symposium, American Statistical Association, Washington, DC, pp. 117–122.
Zobel, S. P., 1958, One-sided statistical tolerance limits, Industrial Quality Control, 15(4): 35.
11
Attributes Sampling Schemes

Sampling Schemes
The sampling plans presented in the foregoing chapters provide the basis for more
sophisticated sampling designs. Sampling plans are frequently used in consort to pro-
duce levels of protection not attainable by any of the component plans individually.
Such combinations of sampling plans are called “sampling schemes” or “sampling sys-
tems.” The International Organization for Standardization (ISO) in ISO 3534-2 (2006)
has defined them as follows:

Acceptance sampling plan—A plan that states the sample sizes to be used and the
­associated criteria for lot acceptance
Acceptance sampling scheme—A combination of acceptance sampling plans with
switching rules for changing from one plan to another
Acceptance sampling system—A collection of acceptance sampling plans or acceptance
sampling schemes together with criteria by which appropriate plans or schemes
may be chosen

The sampling scheme then consists of a set of plans that are selected to be used as indicated
by a set of switching rules. These rules allow the user to go from one plan to another in a
prescribed fashion to obtain levels of performance not available from using just one plan.
A switching rule is defined by ISO 3534-2 (2006) as

Switching rule—An instruction within a sampling scheme for changing from one
acceptance sampling plan to another of greater or lesser severity of sampling
based on the demonstrated quality history

Sampling plans are the basic elements of sampling schemes, while sampling systems may
be considered to involve a grouping of one or more sampling schemes.

Quick Switching Systems

As a simple example of a sampling scheme, consider the quick switching system (QSS)
proposed by Romboski (1969). Probably the most straightforward of all sampling schemes
is his QSS-1 plan that proceeds as follows.

277
278 Acceptance Sampling in Quality Control

The following are given:

n as the sample size of tightened and normal plans

cN as the acceptance number under normal inspection
cT as the acceptance number under tightened inspection
PN as the probability of acceptance under normal sampling plan
PT as the probability of acceptance under tightened sampling plan

Utilize the following switching rules:

1. Start using the normal inspection plan.

2. Switch to the tightened inspection plan immediately after a rejection.
3. When on tightened inspection, switch to normal inspection immediately after
accepting a lot.
4. Alternate back and forth as dictated by these rules.

It is immediately obvious that there must be a flow of lots for the scheme to be applied.
Schemes require Type B sampling plans and are not intended for Type A sampling
involving single  lots since the switching rules are applied to a sequence of lots over
time. Also, note that there are three operating characteristic (OC) curves involved here,
namely, those for the tightened plan, the normal plan, and their combination in the com-
posite OC curve for the scheme. Romboski (1969) calculates the scheme probability of
acceptance, Pa, as
PT
Pa =
( N ) + PT
1 - P

Now suppose we use n = 20, c = 0 as the tightened plan and n = 20, c = 1 as the nor-
mal plan. These plans have a lot tolerance percent defective (LTPD) of 11.5% and 19.5%,
respectively, using the Poisson approximation. When combined in a quick switching for-
mat, the resulting LTPD is 12.64%—close to the c = 0 plan. Moreover, the tightened and
normal plans have an acceptable quality level (AQL) (Pa = 0.95) of 0.255% and 1.77%,
respectively, while the quick switching–combined AQL is 1.54%—close to the c = 1 plan.
Thus, the scheme has captured the best features of both its constituents. Furthermore, the
operating ratio of the quick switching scheme is 8.21, while the component plans have
operating ratios of 44.9 and 10.9, respectively. So the scheme is more discriminating than
either of its components.
As a check on our computations, we may wish to compare the composite probability of
acceptance for the LTPD of 0.1264 using Romboski’s formula and Poisson probabilities.
We have

PT = F ( n = 20, c = 0, p = 0.1264, np = 2.528 ) = 0.0798

PN = F ( n = 20, c = 1, p = 0.1264, np = 2.528 ) = 0.2816
0.0798
QSS Pa = = 0.10
(1 - 0.2816 ) + 0.0798

as it should be. The QSS is discussed in detail in Chapter 17.

Attributes Sampling Schemes 279

Tightened–Normal–Tightened Plans
The QSS maintains the same sample size for both tightened and normal inspection,
­varying the acceptance number to achieve different probabilities of acceptance. A related
sampling scheme, the tightened–normal–tightened (TNT) plans proposed by Calvin
(1977), maintains the same acceptance number for the tightened and normal plans while
varying the sample size between them. This is especially useful when c = 0 plans are
required since it provides operating ratios on the order of 23, much less than that in
using a c = 0 plan alone. The schemes are applied as follows:

1. Start on tightened inspection using the plan (n, cT).

2. Switch to normal inspection when t consecutive lots are accepted and then use the
normal plan (n, cN).
3. Switch to tightened inspection when 2 out of s + 1 lots are rejected. Use the plan (n, cN).
4. Go to step 2.

The TNT plans are discussed more extensively in Chapter 17. Note that the combination
t = 5 and s = 4 provides a switching procedure close to that in Military Standard 105E
(MIL-STD-105E) and its derivatives. The 105 series (MIL-STD-105A through MIL-STD-
105E) has a long history dating back to World War II. This acceptance sampling system is,
without doubt, the most used and most copied set of standards in the world.

MIL-STD-105E and Derivative Standards

The MIL-STD-105 series of standard is an outgrowth of statistical contributions to the
war effort in World War II. It came from a need for a sampling system that did not
require 100% inspection for use in testing munitions and for other destructive tests. The
result was the Army Service Forces inspection tables that came out in 1942 and 1943.
Improvement led to MIL-STD-105A, B, …, E in subsequent years. The Army discontinued
the support for military statistical standards on February 27, 1995, proposing the civilian
standards. Meanwhile, other standards-writing bodies, such as the American National
Standards Institute (ANSI), the ISO, and the International Electrotechnical Commission,
developed their own derivatives of 105 as civilian standards. MIL-STD-105E, the last of
the seminal series, will be discussed here and differences with these derivative standards
will be indicated. MIL-STD-105E is, after all, regarded as a classic system worldwide.
Many sampling schemes are included in what are called AQL systems. AQL refers to
the acceptable quality level,* that is, what has been called the producer’s quality level in
single-sampling plans. These systems are intended to be applied to a stream of lots. Such
plans specify an upper limit on quality, the AQL, not to be exceeded by the producer with-
out the penalty of an excessive number of rejected lots. That is, for levels of quality less
than the AQL, rejection will be relatively infrequent, say, less than 1 in 10, while for levels
of quality in excess of the AQL, rejections will be more frequent, say, more than 1 in 10.
This is achieved by switching back and forth between normal, tightened, and reduced
plans included in the system. Tighter plans are used when quality levels are shown to

* In more recent attributes sampling standards, the AQL is referred to as the acceptance quality limit.
280 Acceptance Sampling in Quality Control

be poor, while looser plans involving smaller sample sizes are utilized when quality is
shown to be good. Over a continuing supply, schemes can be devised to incorporate the
best properties of the plans included as elements. Frequently, schemes are selected within
a system in relation to the lot size involved.
MIL-STD-105E (p. 2) defines the AQL as follows:
When a continuous series of lots is considered, the AQL is the quality level which, for
the purposes of the sampling inspection, is the limit of a satisfactory process average.

The definition of AQL in acceptance sampling has created controversy over the years because of
the implication that nonzero levels of quality are acceptable. Accordingly, the ISO has changed
the acronym AQL to stand for acceptance quality limit, defined as the “worst tolerable quality
level.” It is evident that the meanings of the two definitions are essentially the same. Since the
terms are in transition, we will generally use the acronym AQL wherever possible and depend
on the text to highlight any differences from the definitions presented earlier.
MIL-STD-105E (1989) is not a sampling plan. It is a sampling system. As such, it
­combines several individual sampling plans in schemes constructed to employ economic,
­psychological, and operational means to motivate the producer to sustain the quality at
levels less than or equal to the AQL. The procedure for switching between plans is ­essential
to the system; it is so designed as to exert pressure on the producer to take corrective action
when quality falls below prescribed levels and to provide rewards, in terms of reduced
sample size, for quality improvement.
The standard ties together sets of three attributes sampling plans, each at a different level
of severity, into a unified procedure for lot acceptance through the use of its switching
rules. These action rules determine the level of severity to be employed depending on the
level of quality previously submitted. Thus, inspection of a succession of lots is intended to
move among the specified set of tightened, normal, and reduced sampling plans as qual-
ity levels degenerate or improve. Switching between tightened and normal plans is made
mandatory by the standard, while the use of reduced plans is optional.
The MIL-STD-105E system, as such, does not allow for application of individual plans with-
out the use of the switching rules, since such an approach can lead to serious loss of p
­ rotection
from that achieved when the system is properly applied. Quality levels are ­specified in terms
of AQL for the producer, while consumer protection is afforded by the switching rules that
lead to tighter plans when quality is poor. The operation of MIL-STD-105D has been described
in detail by Hahn and Schilling (1975) and is the subject of ­several military and international
handbooks. The D and E versions of MIL-STD-105 exhibit minor editorial changes while the
tables are essentially the same; however, the rule for discontinuing inspection was modified
from 10 lots on tightened to 5 rejections while on tightened inspection.
When an isolated lot is to be inspected, special tables of limiting quality (LQ) are
­presented in the standard. Used in such instances, MIL-STD-105E merely represents a
convenient collection of individual plans indexed by the LQ table. In no sense, however,
this is the use for which the MIL-STD-105E system was designed.
Unfortunately, the standard may sometimes be misused, particularly in nonmilitary
applications, through the selection and use of normal plans only—disregarding the
tightened and reduced plans and the switching rules. This deprives the consumer of
the protection provided by the tightened plan when quality is poor, and it foregoes the
advantage to the producer of smaller sample sizes and slightly increased protection
afforded by the reduced plan when quality is good.
The operation of MIL-STD-105E is straightforward. Lot sizes are linked to sample size
by a system of code letters. Matched sets of single, double, and multiple plans provide
Attributes Sampling Schemes 281

General Master Limit numbers Operating

for reduced characteristics
description tables inspection and derivatives

Limit
numbers
for reduced
Sample Single Double Multiple AOQL Limiting
size code factors
letters sampling sampling sampling quality
(single Average
sampling) sample
size curves
Tightened

Tightened

Tightened
Detailed
Reduced

Reduced

Reduced

Pa = 10%
Normal

Normal

Normal

Pa = 5%
Tightened
plans and

Normal
OC curves

Percent defective

Percent defective
Defects/100

Defects/100
FIGURE 11.1 
Structure of MIL-STD-105E.

a complete choice among these types of plans in application. The average sample size of
double and multiple plans can be arrived at from average sample number (ASN) curves,
which are given. MIL-STD-105E also contains tables presenting the average outgoing qual-
ity limit (AOQL) resulting from the use of its normal plans together with 100% inspection
of rejected lots. Complete sets of OC curves and probability points of the normal and tight-
ened plans are contained in the standard. The standard is written in terms of inspection for
defectives (expressed in percent defective) and also for defects (expressed in defects per
100 units). The approach and operation of the scheme are the same for both and so they
will be used interchangeably here for economy of presentation. Their measures of perfor-
mance, however, are based on different probability distributions (binomial and Poisson)
and so they must be addressed separately where OCs and other measures are concerned.
The structure of MIL-STD-105E is shown in Figure 11.1.

Operation
Proper use of the MIL-STD-105E sampling system demands close attention and adherence
to the rules for switching among the sets of three plans (tightened, normal, and reduced),
which are presented. In doing so, the producer receives adequate protection against exces-
sive rejections when quality is better than the AQL, while the consumer receives increased
protection when quality is running worse than the AQL. The operation of the switching
rules is shown in Figure 11.2.
282 Acceptance Sampling in Quality Control

Start
Preceding 10 lots accepted,
with total defectives
less than limit number, and
production steady, and Two out of five
approved by responsible consecutive
authority lots rejected

Reduced Normal Tightened

Lots rejected or Five consecutive

lots accepted but defects lots accepted
found lie between
Total of five
Ac and Re of plan, or
lots rejected
production irregular or
on tightened
other conditions warrant

Discontinue
inspection

FIGURE 11.2 
Switching rules for MIL-STD-105E. (Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10, 77, 1978;
Schilling, E.G. and Johnson, L.I., J. Qual. Technol., 12(4), 220, 1980; Schilling, E.G. and Sommers, D.J., J. Qual. Technol.,
13(2), 83, 1981. With permission.)

A MIL-STD-105E scheme always starts with the normal inspection plan. The plan ­continues
to be used until sufficient evidence is generated to indicate that a switch to the tightened or
reduced plan is appropriate. Note that MIL-STD-105E makes use of the reduced plan optional.
Although for full economic benefit of the procedure, it should be utilized where possible.
A switch to tightened inspection roughly involves moving to the acceptance criteria of
the next lower AQL category while retaining the sample size used in the normal plan.
This results in a more stringent plan with less consumer’s risk at the expense of increased
producer’s risk. Tightened inspection is imposed when two out of five consecutive lots are
rejected on original inspection. Normal inspection is reinstated when five consecutive lots
are accepted on original inspection.
A switch to reduced inspection involves changing both the sample size and the accep-
tance number. Sample size is roughly reduced two sample size code letter categories
below that originally used for normal inspection. The final acceptance and rejection
­numbers are separated by a gap. If the number of defectives found falls in the gap, the
lot is accepted but the scheme reverts to the normal plan on the next lot. The gap is used
to prevent rejection of a lot on reduced inspection when it might be accepted under
­normal inspection. Otherwise, the acceptance and rejection numbers are used in the
­conventional manner on reduced inspection. A shift is made to reduced inspection when

1. The preceding 10 lots have been accepted on original inspection under normal
sampling
2. The total number of defectives from the preceding 10 lots is less than or equal to
the limit numbers given in Table VIII of the standard (The results from all samples
(not just first samples) should be used if double or multiple sampling is employed.)
Attributes Sampling Schemes 283

3. Production must be steady

4. Reduced inspection is considered desirable by responsible authority

MIL-STD-105E Table VIII is reproduced here as Appendix Table T11.1. To use the table, the
accumulated sample size from the last 10 lots is entered and the limit number read from
the AQL. When the accumulated sample size is not sufficient for reduced inspection, addi-
tional lots must be taken until a limit number can be obtained from the table. Clearly, the
additional lots must be from the same uninterrupted sequence.
Normal inspection must be reinstated from reduced when

1. A lot is rejected
2. The results of inspection of a lot fall in the gap between the reduced acceptance
and rejection numbers
3. Production becomes irregular or delayed
4. Other conditions warrant

MIL-STD-105E was intended to be used with a continuing series of lots or batches.

However, occasionally specific plans may be selected from the standard and used without
the switching rules. This is not the intended application of MIL-STD-105E and its use in
this way should not be referred to as inspection under MIL-STD-105E. When employed
in this manner, the standard simply represents a repository for a collection of individual
plans indexed by AQL. In this sense, AQL has no operational meaning and the OCs and
other measures of a plan so chosen must be assessed individually for that plan from the
tables of performance provided in MIL-STD-105E. It is a convenience to the user that tables
are provided to be used in this way. They are described in the following section.

Selection
The selection of a set of tightened, normal, and reduced plans from MIL-STD-105E is
fairly straightforward. The key elements in the selection of a plan are lot size and AQL.
The definition of a lot is often governed by the operational situation and the available
information. Lots may be composed of the material delivered at one time or produced at
the same time (a day or a month) or made under a particular set of operating conditions
(raw material, operator, etc.). The standard suggests that “as far as practicable, each lot
should consist of units of product manufactured under essentially the same conditions
and at essentially the same time.” A quest for homogeneity tends toward small lots, while
large lots are desirable in allowing for larger sample sizes with greater discrimination
between good and bad quality. Thus, determination of lot size is often a compromise
­frequently settled by practical considerations.
The AQL is central to the entire MIL-STD-105E system. It must be set with due
­consideration for the producer’s process capability and the consumer’s need for a
­reasonable quality level relative to the state of the art. The ideal AQL would be set
in terms of process capability studies to determine the reasonable levels and costs of
quality for the producer’s process as well as the tolerance of the consumer to changes
in quality level and the associated costs. An excellent discussion of process capability
studies will be found in Mentch (1980).
284 Acceptance Sampling in Quality Control

A further consideration in the determination of AQLs is defect class. MIL-STD-105E

uses the defect classification:

Critical: Likely to result in hazardous or unsafe conditions

Major: Likely to result in failure or to reduce the usability materially
Minor: Not likely to reduce the usability materially

AQLs are usually assigned to each category with each defect type in the category counted
against the category AQL.
Ultimately, the AQL to be used in sampling must be determined by negotiation between
the producer and the consumer with due consideration of the trade-offs in both physical
and economic sense. The sequence of steps involved in the selection of a set of plans from
MIL-STD-105E is shown in Figure 11.3.

Start

Check producer’s process

average-control chart

Determine
lot size and AQL

Determine sampling procedure

and select plan

Single Double Multiple

Tightened–Table III-B

Tightened–Table IV-B
Tightened–Table II-B

Reduced–Table III-C

Reduced–Table IV-C
Reduced–Table II-C

Normal–Table III-A

Normal–Table IV-A
Normal–Table II-A

Check operating characteristics

to insure protection is as desired

Apply switching rules

(paragraph 8)

FIGURE 11.3 
Check sequence for selecting a plan from MIL-STD-105E.
Attributes Sampling Schemes 285

Once the lot size and AQL have been determined, a set of sampling plans can be
found. The lot size is used to enter Table I of the standard, reproduced here as Appendix
Table T11.2. A sample size code letter is then obtained appropriate to the inspection level
to be used. Inspection Level II is normally used unless some other inspection level is
specified. Inspection Levels I and III allow for control of discrimination (lower or higher)
depending on past history and operating circumstances. The special inspection levels, S-1
through S-4, are generally used with expensive or destructive tests where sample size is at
a premium, and more extensive inspection is not economic or not warranted on the basis
of past history and the intent of application of the plan.
The advantages and disadvantages of single-, double-, and multiple-sampling plans
have previously been discussed. Single-sampling plans are easy to administer and under-
stand. Double-sampling plans allow for a reduction in average sample size at the expense
of the possibility of taking an additional sample. Multiple-sampling plans are somewhat
difficult to administer but provide the greatest economy in terms of ­average sample size.
The choice, again, depends on the operational situation and the experience and reli-
ability of the inspection personnel involved. MIL- STD-105E does provide ASN curves
to help in the allocation of inspection effort when double- and multiple-sampling plans
are involved.
Once the sampling procedure has been selected, a set of plans is found in the appropriate
tables of the standard. Figure 11.4 shows how the sampling plans are selected and gives
the decision rules to be applied in application of the plan.

Step Tightened Normal Reduced

MIL-STD-105E Table I
Preparatory Determine sample size code letter from
lot size and inspection level (ii)
Determine criteria
Single sampling MIL-STD-105E MIL-STD-105E MIL-STD-105E
Table IIB Table IIA Table IIC
Double sampling MIL-STD-105E MIL-STD-105E MIL-STD-105E
Table IIIB Table IIIA Table IIIC
Multiple sampling MIL-STD-105E MIL-STD-105E MIL-STD-105E
Table IVB Table IVA Table IVC
Decision rules Accept if total Accept if total Accept if total
defects ≤ Ac defects ≤ Ac defects ≤ Ac
Reject if total Reject if total Reject if total
defects ≥ Re defects ≥ Re defects ≥ Re
Accept lot but
switch if total

sample fall in gap

Switching See Figure 11.2 for switching rules

FIGURE 11.4 
Application of MIL-STD-105E.
286 Acceptance Sampling in Quality Control

The master tables for plan selection are reproduced herein appendix as follows:

Appendix Table MIL-STD-105E Table Content

T11.1 Table VIII Limit number for reduced inspection

T11.2 Table I Sample size code letters
T11.3 Table IIA Single sampling, normal inspection
T11.4 Table IIB Single sampling, tightened inspection
T11.5 Table IIC Single sampling, reduced inspection
T11.6 Table IIIA Double sampling, normal inspection
T11.7 Table IIIB Double sampling, tightened inspection
T11.8 Table IIIC Double sampling, reduced inspection
T11.9 Table IVA Multiple sampling, normal inspection
T11.10 Table IVB Multiple sampling, tightened inspection
T11.11 Table IVC Multiple sampling, reduced inspection

These tables completely specify the plans included in MIL-STD-105E. In the selection of
specific plans from the tables, however, it must be emphasized that the vertical arrows
direct the user to a completely new set of acceptance criteria, that is, both the sample size
and acceptance number of the indicated plan must be used in satisfying the intent of the
arrow.

Measures
MIL-STD-105E contains detailed tables showing the measures of performance of individual
plans. This includes the following:

Appendix Table MIL-STD-105E Table Content

T11.12 Table VB AOQL for normal plans

T11.13 Table VIA AOQL for tightened plans
T11.14 Table VIA LQ in percent defective for Pa = 10%
T11.15 Table VIB LQ in defects = 100 units for Pa = 10%
T11.16 Table VIIA LQ in percent defective for Pa = 5%
T11.17 Table VIIB LQ in defects = 100 units for Pa = 5%
F11.18 Table IX ASN curves for double and multiple sampling
T11.19 Table X OC curves and probability points for plans by
code letter (Code F only)

These tables are presented here as Appendix Tables T11.12 through T11.19 as indicated
earlier.
The distinction between percent defective and defects per 100 units is particularly
important when dealing with OCs and other measures of performance. Since Type B OCs
are involved, the binomial distribution is exact in assessing the percent defective, while the
Poisson distribution is employed in determining the defects per 100 units. This is carried
Attributes Sampling Schemes 287

out through MIL-STD-105E and appears explicitly in Table X. The Poisson distribution,
however, is used as an approximation to the binomial except for AQLs of 10.0 or less and
sample sizes of 80 or less. This simplifies the presentation somewhat with little loss of
accuracy. The AOQL tables are based on the Poisson distribution as are the ASN curves.
The tables of OCs and other measures are fairly self-explanatory; however, certain
­features should be pointed out:

1. AOQL factors are approximate in Table V and can be corrected by multiplying by

(1 − [sample size/lot or batch size]).
2. ASN curves in Table IX are selected by the single sample size acceptance number c.
The vertical axis is then interpreted in proportion to the single sample size n at the
top, and the horizontal axis in terms of unity values np. As a result, a comparison
of single, double, and multiple average sample sizes can be made for any matched
set in the standard. For example, for Code F, 2.5% AQL, the single-sampling plan
is n = 20, c = 1. Using the c =1 set of curves, the vertical axis becomes 0, 5, 10, 15, 20
and the horizontal axis becomes 0, .05, .10, .15, .20. The double and multiple plans,
then, have approximately the same average sample size at a proportion defective
of about p = .10, that is, where np = 2. The arrows show the position of the AQL,
obviously at a proportion defective of .025 for the plan n = 20, c = 1. This is made
possible by the fact that for a given single-sampling acceptance number, the prod-
uct (n × AQL) is constant for all sample sizes.

Scheme Properties
The OC curves and other measures of performance given in MIL-STD-105E relate to the
performance of the constituent individual plans and so can be used to assess its operation
at any given stage or to determine how the plans will perform in moving from normal to
tightened or reduced inspection. This is helpful in determining the AQLs. Unfortunately,
the standard does not give measures of performance of the system as a whole, including
the switching rules. Detailed tables of scheme performance patterned after the MIL-STD-
105E tables cited earlier have been prepared by Schilling and Sheesley (1978). They are
based on the work of Stephens and Larson (1967) and Burnett (1967), which did much
to develop the theory of evaluation of scheme characteristics. They are also included in
the ANSI/ASQ Z1.4 standard, which is the ANSI equivalent of MIL-STD-105E. While the
original work was done with MIL-STD-105D, the results apply to MIL-STD-105E since
discontinuation of inspection was not included in the evaluation.
A representation of the Schilling–Sheesley tables is given in the appendix as follows:

MIL-STD-105E Individual
Appendix Scheme Table Plans Table Content

T11.20 Table XI MIL-STD-105E scheme AOQL

T11.21 Table XII MIL-STD-105E scheme LQ for Pa = 10%
T11.22 Table XIII MIL-STD-105E scheme LQ for Pa = 5%
T11.23 Tables XIV, XV MIL-STD-105E scheme Pa, ASN, AOQ,
ATI (Code F only)
288 Acceptance Sampling in Quality Control

The first three tables correspond directly to those given only for the individual nor-
mal and tightened plans in MIL-STD-105E. The fourth is an example of the complete
listing of measures by code letter and provides values for examples to follow. They
­characterize the performance of the standard when it is properly used, with the switch-
ing rules.
In application of the MIL-STD-105E system, it is intended that a switch to tightened
inspection with possible discontinuation of inspection will, in the case of poor quality,
provide a psychological and economic incentive for the producer to improve the level of
quality submitted in actual application; this may or may not be the case. When used in
internal inspection to take advantage of the increased protection and economy afforded
by the switching procedure, a scheme may be used with no intention to discontinue the
inspection. Further, in the early stages of process development, producers may expect to
have a large proportion of lots rejected and it may be impossible to improve the process
given the state of the art. MIL-HDBK-53 (1965) points out that when inspection is discon-
tinued and “If the supplier otherwise has an excellent quality history for similar products,
the specified AQL should be investigated.” Thus, the AQL and not the process may be
changed. It is quite possible, however, as pointed out by Stephens and Larson (1967) that
“the actual behavior of the process under the influence of the sampling procedure may
be … very dynamic.”
In discussing the problem of evaluating the performance of a sampling system, such as
MIL-STD-105E, which may itself induce such process changes, Stephens and Larson

… adopt a somewhat simpler model which is tractable and which permits relative com-
parisons to be made between different plans or … different sets of plans … which allows
an evaluation of the operating behavior of the system of plans for different values of frac-
tion defective. This is the same type of approach taken in the presentation of an ordinary
OC curve for a sampling plan.

The same approach has also been used by Pabst (1963) and by Dodge (1965). A producer
would not usually be expected to operate at the LQ level of any simple sampling plan
or complex sampling scheme for very long without taking action of one kind or another.
However, ordinary OC curves do not reflect such actions.
The Stephens–Larson model as evaluated by Schilling and Sheesley (1978) does not
incorporate considerations of possible process changes resulting from psychological pres-
sures inherent in the use of the switching rules or discontinuation of inspection. After dis-
continuation, the Schilling–Sheesley tables essentially assume the restart under tightened
inspection with no change in fraction defective. Thus, the term scheme performance, as
used with respect to the scheme OC curve, has a very special meaning. It refers to how
the MIL-STD-105E system of switching rules would operate at a given process level under
the assumption that the process stays at that level even after discontinuation of inspec-
tion. Thus, discontinuation does not play a part in the Schilling–Sheesley evaluation and
so the values apply regardless of the rules for discontinuation. It should be noted that
this gives a conservative worst-case description of the performance of a scheme in the
sense that if ­psychological pressures were operative, the probability of acceptance at low
levels of fraction defective would be increased while the probability of acceptance at high
levels of fraction defective would be decreased relative to the values given by Schilling
and Sheesley.
The compilation of complete tables of measures of scheme performance allows the
­following approximate procedure to be used when the stream of consecutive lots, on which
Attributes Sampling Schemes 289

MIL-STD-105E is based, is broken to produce an isolated lot (known not to be the part of
the stream) or a short sequence of lots of a unique character:

1. Obtain the LQ for the scheme at Pa = 10% from Schilling–Sheesley (Appendix

Table T11.21), using the appropriate AQL and sample size code letter.
2. Select the plan from the MIL-STD-105E LQ table with Pa = 10% (Appendix
Tables T11.14 and T11.15), which has the LQ of the scheme at the AQL listed.

This procedure will guarantee about the same protection on the isolated lot as would
have been obtained under the use of the switching rules with the continuing series of lots.
A more refined approach is given later.

Implementation of MIL-STD-105E
The implementation of MIL-STD-105E is probably best explained by example. Suppose the
producer and the consumer agree on an AQL of 2.5% and lot sizes are expected to be N = 100.
Inspection Level II will be used since no other inspection level was agreed upon. Using the
lot size of 100 and Inspection Level II, the sample size code letter table (Appendix Table
T11.2) gives Code F. Using the master tables for tightened, normal, and reduced inspection,
we find the following set of matched single, double, and multiple plans to apply.

Tightened Normal Reduced

Code F, 2.5 AQL n Ac Re n Ac Re n Ac Re
Single 32 1 2 20 1 2 8 0 2
Double 20 0 2 13 0 2 5 0 2
20 1 2 13 1 2 5 0 2
Multiple 8 # 2 5 # 2 2 # 2
8 # 2 5 # 2 2 # 2
8 0 2 5 0 2 2 0 2
8 0 3 5 0 3 2 0 3
8 1 3 5 1 3 2 0 3
8 1 3 5 1 3 2 0 3
8 2 3 5 2 3 2 1 3

The plans under tightened inspection are found by the use of the arrow, which directs the user
to the next set of sample sizes and acceptance numbers. The symbol # in multiple sampling
indicates that no acceptance decision can be made at that stage of the sampling plan. Notice
that the final acceptance and rejection numbers under reduced inspection d
­ iffer by more than 1,
thus showing the gap that can lead to lot acceptance with a switch to normal inspection.
Suppose single samplings were used. The first lot would be inspected using the plan
n = 20, c = 1. This plan would continue in use on subsequent lots until a switch was called
for. At that time, the plan n = 32, Ac = 1, Re = 2 or n = 8, Ac = 0, Re = 2 would be used
depending on whether switch was to tightened or reduced inspection. For example, con-
sider the following sequence of lot acceptance (A) and rejection (R).

A A R A A A R R A R A A A A A A A A A A A A A A A A A A A A R
290 Acceptance Sampling in Quality Control

Inspection would start using the normal plan. At the second rejection, a switch to tight-
ened inspection is instituted since two out of five lots are rejected under normal inspection.
Tightened inspection continues until the 15th lot signals a switch to normal. A switch to
reduced inspection is called for after 10 lots are accepted under normal inspection. However,
the total number of defectives in the last 10 lots must be less than the limit number of 2 found
in MIL-STD-105E Table VIII for 200 accumulated sample units. If one defective was found,
the switch would be made, only to revert back to normal inspection with the rejection at the
end of the sequence. This, of course, assumes the other conditions for switching were met.
The measures of performance of these individual plans are easily found from the tables
given in the appendix. They are

AOQL (normal) = 4.2% (3.4% corrected)

AOQL (tightened) = 2.6% (1.8% corrected)
10% LQ (normal, percent) = 18%
10% LQ (normal, defects) = 20 defects per 100 units
5% LQ (normal, percent) = 22%
5% LQ (normal, defects) = 24 defects per 100 units
ASN at AQL (double) ≃15
ASN at AQL (multiple) ≃18

MIL-STD-105E Table X for Code F reproduced here as Appendix Table T11.19 shows
­probability points for tightened and normal inspection for percent defective. The plan
n = 8, c = 1 was similarly evaluated for reduced inspection. Appendix Table T11.23 shows
the probability points of the resulting scheme as a whole as computed by Schilling and
Sheesley (1978). They may be compared as follows:

Pa Normal Tightened Reduced Scheme

.99 0.75 0.475 2.00 0.978

.95 1.80 1.13 4.64 1.85
.90 2.69 1.67 6.88 2.47
.75 4.81 3.01 12.1 3.66
.50 8.25 5.19 20.1 5.40
.25 12.9 8.19 30.3 8.21
.10 18.1 11.6 40.6 11.6
.05 21.6 14.0 47.1 14.0
.01 28.9 19.0 58.8 19.0

Notice that the scheme performance is slightly looser than that of the normal inspection
plan for levels of quality well below the AQL but is much tighter for levels of quality above
the AQL. In fact, the scheme performance is close to that of the tightened plan at or below
the indifference quality level. This illustrates the advantage of using the scheme over any
of its individual component plans. This can also be seen in the composite OC curve shown
in Figure 11.5.
Since different sample sizes are involved in the plans constituting the scheme for a given
code letter and AQL, the sample size for the scheme can only be represented as an expected
value. This is the ASN for the scheme. Although the scheme OC curve shows minimal
increase in probability of acceptance over that for the normal plan alone when quality is
Attributes Sampling Schemes 291

1.0

0.9

0.8

0.7

0.6
Tightened Reduced
Pa

0.5

0.4

0.3

0.2

0.1 Normal
Scheme
0
2 4 6 8 10 12 14 16 18 20
P%

FIGURE 11.5 
Scheme OC curves, Code F, 2.5% AQL. (Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol.,
10, 79, 1978. With permission.)

35
Scheme Tightened
30

25

20 Normal
ASN

15

10
Reduced
5

0
02 04 06 08 10 12 14 16 18 20
P%

FIGURE 11.6 
Scheme ASN curves, Code F, 2.5% AQL. (Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol.,
10, 80, 1978. With permission.)

good, the reduction in the ASN in that region is substantial because of the possibility of
going to reduced inspection. This may be seen in the ASN curve for Code F, 2.5% AQL,
shown in Figure 11.6. The sample sizes for the component plans are also indicated.
When rectification is employed, the AOQL for the scheme is much improved over levels
reported for the normal inspection plan, although they are not as low as that given of the
tightened plan. This can be seen for Code F, 2.5% AQL in the average outgoing quality
(AOQ) curve shown in Figure 11.7.
292 Acceptance Sampling in Quality Control

10
Reduced
9

6
AOQ

2
Normal
1
Tightened Scheme
0
2 4 6 8 10 12 14 16 18 20
P%

FIGURE 11.7 
Scheme AOQ curves, Code F, 2.5% AQL. (Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol.,
10, 80, 1978. With permission.)

120 Scheme
Normal
100

80
Tightened
ATI

60 Reduced

40

20

0
2 4 6 8 10 12 14 16 18 20
P%

FIGURE 11.8 
Scheme ATI curves, Code F, 2.5% AQL. (Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol.,
10, 80, 1978. With permission.)

The average total inspection is given as a guide to determine the inspection effort require-
ments when rectification is used with an MIL-STD-105E scheme. A plot of the ATI curve for
Code F, 2.5% AQL is given in Figure 11.8.
Curves such as those shown can easily be constructed from the tables presented by
Schilling and Sheesley (1978).
A comparison of single, double, and multiple plans for Code F, 2.5% AQL indicates that
the OC curves of the schemes using these sampling procedures are about as well matched
as those of the constituent individual plans. This can be seen in Table 11.1. It also suggests
TABLE 11.1
Attributes Sampling Schemes

Scheme Pa and ASN Compared to the Normal Plan Only for Single, Double, and Multiple Sampling (Code F, 2.5% AQL)
Probability of Acceptance ASN

Scheme Normal Only Scheme Normal Only

P Single Double Multiple Single Double Multiple Single Double Multiple Single Double Multiple

0.978 99 98.5 99.6 98.4 98.0 98.9 14.6 11.7 8.7 20 14.5 16.4
1.85 95 93.2 96.4 94.8 93.5 96.1 19.1 15.9 16.3 20 15.5 17.4
2.47 90 87.2 92.1 91.4 89.4 93.1 21.5 18.3 19.0 20 16.1 17.9
3.66 75 70.6 77.6 83.5 80.3 85.5 26.2 23.3 24.0 20 17.0 18.7
5.40 50 46.7 49.2 70.6 66.1 71.9 30.8 26.9 28.2 20 17.7 19.1
8.21 25 23.9 22.4 50.3 45.4 49.4 32.0 26.4 25.1 20 18.0 18.5
11.6 10 10.4 8.1 30.8 27.1 28.2 32.0 24.5 20.4 20 17.5 16.7
14.0 5 5.7 3.8 20.8 18.3 18.1 32.0 23.2 17.8 20 16.9 15.2
19.0 1 1.6 0.8 8.4 7.7 6.6 32.0 21.4 14.1 20 15.6 12.4
Source: Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10, 82, 1978. With permission.
293
294 Acceptance Sampling in Quality Control

that savings in sample size may result when using double or multiple plans with the
switching procedure, particularly below AQL levels of percent defective.
Thus, there are significant advantages in the use of the switching rules to achieve OCs
and other measures of performance not attainable through individual sampling plans.

Matching Individual Sampling Plans

to MIL-STD-105E System Performance
Occasionally, it is necessary or economically desirable to abandon the scheme aspect
of MIL-STD-105E in favor of a single-sampling plan. This may be because a unique or
­isolated lot and not a continuing stream is to be inspected. It may be that the stream of
lots is too short to provide effective use of the MIL-STD-105E switching rules. When this
is the case, the sampling plan for normal inspection is sometimes incorrectly selected to
be used alone without the switching rules. This results in less consumer protection than
would be afforded by the use of the overall scheme. A more conservative approach for
the consumer would be to use the tightened plan alone, but this can result in an objec-
tionably high level of rejection for the producer of quality at or better than the AQL.
For  example, for Code H, 1.5% AQL, the scheme will accept 1.59% defective quality
95% of the time and reject 7.56% defective 90% of the time. The normal plan will accept
1.59% defective 81% of the time while the tightened plan will reject 7.56% defective 73%
of the time.
It is possible to use the unity values developed by Schilling and Johnson (1980) shown
in Appendix Table T6.1 to derive a unique individual sampling plan to match the scheme
performance of MIL-STD-105E. The plan obtained will usually require a larger sample size
than that given for the normal plan since protection under the scheme is better than under
the normal plan taken alone. The difference in sample size reflects the advantage in protec-
tion obtained by using the switching rules.
Appendix Table T11.24 from Schilling and Johnson (1980) shows values of the operating
ratio R for the AQL–code letter combinations of the MIL-STD-105E system. It was derived
from the tabulations of the MIL-STD-105E scheme performance by Schilling and Sheesley
(1978) and includes switching among tightened, normal, and reduced plans. Appendix
Table T11.21, from Schilling and Sheesley (1978), shows the LTPD (LQ = 10%) associated
with the code letter–AQL combinations of the MIL-STD-105E system. To use Tables T11.24
and T-21 to obtain a unique plan having the performance of an MIL-STD-105E scheme,
perform the following:

1. Decide if single, double, or multiple sampling is to be used.

2. Use Table T11.24 to obtain the appropriate MIL-STD-105E scheme operating ratio.
3. Use Table T11.21 to obtain the LTPD resulting from the use of the MIL-STD-105E
scheme.
4. Use this operating ratio and a value of p2 equal to the LTPD to determine a match-
ing individual plan from Appendix Table T6.1.

For example, if MIL-STD-105E specifies Code F, 2.5% AQL, a matching individual sam-
pling plan must have R  =  6.70 and LTPD = 12.2%, when the Poisson approximation to
Attributes Sampling Schemes 295

the binomial is used. A single, double, or multiple plan may be selected. Application of
Appendix Table T6.1 produces the following possibilities using plans 2S, 2D, and 2M.

Sample Sample
Sample Size Ac Re Sample Size Ac Re

Single 1 44 2 3 Multiple 1 12 # 2
2 12 0 3
3 12 0 3
Double 1 28 0 3 4 12 1 4
2 28 3 4 5 12 2 4
6 12 3 5
7 12 4 5

Any of these plans will provide scheme performance protection equivalent to the MIL-STD-
105E, Code F, 2.5% AQL scheme. It should be noted that the average sample size for the MIL-
STD-105E scheme at the AQL is about 22, while the single-sampling plan to match the scheme
has a sample size of 44. This illustrates the advantage of use of the switching rules, which are
incorporated in MIL-STD-105E. Also, use of the normal inspection plan alone, without the
switching rules, would result in an operating ratio of 10.96 and in considerably less consumer
protection than that of the scheme or of the plan derived earlier to match the scheme.
Appendix Table T11.25 may also be used in reverse to find an MIL-STD-105E sampling
scheme to match an individual sampling plan. The procedure is as follows:

1. Find the operating ratio of the individual plan.

2. Find the LTPD of the individual plan.
3. Locate the diagonal of Table T11.24 showing the operating ratios just less than or
equal to that of the given plan.
4. On the corresponding diagonal in Table T11.21, find the sample size–code letter
combination that has the desired LTPD for the MIL-STD-105E scheme.
5. Use this MIL-STD-105E scheme, with the switching rules, in lieu of the indi-
vidual plan.

For example, the plan n = 20, c = 2 has an operating ratio of 6.5 with 26.6% LTPD, using
the Poisson approximation to the binomial. Table T11.24 shows v ­ alues of R close to 6.5 on
the second diagonal. The second diagonal of Table T11.21 gives LTPD = 19.4 for Code E,
4.0 AQL, which is closest to that desired. Use of this code letter–AQL combination, with
the switching rules, will give an average sample size of about 15 at the AQL with the same
scheme performance as the plan n = 20, c = 2.
Occasionally, the acceptance criteria of an MIL-STD-105E plan must be altered to meet the
operating conditions. Suppose it is necessary to destructively sample 13 units under Code E, 1.0
AQL, normal inspection from a shipment of 84 units randomly packed a dozen to a box. The
units are to be resold and it is desirable to reduce the sample size to 12 so that 7 full cartons will
remain after sampling. To assess the effect of a sample size of 12, use the following procedure:

1. The original normal plan from MIL-STD-105E is n = 13, c = 0.

2. The operating ratio of the normal plan is R = 44.9 with p1 = .004 and p2 = .177.
296 Acceptance Sampling in Quality Control

3. If a sample size of 12 is used with the same acceptance number, we have from
Appendix Table T6.1
.0513
p1 = = .004
12

and
2.303
p2 = = .192
12

If the slight degradation in consumer protection can be tolerated by the consumer,

a switch to the plan n = 12, c = 0 may be reasonable.

Thus, since matched MIL-STD-105E criteria are used in the Schilling–Johnson (1980)
tables, they can be employed to assess the effect of any changes from the nominal sample
sizes given in that standard to other values made necessary by operating conditions or
to compensate for such changes. Individual sampling plans can also be derived to match
MIL-STD-105E scheme performance for use under conditions in which switching is dif-
ficult or impossible. These tables also provide unity values for a very low probability of
acceptance for use in reliability, safety, and compliance sampling. Sufficient values are
given to allow the OC and ASN curves to be evaluated as necessary.

MIL-STD-105 Derivatives
MIL-STD-105A was issued in 1950 with subsequent minor changes in MIL-STD-105B (1958)
and MIL-STD-105C (1961). In 1963, a major revision was undertaken resulting in MIL-
STD-105D. With some editorial changes, MIL-STD-105E was issued in 1989. Unfortunately,
the U.S. Department of Defense (DoD) discontinued the series of military standards on
February 27, 1995, with the objective of utilizing the civilian standards as cost savings.
In doing so, they canceled the premier acceptance sampling standard around the world.
Meanwhile, other standards-writing bodies developed spin-offs from the 105 series. Until
the discontinuation of MIL-STD-105E, these were largely simply copies, page for page, of
the current version of MIL-STD-105E. However, since the discontinuation of MIL-STD-
105E, these civilian standards have represented the 105 concept, keeping the probabilities
of acceptance as close as possible to 105E. These include American Society for Testing and
Materials (ASTM) International, the American Society for Quality (ASQ), ANSI, and ISO.
ASTM’s E11 Committee on Quality and Statistics actively converted many DoD military
acceptance sampling standards over several years led by the authors of this text. Those
standards will be discussed briefly in this text where appropriate. For companies that
have historically used these standards for inspection, the ASTM-converted standards, for
example, E2234, which replaces MIL-STD-105E, offer current refereed documents rather
than military standards that the DoD no longer supports.
ANSI and ASQ continue to offer the Z1 standards for attributes (Z1.4) and variables (Z1.9)
sampling plans. The U.S. Technical Advisory Group (TAG) to ISO’s Committee 69 on Statistical
Methods continues to lead the Z1 Committee that is responsible for the Z1.4 and Z1.9 stan-
dards, and both authors have led the effort to update these standards over several  years.
Attributes Sampling Schemes 297

The  U.S. TAG also contributes to the ISO acceptance sampling procedures for attributes
(ISO 2859) and variables (3951) sampling plans. Examples of these standards are as follows.

ANSI/ASQ Standard Z1.4

It is an American national standard with direct lineage to MIL-STD-105E. It is recom-
mended by the U.S. DoD as the replacement to MIL-STD-105E. It is best used in house and
in domestic transactions.

ASTM International Standard E2234

E2234 Standard Practice for Sampling a Stream of Product by Attributes Indexed by AQL is an
ASTM standard that contains and maintains the original MIL-STD-105E content as closely as
possible. This practice establishes lot or batch sampling plans and procedures for inspection
by attributes using MIL-STD-105E as a basis for sampling a steady stream of lots indexed
by AQL. This practice provides the sampling plans of MIL-STD105E in an ASTM format for
use by ASTM committees and others. It recognizes the continuing usage of MIL-STD-105E
in industries supported by ASTM. Most of the original text in MIL-STD-105E is preserved in
Sections 4 through 6 of this practice. It is intended to provide a source for use in conjunction
with ASTM and other standards, which directly referenced MIL-STD-105E. It is best used for
testing in a laboratory environment and with methodology in support of other standards.

ISO Standard 2859

It is an ISO international standard (1974a,b) that incorporates modifications of the original
MIL-STD-105 concepts, which reflect the state of the art. It is best used in international trade.
ISO 2859-1 has undergone substantial modification from MIL-STD-105E, while ASTM
E2234 and ANSI/ASQ Z1.4 have been subject to minor changes, none of which involved the
tables central to the operation of the system. Figure 11.9 summarizes some of these changes.
ISO has also developed a series of schemes in support of the AQL system in ISO 2859-1.
These include the following:

ISO 2859-1 “Sampling schemes indexed by acceptance quality limit (AQL) for lot-
by-lot inspection”—This is the ISO version of MIL-STD-105 and presents the basic
tables and subsidiary matter for the sampling system.
ISO 2859-2 “Sampling plans indexed by limiting quality (LQ) for isolated lot
inspection”—Procedures and tables are presented for sampling isolated lots.
ISO 2859-3 “Skip-lot sampling procedures”—Skip-lot sampling procedures are
­presented to be used with ISO 2859-1.
ISO 2859-4 “Procedures for assessment of declared quality levels”—This standard
­provides procedures, particularly useful in reviews and audits, for the assessment
of declared quality levels.
ISO 2859-5 “System of sequential sampling plans indexed by acceptance quality limit
(AQL) for lot-by-lot inspection”—Sequential sampling plans are given matching
the plans of ISO 2859-1, which allow the 2859-1 system to be applied using the
sequential plans.
ISO 2859-10 “Introduction to the ISO 2859 series of standards for sampling for
­inspection by attributes”—This is a general introduction to the ISO 2859 series and
provides insight into the application of the plans in the series.
298 Acceptance Sampling in Quality Control

Characteristics MIL-STD-105E ANSI/ASQ Z1.4 ASTM E2234 ISO 2859-1

Switch to
tightened consecutive lots consecutive lots consecutive lots consecutive lots
rejected rejected rejected rejected
Switch from Five consecutive Five consecutive Five consecutive Five consecutive lots
tightened lots accepted lots accepted lots accepted accepted
Switch to reduced 10 lots accepted 10 lots accepted and 10 lots accepted and Complicated
and passes limit passes limit passes limit switching score >30
number number number
Switch from Lot rejected Sample Lot rejected Sample Lot rejected Sample Lot rejected
reduced rejects in gap rejects in gap rejects in gap
between Ac and Re between Ac and Re between Ac and Re
Reduced table Gap between Ac Gap between Ac Gap between Ac Plans changed and
gap and Re and Re and Re gap eliminated
Discontinuation Five lots rejected on Five lots rejected on Five lots rejected on Five lots rejected on
of inspection tightened tightened tightened tightened
Terminology Defect Nonconformity Defect Nonconformity
Defective Nonconforming unit Defective Nonconforming item
Limiting quality Limiting quality Limiting quality Consumer’s risk
Quality
Defect Critical Group A Critical Class A
Major Group B Major Class B
Minor Group C Minor Class C
Arrows between Arrows only Arrows only Arrows only Arrows or fractional
Ac = 0 and Ac = 1 acceptance number
plans with
complicated
acceptance score
when lot size varies
Double Same as 105E Same as 105E Same as 105E Some plans changed
for better ASN
Multiple Seven stages Seven stages Seven stages
stages
AQL Acceptable quality Acceptance quality Acceptance quality Acceptance quality
level limit limit limit

FIGURE 11.9 
Differences between major AQL standards.

It should be noted that this series of standards is predicated on the concept of a flow of
lots, with the exception of Procedure A of ISO 2859-2. They assume a Type B sampling
­distribution utilizing the binomial and Poisson distributions accordingly.

MIL-STD-1916 (1996) DoD Preferred Methods for Acceptance of Product

MIL-STD-1916 is not a derivative of MIL-STD-105. Indeed, it presents a unique
approach utilizing a combination of an evaluation of the quality management system,
with the use of a statistical sampling plan as an alternative. A matched set of c = 0 attri-
butes, variables, and continuous sampling schemes is given, to be used with a set of
straightforward switching rules on a flow of lots. MIL-STD-1916 is addressed in detail
in Chapter 17.
Attributes Sampling Schemes 299

Further Considerations
The background of MIL-STD-105E and its development out of the 105 series are given
in an excellent paper by Pabst (1963). It explains some of the intricacies of the system
and its development. The theory behind the structure of the MIL-STD-105E tables is well
presented in a paper by Hill (1973). A detailed explanation of the procedural aspects of
the use of the system is given by Hahn and Schilling (1975). An extensive and informa-
tive investigation of the properties of MIL-STD-105E schemes is presented in a paper by
Stephens and Larson (1967). Scheme properties are also investigated by Schilling and
Sheesley (1978), and measures of performance are tabulated.
Based on his work with Torrey on continuous sampling plans, Dodge (1965) has pointed
out that the scheme OC curve resulting from the combination of two plans into a scheme
using the MIL-STD-105E normal–tightened (only) switching rules is easily determined.
Consider a normal plan N and a tightened plan T so combined. Then at a given proportion
defective with associated probabilities of acceptance PaN and PaT , respectively, the system
probability of acceptance is determined by calculating

2 - Pa4N
a=
(
(1 - PaN ) 1 - Pa4N )
1 - Pa5T
b=
(1 - PaT ) Pa5T

Then a/(a + b) represents the proportion of the time the plan will be on normal inspection and
b/(a + b) represents the proportion of time the plan will be on tightened inspection so that

aPaN + bPaT
Pa =
a+b

(Those familiar with CSP-2 will recognize a = fv, b = u, i = 5, k = 4.)

For example, if under Code F, 2.5 AQL, the reduced procedure is not used, the scheme
probability of acceptance at the AQL may be determined from the plans

N : n = 20 c = 1 Pa N = .9118

T : n = 32 c = 1 PaT = .8097
as
2 - .9118 4
a= = 48.05
(
(1 - .9118 ) 1 - .9118 4 )
5
1 - .8097
b= = 9.844
(1 - .8097 ) .8097 5
48.05 ( .9118 ) + 9.844 ( .8097 )
Pa = = .894
48.05 + 9.844

Note that the Dodge formula gives the scheme performance of MIL-STD-105E when only
the normal and tightened plans are used and reduced inspection is omitted.
300 Acceptance Sampling in Quality Control

Software Applications
Minitab
Minitab cannot generate MIL-STD-105E (or equivalent) single attributes sampling plans
but certainly can evaluate any given plan. Unfortunately, as stated earlier, Minitab cannot
generate or evaluate double or multiple attributes sampling plans.

Snap Sampling Plans!

Snap Sampling Plans! can only produce MIL-STD-105E (ANSI Z1.4) single-sampling
plans. As an example, suppose the producer and the consumer agree on an AQL of
2.5% and lot sizes are expected to be N = 100. Inspection Level II will be used since
no other inspection level was agreed upon. Using the lot size of 100 and Inspection
Level  II, the sample size code letter table in MIL-STD-105E (Appendix Table T11.2)
specifies Code F.
From the main menu, choose the Mil Std 105/ANSI Z1.4 button. Figure 11.10 shows the
completed dialog box. Click on the Sample Plan button to generate the plan. The single-
sampling plan is presented in Figure 11.11. The sampling plan is n = 20, c = 1, r = 2.

FIGURE 11.10 
Snap Sampling Plans! dialog box to generate a MIL-STD-105E attributes single-sampling plan.
Attributes Sampling Schemes 301

MIL-STD-105/ANSI Z1.4
Sampling plan

Lot size Inspection level Sampling level AQL %

100 II Normal 2.500

Results Sample sizea Accept Reject

20 1 2

Accept number is the number of allowed rejects.

If the inspection sample has >= the reject number, then reject the lot.

For “reduced” sampling levels, if the actual number of defects from the
inspection sample is greater than the accept number but less than the
reject number, then accept the lot and return to “normal sampling level.”
a
If the sample size is greater than the lot size then conduct 100%
inspection.

FIGURE 11.11 
Snap Sampling Plans! MIL-STD-105E attributes single-sampling plan.

Statgraphics
Statgraphics has the ability to produce MIL-STD-105E (ANSI Z1.4/ISO 2859) single,
double, or multiple attribute sampling plans with OC and ASN curves. Suppose the pro-
ducer and the consumer agree on an AQL of 2.5% and lot sizes are expected to be N = 100.
Inspection Level II will be used since no other inspection level was agreed upon. Using the
lot size of 100 and Inspection Level II, the sample size code letter table in MIL-STD-105E
(Appendix Table T11.2) specifies Code F. The matching single, double, and multiple plans
were determined to be

Tightened Normal Reduced

Code F, 2.5 AQL n Ac Re n Ac Re n Ac Re

Single 32 1 2 20 1 2 8 0 2
Double 20 0 2 13 0 2 5 0 2
20 1 2 13 1 2 5 0 2
Multiple 8 # 2 5 # 2 2 # 2
8 # 2 5 # 2 2 # 2
8 0 2 5 0 2 2 0 2
8 0 3 5 0 3 2 0 3
8 1 3 5 1 3 2 0 3
8 1 3 5 1 3 2 0 3
8 2 3 5 2 3 2 1 3

In order to generate the single-sampling plan under normal inspection, choose SPC >
Acceptance Sampling > Attributes > MIL-STD-105E (ANSI Z1.4/ISO 2859) and complete
the dialog box as shown in Figure 11.12 and click OK. Select all the graph types on the
Tables and Graphs dialog box as shown in Figure 11.13 and click OK.
302 Acceptance Sampling in Quality Control

FIGURE 11.12 
Statgraphics dialog box to generate a MIL-STD-105E attributes single-sampling plan.

FIGURE 11.13 
Statgraphics dialog box to select graph types.

MIL-STD-105E (ANSI Z1.4/ISO 2859)

Attribute: Percent nonconforming
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Normal
Sampling plan: Single (Code F)

Stage Sample Size Acceptance Number Rejection Number

1 20 1 2
Attributes Sampling Schemes 303

The StatAdvisor
The derived sampling plan has a single stage. Take a sample of 20 units. If the number of
nonconforming units is no more than 1, accept the lot. If the number of nonconforming units
is 2 or more, reject the lot.

The type of inspection can be changed in the dialog box in Figure 11.12 to tightened and
reduced to produce those plans.

MIL-STD-105E (ANSI Z1.4/ISO 2859)

Attribute: Percent nonconforming
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Reduced
Sampling plan: Single (Code F)

Stage Sample Size Acceptance Number Rejection Number

1 8 0 2

The StatAdvisor
The derived sampling plan has a single stage. Take a sample of 8 units. If the number of
­nonconforming units is no more than 0, accept the lot. If the number of nonconforming units
is 2 or more, reject the lot. Otherwise, accept the lot but return to normal inspection.

MIL-STD-105E (ANSI Z1.4/ISO 2859)

Attribute: Percent nonconforming
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Tightened
Sampling plan: Single (Code F)

Stage Sample Size Acceptance Number Rejection Number

1 32 1 2

The StatAdvisor
The derived sampling plan has a single stage. Take a sample of 32 units. If the number of
nonconforming units is no more than 1, accept the lot. If the number of nonconforming units
is 2 or more, reject the lot.

Let’s generate the double-sampling plan under normal inspection. The reduced and
tightened plans will not be shown here as the process is similar to that used for single-
sampling plans. Choose SPC > Acceptance Sampling > Attributes > MIL-STD-105E (ANSI
Z1.4/ISO 2859) and complete the dialog box as shown in Figure 11.14.
304 Acceptance Sampling in Quality Control

FIGURE 11.14 
Statgraphics dialog box to generate a MIL-STD-105E attributes double-sampling plan.

MIL-STD-105E (ANSI Z1.4/ISO 2859)

Attribute: Percent nonconforming
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Normal
Sampling plan: Double (Code F)

Stage Sample Size Cumulative Size Acceptance Number Rejection Number

1 13 13 0 2
2 13 26 1 2

The StatAdvisor
The derived sampling plan has two stages. In the first stage, take a sample of 13 units. If the
number of nonconforming units is no more than 0, accept the lot. If the number of noncon-
forming units is 2 or more, reject the lot. Otherwise, take a second sample of 13 units. If the
total number of nonconforming units in both samples is no more than 1, accept the lot. If
the total number of nonconforming units in both samples is 2 or more, reject the lot.

Finally, generate the multiple-sampling plan under normal inspection. Again, the
reduced and tightened plans will not be shown here as the process is similar to that used
for single-sampling plans. Choose SPC > Acceptance Sampling > Attributes > MIL-STD-
105E (ANSI Z1.4/ISO 2859) and complete the dialog box as shown in Figure 11.15.
Attributes Sampling Schemes 305

FIGURE 11.15 
Statgraphics dialog box to generate a MIL-STD-105E attributes multiple-sampling plan.

MIL-STD-105E (ANSI Z1.4/ISO 2859)

Attribute: Percent nonconforming
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Normal
Sampling plan: Multiple (Code F)

Stage Sample Size Cumulative Size Acceptance Number Rejection Number

1 5 5 Acceptance not permitted 2
2 5 10 Acceptance not permitted 2
3 5 15 0 2
4 5 20 0 3
5 5 25 1 3
6 5 30 1 3
7 5 35 2 3

The StatAdvisor
The derived sampling plan has seven stages. At each stage, take a sample of 5 units. If the
number of nonconforming units in all samples is no more than the acceptance number for
that stage, accept the lot. If the number of nonconforming units in all samples is equal to or
greater than the rejection number for that stage, reject the lot. Otherwise, move to the next
stage and take another sample.
306 Acceptance Sampling in Quality Control

Problems
1. MIL-STD-105E 1.0% AQL is specified and a lot of 390 pieces is to be inspected. Find
the associated set of normal, tightened, and reduced plans for
a. Single sampling
b. Double sampling
c. Multiple sampling
2. The exact AOQL for the scheme represented in Problem 1 is 0.93%. What is the AOQL
of the constituent?
a. Normal plan
b. Tightened plan
What does this suggest as a rough measure of the AOQL of the tightened plan? Of the
scheme?
3. What is the LQ for Pa = 10% for the tightened and normal plans of Problem 1?
a. Percent defective
b. Defects per 100 units
4. Which type of plan (single, double, multiple) gives minimum average sample size at
the AQL for the tightened plan of Problem 1?
5. What action should be taken if, after a switch, the sixth lot is the first lot rejected
(and the switch was to)?
a. Normal inspection
b. Tightened inspection
c. Reduced inspection
6. For the scheme, Code F, 4.0 AQL, what are the following properties of the scheme for
defects per 100 units?
a. Probability of acceptance at the AQL
b. ASN at the AQL
c. AOQL
d. Average total inspection for lots of size 120 at the AQL
7. What is the probability of having a succession of 10 lots rejected on tightened inspec-
tion after a switch is made if the process is running?
a. The indifference quality level of the tightened plan
b. The LTPD of the tightened plan
8. The reduced plan for Code C, 10% AQL is n = 2, Ac = 0, Re = 2. What is the probabil-
ity of simultaneously accepting a lot but switching back to normal inspection if the
producer’s process is running at 10% defective?
9. The sample sizes in MIL-STD-105E are in a geometric progression with a ratio 101/5.
What would be the next single sample size after S in the tightened table if Code T
were added? What would be the acceptance number for Code T and 0.015 AQL?
What would be the approximate AOQL?
Attributes Sampling Schemes 307

10. A contract requires MIL-STD-105E, 4.0 AQL. A single isolated lot of size N = 140 is to
be inspected. Derive a single-sampling plan that will match the performance of the
MIL-STD-105E scheme specified by the contract. What MIL-STD-105E scheme will
afford performance equivalent to the plan n = 14, c = 3?

References
American Society for Testing and Materials, Standard Practice for Sampling a Stream of Product by
Attributes Indexed by AQL, ASTM Standards E2234, Vol. 14.02, West Conshohocken, PA.
Burnett, T. L., 1967, Markov Chains and attribute sampling plans, IBM Technical Report No. 67-825-
2175, IBM Federal Systems Division, Oswego, NY.
Calvin, T. W., 1977, TNT zero acceptance number sampling, in American Society for Quality Control
Technical Conference Transactions, Philadelphia, PA.
Dodge, H. F., 1965, Evaluation of a sampling inspection system having rules for switching between
­normal and tightened inspection, Technical Report No. 14, The Statistics Center, Rutgers—The
State University, New Brunswick, NJ.
Hahn, G. R. and E. G. Schilling, 1975, An introduction to the MIL-STD-105D acceptance sampling
scheme, ASTM Standardization News, 3(9): 20–30.
Hill, I. D., 1973, The design of MIL-STD-105D sampling tables, Journal of Quality Technology, 5(2): 80–83.
International Organization for Standardization, 1974a, Sampling Procedures and Tables for Inspection
by Attributes (ISO 2859), International Organization for Standardization (ISO), Geneva,
Switzerland.
International Organization for Standardization, 1974b, A Guide to the Use of ISO 2859 “Sampling
Procedures and Tables for Inspection by Attributes”, (ISO 3319), International Organization for
Standardization (ISO), Geneva, Switzerland.
International Organization for Standardization, 2006, Statistics–Vocabulary and Symbols—Part 2,
Applied Statistics, (ISO 3534-2), International Organization for Standardization (ISO), Geneva,
Switzerland, p. 70.
Mentch, C. C., 1980, Manufacturing process quality optimization studies, Journal of Quality Technology,
12(3): 119–129.
Pabst, W. R., Jr., 1963, MIL-STD-105D, Industrial Quality Control, 20(5): 4–9.
Romboski, L. D., 1969, An investigation of quick switching acceptance sampling systems, PhD dis-
sertation, Rutgers—The State University, New Brunswick, NJ.
Schilling, E. G. and L. I. Johnson, 1980, Tables for the construction of matched single, double, and
multiple sampling plans with application to MIL-STD-105D, Journal of Quality Technology, 12(4):
220–229.
Schilling, E. G. and J. H. Sheesley, 1978, The performance of MIL-STD-105D under the switching
rules, Journal of Quality Technology, Part 1, 10(2): 76–83.
Stephens, K. S. and K. E. Larson, 1967, An evaluation of the MIL-STD-105D system of sampling
plans, Industrial Quality Control, 23(7): 310–319.
United States Department of Defense, 1965, Guide to Sampling Inspection, Quality and Reliability
Assurance Handbook (MIL-HDBK-53), Office of the Assistant Secretary of Defense (Installations
and Logistics), U.S. Department of Defense, Washington, DC.
United States Department of Defense, 1989, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
12
Variables Sampling Schemes

Sampling schemes are not restricted to attributes. They may be composed of variables
plans as well. Thus, it was Military Standard 414 (MIL-STD-414) that was issued on June 11,
1957. It has since become a classic companion standard to MIL-STD-105 and has been used
throughout the world.
The protection afforded by this standard is roughly matched to MIL-STD-105A.
However, modifications in the tables incorporated in the MIL-STD-105D version upset
the match somewhat. Commander Gascoigne of the British Navy showed how to restore
the ­balance, and his simple method has been incorporated into civilian sampling systems.
The MIL-STD-414 sampling system will be discussed in depth here as an example of a clas-
sic variables system, and its relation to other systems will be indicated.
On April 1, 1996, the Department of Defense reentered standard development with
MIL-STD-1916, which contains variables, attributes, and continuous sampling together
with process control. It is presented in depth in Chapter 17 and will also be discussed here.

MIL-STD-414
Unlike MIL-STD-105E, MIL-STD-414 is a sampling system utilizing variables inspection.
It was devised by the military, as a consumer, to be used to assess the percent defective
beyond contractual limits. Since it is a sampling system, it incorporates switching rules
to move from normal to tightened or reduced inspection and return to achieve consumer
protection. These switching rules must be used if the standard is to be properly applied.
The switching rules differ somewhat from those used in MIL-STD-105E. The standard
assumes underlying normality of the distribution of the measurements to which it is
applied and is intended to be used with a steady stream of lots.
MIL-STD-414 allows for the use of three alternative measures of variability: known stan-
dard deviation (σ), estimated standard deviation (s), or average range of subsamples of
five (R). If the variability of the process producing the product is known and stable, it is
profitable to use σ. The choice between s and R when σ is unknown is an economic one.
The range requires larger sample sizes but is easier to compute and understand. Operating
characteristic (OC) curves given in the standard are based on the use of s, the σ, and the R
plans having been matched as closely as possible to those using s.
The basic statistic to be calculated in applying MIL-STD-414 may be considered to be
the standardized distance from the sample mean to the specification limit. For an upper
specification limit U, when σ is known

U-X
tU =
s

309
310 Acceptance Sampling in Quality Control

When σ is unknown

U -X
tU =
s

or

U-X
tU =
R

which is substituted depending on the measure of variability chosen. A comparison of

tU with the acceptance constant k will show whether the sample mean is in the region
of acceptance.
MIL-STD-414 offers an alternate procedure to use the acceptance constant k, the M
method discussed in Chapter 10. This involves using a statistic similar to those previously
mentioned to estimate proportion defective in the lot and is referred to in the standard as
Form 2. The k method, involving a simple comparison of t with k to determine the accept-
ability, is called Form 1. Form 2 is the preferred procedure since the switching rules cannot
be applied unless the fraction defective p̂ of each lot is estimated from the sample.
MIL-STD-414 is complex. It consists of sections indexed by measure of variability, type
of specification (single or double), and form number of the acceptance procedure. Only
Form 2 is officially available for the case of double specification limits. The standard’s
structure is shown in Figure 12.1.

MIL-STD-414

Section A Section B Section C

(Variable unknown) Section D
General Standard deviation (Variable unknown)
Variable known
description method Range method
Criteria T – R*, MAR
Criteria T – R*, MSD
Double specification

Double specification

Double specification
Single specification

Single specification

Single specification

Criteria T – R*
Estimating p

Estimating p

Estimating p
Form 1

Form 2

Form 2

Form 1

Form 2

Form 2

Form 1

Form 2

Form 2
M

M
k

FIGURE 12.1
Content of MIL-STD-414. (Asterisk [*] indicates tables for estimating p and criteria for tightened and reduced
inspection.) (Reprinted from Schilling, E.G., Qual. Prog., 7(5), 19, 1974. With permission.)
Variables Sampling Schemes 311

Application of MIL-STD-414 follows the pattern of MIL-STD-105E. Note that MIL-

STD-414 and MIL-STD-105E plans are not matched. The classification of defects used in
MIL-STD-414 is the same as that used in MIL-STD-105E: critical, major, and minor. Sample
sizes are determined from lot size and acceptable quality level (AQL), and after choosing
the measure of variability to be used and the form of acceptance procedure, appropriate
acceptance constants are obtained from the standard.
MIL-STD-414 has a liberal supply of excellent examples. The reader should refer to the
standard for detailed numerical examples of its application.
The necessary assumption of a known, stable underlying normal distribution of
individual measurements inherent in the MIL-STD-414 variables plans is a serious
­
­limitation in their application. Use of MIL-STD-414 plans without investigating the true
nature of the underlying distribution is foolhardy, for the results can be very bad indeed.
Nonetheless, sensible evaluation of the nature of the underlying distribution and imple-
mentation of prudent procedures to ensure stability can provide sufficient ­justification
for the use of MIL-STD-414. This is particularly true for in-process and final inspection
where the distribution of the process producing the product is not beyond the control or
investigation of those applying the plan. The rewards for painstaking, thorough analysis
are great in terms of sample size and worthwhile information on the process involved.

Operation
Since it is a sampling system, proper use of MIL-STD-414 requires diligent use of the
switching rules. It is with this procedure that protection is afforded by both the producer
and the consumer through tightening and relaxing the severity of inspection consistent
with the demonstrated performance of the producer. The operation of the switching rules
is shown schematically in Figure 12.2.
An MIL-STD-414 sampling scheme always starts on normal inspection, which is contin-
ued until a switch to tightened or reduced inspection is warranted. Normal inspection is
reinstituted when the conditions justifying tightened or reduced inspection can be shown
to apply no longer. The switching rules of MIL-STD-414 are such that the probability of
switching from normal to tightened or reduced inspection, respectively, is <.005 when
quality is running at the level of the AQL.
An important part of the switching procedure is the estimated percent defective in each
lot, obtained using the M method of Chapter 10 from the tables given as part of Form 2 of
the standard. The estimated process average, which is the mean of these percent defective,
is also employed in switching.
A switch to tightened inspection involves changing the acceptance criterion to the
next lower AQL category, while retaining the sample size associated with the code letter
involved. This leads to decreased consumer’s risk at the expense of an increase in the pro-
ducer’s risk. Tightened inspection is instituted under the following conditions:

1. More than T of the last 10 lots (or such other number of lots as designated) have
estimates of percent defective, obtained through the use of the M method (Form 2),
exceeding the AQL.
2. The process average obtained from the estimated percent defective of the last
10 lots (or such other number of lots as designated) is greater than the AQL.
312 Acceptance Sampling in Quality Control

Start

Normal
inspection

No >T of prior 10
lots have pˆ > AQL
Yes

Production steady No
No p > AQL
Yes
Yes
Prior 10 lots passed
No under normal
Yes Tightened
pˆ L of each lot < lower inspection
No limit for reduced
Yes

p > AQL
Yes No
Reduced
inspection pˆ = Est. % defective for one lot
p = Avg. % defective for last
10 lots (or as designated)

Previous lot passed

No
Yes
p ≤ AQL
No
Yes
Production steady
No Yes

FIGURE 12.2
MIL-STD-414 normal–tightened–reduced. (Reprinted from Schilling, E.G., Qual. Prog., 7(5), 18, 1974. With
permission.)

Values of T are given in each of the sections (see Figure 12.1) for application against the
last 5, 10, or 15 lots, which have been designated. If the sample size code letter is not the
same for all the previous lots, the table of T is entered using the code letter of the smallest
sample size involved. As an example, MIL-STD-414 Table B.6, which gives T values for the
standard deviation section, is presented as Appendix Table T12.1.
Normal inspection is reinstated from tightened when
The estimated process average of the last 10 lots (or such other number of lots as desig-
nated) is equal to or less than the AQL.

A switch to reduced inspection involves changing both the sample size and the acceptance
criteria to obtain a reduction in the sample size. This reduction is typically around 40%.
Variables Sampling Schemes 313

The producer’s risk is decreased slightly thereby, while the consumer’s risk is increased.
Reduced inspection is instituted when

1. Production is at a steady rate

2. The preceding 10 lots (or such other number of lots as designed) have been accepted
under normal inspection
3. The estimated percent defective for each of the preceding 10 lots (or such other
number of lots as designated) is less than the applicable lower limit number tabu-
lated, or, for certain plans having small sample size and low AQL, the estimated
lot percent defective must be zero for a specified number of consecutive lots

Values of the lower limit number (or number of consecutive lots) are given in each of the
sections (see Figure 12.1) for application against the preceding 5, 10, or 15 lots, whichever
has been designated. As an example, MIL-STD-414 Table B.7 showing the limit numbers
for reduced inspection is presented in Appendix Table T12.2.
Normal inspection is reinstated from reduced when

1. A lot is rejected
2. The estimated process average from the last 10 lots (or such other number of lots
as designated) is greater than the AQL
3. Production becomes irregular or delayed
4. Other conditions warrant that normal inspection should be instituted

These switching rules are somewhat more complicated than those of MIL-STD-105E and
are patterned after those used in MIL-STD-105A. Nevertheless, their use is economically
effective in reducing sample size with increased protection over that which could be
achieved by use of single plans alone.
The process average is defined as the average percent defective, based upon a group of
lots submitted for original inspection. It is constructed using estimates of percent defective
from a specific number of preceding lots from the first submissions only. A product known
to have been produced under atypical conditions is excluded from the estimated process
average. Normally, it is computed as the arithmetic mean of the estimated percent defec-
tive from the last 10 lots unless some other number of lots has previously been designated.

Selection
The selection of a set of plans for normal, tightened, and reduced inspection is more com-
plicated in MIL-STD-414 than in MIL-STD-105E in that MIL-STD-414 offers complete sets
of plans and procedures for each of the three methods for estimated variability. In fact,
MIL-STD-414 could easily be separated into three self-contained standards each based on
its own measure of variability, ŝ. As seen in Figure 12.1, they are the following:

Standard deviation method (Section B)

( )
2
å X-X
ŝ = s =
n -1
314 Acceptance Sampling in Quality Control

Range method (Section C)

R
sˆ =
d2*

Variability known (Section D)

ŝ = s

Section A applies to each measure of variability and presents a general description of the
sampling plans, gives AQL ranges to be covered in the standard, supplies sample size code
letters, and presents OC curves.
Obtaining a plan from MIL-STD-414 involves more than selection of a measure of
­variability, however. The sequence for selection of a set of plans is given in Figure 12.3.

Start

Check normality of
individual measurements

Determine lot
size and AQL

Determine form
of acceptance procedure

Form 1 Form 2
(k method) (M method)

Decide measure Decide measure

of variability of variability

σ s R σ s R

Select plan Select plan Select plan Select plan Select plan Select plan
as in as in as in as in as in as in
Section D Section B Section C Section D Section B Section C

Check OC curves to
ensure protection
is as desired

Apply plan

Apply
switching
rules

FIGURE 12.3
Check sequence for selecting a plan from MIL-STD-414. (Reprinted from Juran, J.M., Quality Control Handbook, 5th ed.,
McGraw-Hill, New York, 1999. With permission; Section 25, Sampling by Variables by E.G. Schilling, pp. 25–15.)
Variables Sampling Schemes 315

First, the underlying distribution of measurements to which the plan is to be applied

should be checked for normality. This involves probability plots, statistical goodness-of-fit
tests, control charts, and other statistical procedures as appropriate. MIL-STD-414 assumes
a normal distribution of measurements, and this assumption needs to be constantly veri-
fied during application of the standard because the central limit theorem does not apply.
As in the application of MIL-STD-105E, the lot size and AQL must be determined. If the
AQL chosen is not one of those used to index MIL-STD-414 plans, Table A.1 of Section A
allows for conversion to one of the specified values. Fortunately, the AQL values given
in MIL-STD-414 are the same as those used in MIL-STD-105E. As in MIL-STD-105E, lot
size is used to determine the sample size code letter using Table A.2 of Section A. Five
inspection levels are given. Since assignment of code letters to lot size ranges is primarily
based on economic and other nonstatistical considerations in Type B sampling situations,
the alternative inspection levels in both systems provide some flexibility in this regard.
Unlike MIL-STD-105E, inspection level IV is normally used unless some other inspection
level is specified. Also, the lot size ranges are not the same in the two standards. Tables A.1
and A.2 of MIL-STD-414 are reproduced here in the Appendix as Tables T12.3 and T12.4,
respectively.
The choice between the k method (Form 1) and the M method (Form 2) is an important
initial decision. These methods are described in Chapter 10. For single specification limits,
Form 1 is more straightforward. Values of the maximum standard deviation (MSD) and
maximum allowable range (MAR) are provided in the standard for use of the k method
with double specification limits. These are useful when a plan is to be plucked out of
the standard to be used singly and not as part of an MIL-STD-414 scheme. However,
MIL-STD-414 recognizes only the M method (Form 2) when the standard is to be applied
to double specification limits. Furthermore, the switching rules cannot be used unless
Form 2 tables and procedures are used. Therefore, Form 2 is to be recommended for use
with an MIL-STD-414 scheme if only for reasons of economy of effort. Form 1 is easier to
explain and administer, however, and the associated sampling plans are recommended
if individual plans are to be taken from the standard and used not as MIL-STD-414, but
separately as individual variables plans for proportion nonconforming.
It is, of course, necessary to select a measure of variability to be used. If control charts
have confirmed the existence and consistency of a known standard deviation, variability
known (Section D) will be the most economic source of sampling plans. If the standard
deviation is unknown (but the distribution shape consistently stays normal), plans using
the standard deviation method (Section B) or the range method (Section C) may be chosen.
The range plans are easier to explain, calculate, and understand than those using sample
standard deviation but are also less efficient. Calculators and computers have facilitated
the computation of the standard deviation. The choice should be made in keeping with the
sampling situation and the competence of the inspection personnel.
Having made these decisions, the specific set of normal, tightened, and reduced plans is
selected from the standard. Vertical arrows shown in the table are used in the same man-
ner as those in MIL-STD-105E. Figure 12.4 from Schilling (1974) shows how the plans are
implemented once they have been selected. The sample size and acceptance criteria are
obtained from appropriate tables. The statistics associated with a specific plan are then
computed. These are listed in Figure 12.4 and are explained by worked examples in each
section of MIL-STD-414. The statistic is compared directly to the acceptance criteria in the
manner of the k method when using Form 1. Form 2 requires estimation of the percent
defective in the lot to obtain the percent estimated to be above an upper specification limit
pU or below a lower limit pL. This is done from special tables. The total estimated percent
316 Acceptance Sampling in Quality Control

Step Section Form 1 Form 2

Preparatory — Obtain k and n from appropriate Obtain M and n from appropriate
tables tables
Determine Section B (s) U–x U– x
zU = QU =
criteria s s
x –L x–L
zL = QL =
s s
Section C (R) U–x (U – x )c
zU = QU =
R R
x–L ( x – L)c
zL = QL =
R R
Section D (σ) U– x (U – x )v
zU = QU = σ
σ
x–L ( x – L)v
zL = QL =
σ σ
Estimation — — Enter table with n and QU or QL
to get pU or pL
Action Single specification Accept if zU ≥ k or zL ≥ k Accept if pU ≤ M or pL ≤ M
Double specification Accept if* zU ≥ k, zL ≥ k Accept if pU + pL ≤ M
–
And s ≤ MSD or R ≤ MAR

n
c = scale factor v=
n –1

FIGURE 12.4
Application of MIL-STD-414. (Reprinted from Schilling, E.G., Qual. Prog., 7(5), 20, 1974. With permission.)

defective is then compared to M taken from the Form 2 table for acceptance or rejection
when applied to double specification limits.
Detailed examples of the selection of plans from MIL-STD-414 and their operation are
given in later sections of this chapter.

Measures
Only the OC curves are given as measures of the plans contained in MIL-STD-414. These
are for individual plans and not for the scheme as a whole. Since the plans for the standard
deviation, range, and variability known methods are matched, and the k and M methods are
equivalent for single specification limits, only one set of OC curves is given. These are for the
standard deviation method. The others are assumed sufficiently well matched to be repre-
sented by those shown. The OC curve of the plan Code F, 2.5% AQL is shown in Figure 12.5.
The OC curves of MIL-STD-414 may be used to select individual plans to be used out-
side the MIL-STD-414 sampling system. In this case, the OC curve desired is found and
the acceptance criteria determined from the corresponding table of normal plans. In no
sense should the resulting plan be referred to as an MIL-STD-414 plan since MIL-STD-414
implies full use of the sampling system based on the switching rules. Nevertheless, it is a
natural compendium of variables plans for proportion nonconforming and can be used to
effectively select individual plans for special applications.
Variables Sampling Schemes 317

100
90
% of lots expected to be accepted

80
70
60
50
40
1.0 1.5 2.5 4.0 6.5 10.0 15.0
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Quality of submitted lots (in percent defective)

FIGURE 12.5
MIL-STD-414 OC curve for Code F, standard deviation method. (Reprinted from U.S. Department of Defense
MIL-STD-414, Table A.3, p. 11, 1957. With permission.)

Implementation of Form 2
Implementation of MIL-STD-414 is best shown by example. Since Form 2, the M method,
is the preferred procedure in that the switching rules are based on its estimates, it
will be presented first. Also, the standard deviation method is used since the range
method and variability known involve only slight modifications of the procedure (see
Figure 12.4). Double specification limits are shown since the single specification limit
procedure follows from that given. Consider the following example, adapted from MIL-
STD-414 (p. 69).
The specification for electrical resistance of a certain electrical component is 650.0 ± 30 Ω.
A lot of 100 items is submitted for inspection with AQL = 2.5% for the upper and AQL = 1%
for the lower specification limits. Suppose the values of sample resistances are as follows:
643, 651, 619, 627, 658, 670, 673, 641, 638, and 650.
Assume that the electrical resistances of this device have been shown to be normally
distributed. In fact, these data were plotted on normal probability paper in Chapter 3.
The sample size code letter table (Table T12.4) shows Code F to apply under inspection
level IV, which is used unless some other level is specified. The master table for normal
and tightened inspection (Table T12.5) and the master table for reduced inspection (Table
T12.6) give the following criteria for the plans involved in the MIL-STD-414 sampling
scheme:

1% AQL 2.5% AQL

Tightened n = 10, M = 2.17 n = 10, M = 4.77

Normal n = 10, M = 3.26 n = 10, M = 7.29
Reduced n = 4, M = 5.50 n = 4, M = 16.45
318 Acceptance Sampling in Quality Control

Since the scheme starts on normal inspection, we will illustrate application of the n
­ ormal
plan. We find

X = 647 s = 17.2

The quality indices are

QL =
( X - L ) = 647 - 620 = 1.57
s 17.2

QU =
( U -X ) = 680 - 647 = 1.92
s 17.2

Using the table for estimating the lot percent defective (Table T12.7), the values of QL and
QU and the sample size are cross tabulated to give estimates of percent defective of

Lower specification: pL = 4.92

Upper specification: pU = 1.62

Overall: p = pL + pU = 6.54

These estimates are compared against the respective critical values of M to obtain

pL = 4.92 > ML = 3.26 Reject

pU = 1.62 < MU = 7.29 Accept

p = 6.54 < MU = 7.29 Accept

The lot is rejected since the lower estimated percent defective does not meet the acceptance
criterion M. When there are different AQLs for the lower and upper specifications, it is neces-
sary to test the upper and lower estimated percent defective separately. Furthermore, the total
estimated percent defective p is tested against the value of M associated with the larger AQL.
These estimates of percent defective are slightly smaller than those obtained using
the probability plot of Chapter 4 due to estimation by the minimum variance unbiased
technique. This procedure is somewhat more complicated than the procedure required if
both specification limits had the same AQL. Suppose the AQL of 1.0% is applied to both
specification limits. Then the acceptance procedure would simply be to compute the
total estimated percent defective and compare it to the value of M for 1.0% AQL. That is,

p = 6.54 > M = 3.26

and the lot would be rejected. The OC curve for such a plan will be found in Figure 12.5
labeled 1.0% AQL. The OC curves for different AQLs on the lower and upper specification
limits are not given since they would depend upon the split of percent defective between
the specifications.
Having rejected the lot, the estimated value of p = 6.54 would be entered into the process
average to be utilized in the switching procedure.
Variables Sampling Schemes 319

Implementation of Form 1
As an illustration of Form 1, a single specification limit will be used since the standard
does not advocate use of the k method with double specification limits. However, if a Form
1 plan is to be used with double specification limits, the MSD can be found from the table
of values of F for MSD (Table T12.10). It gives values of F, which are used to compute the
MSD as

MSD = F (U - L )

where U and L are the upper and lower specification limits. The plan is applied to each
specification limit separately if s ≤ MSD. Of course, the lot is rejected if s > MSD. Consider
the following example, adapted from MIL-STD-414 (p. 69).
The specification for minimum electrical resistance of a certain electrical component
is 620 Ω. A lot of 100 items is submitted for inspection with an AQL = 1.0%. Suppose
the values of sample resistances are as follows: 643, 651, 619, 627, 658, 670, 673, 641, 638,
and 650.
Assume that the electrical resistances of this component have been shown to be normally
distributed. The sample size code letter table (Table T12.4) shows Code F to apply using
inspection level IV, which is used unless some other level is specified. The master tables
for normal and tightened inspection (Table T12.8) and the master table for reduced inspec-
tion (Table T12.9) give the following criteria for the plans involved in the MIL-STD-414
sampling scheme, Form 1:

Tightened: n = 10 , k = 1.84

Normal: n = 10 , k = 1.72

Reduced: n = 4, k = 1.34

The switching rules begin with normal inspection, and so the normal inspection plan will
be illustrated here. In application of the normal plan, we have

X = 647 s = 17.2

tL =
( X - L ) = 647 - 620 = 1.57
s 17.2

and since 1.57 < 1.72, the lot is rejected. The MSD is, of course, not used with single speci-
fication limits.
To use the switching rules, the estimated percent defective must be determined. This is
done using the Form 2 criteria and tables under the M method. From the previous example
of implementation of Form 2, we have pL = 4.92% with the same data. This is the value that
would be entered into the computations of the process average and compared to appropri-
ate criteria for application of the switching rules.
320 Acceptance Sampling in Quality Control

Implementation of Plans for Range and Variability Known

Implementation of Forms 1 and 2 under the range method or variability known is very much
like that under the standard deviation method. The principal change is in the statistic to be
compared to the acceptance criteria. For Form 1, the statistic remains essentially the same as
that for the standard deviation method with R or σ substituted for s (see Figure 12.4). When
Form 1 is to be used with double specification limits, the MAR, which serves the same pur-
pose as the MSD, is calculated from f factors given in the standard to obtain

MAR = f (U - L )

in a manner similar to the procedure used with s. For Form 2, the statistic is changed by
the addition of a constant so that R/c or σ/v is substituted for s (see Figure 12.4). Here,

n
c = d2* v =
n -1

where d* is the adjusted d2 factor developed by Duncan (1955). This is necessary in order
to obtain the minimum variance unbiased estimate of p characteristic of the M method.
In all range plans, R is the average range of subsamples of 5 in the sample of n. Units
are assigned to the subsamples in the order in which they are drawn (assuming random
sampling). Naturally, for small samples not divisible by 5, the range of the full sample is
used (i.e., 3, 4, and 7).
Aside from these changes, the procedures for implementation of plans for the three mea-
sures are essentially the same.

Match between MIL-STD-414 and MIL-STD-105E

In 1976, the American National Standards Institute (ANSI) Committee Z-1 on Quality
Assurance recommended that a revision of the ANSI version of MIL-STD-414 be made
incorporating some of the suggestions made by Gascoigne (1976) resulting from his
work on British Defense Standard (05-30/1) and with the International Organization for
Standardization (ISO). Principal among these was a method for adjusting the code letter of
the ANSI version of MIL-STD-414 to make its OC curves roughly match those of the ANSI
version of MIL-STD-105D (ANSI Z1.4 1971) at the adjusted code letter and AQL. Revision of
the ANSI version of MIL-STD-414 (ANSI Z1.9) was accomplished by the American Society
for Quality Control Standards Group, and it now appears as ANSI Z1.9 (2003). Table 12.1
shows the match between the revised ANSI Z1.9 (2003) code letter, the MIL-STD-414 code
letter, and the corresponding code letter of MIL-STD-105D. (A comparison with MIL-STD-
105E would be identical.) Plans with these code letters are roughly matched and will allow
switching between variables and attributes plans within the code letters shown at a given
AQL. To preserve the match, MIL-STD-414 AQLs 0.04, 0.065, and 15.00 should not be used
and were dropped from ANSI Z1.9 (2003). For example, MIL-STD-105D, Code J, 1.5 AQL
is roughly matched to MIL-STD-414, Code K, 1.5 AQL, which matches ANSI Z1.9 (2003),
Code J, 1.5 AQL.
Variables Sampling Schemes 321

TABLE 12.1
Matching the Lettersa
MIL-STD-105D (ANSI MIL-STD-414 ANSI Z1.9 (2003)
Z1.4 1971) Code Letter Code Letter Code Letter

B B B
C C C
D D D
E E E
F F F
G G G
H H H
H I I
J K J
K M K
L N L
M O M
N P N
P Q P
a Delete MIL-STD-414 AQLs: 0.04, 0.065, 15.00.

Other changes in ANSI Z1.9 (2003) from the earlier version identical to MIL-STD-414
included an update of terminology and changes in the switching rules, inspection levels,
and other features to match MIL-STD-105D. Standards Z1.4 and Z1.9 may be obtained from
the American Society for Quality (ASQ). Schilling and Sheesley (1984) have addressed
scheme properties of the variables standard ANSI/ASQ Z1.9, producing tables patterned
after those of Chapter 11. Their use is identical.

Conversion of MIL-STD-414 to ANSI/ASQ Z1.9

By following the method of Commander Gascoigne, the tables of MIL-STD-414 are easily
converted into the tables of ANSI/ASQ Z1.9. The procedure is as follows:

1. Eliminate the MIL-STD-414 rows corresponding to Codes J and K, and reletter

the remaining code letters so that MIL-STD-414 code letters K, M, N, O, P, and Q
become J, K, L, M, N, and P.
2. Eliminate the columns corresponding to AQLs 0.04, 0.065, and 15.00.
3. Use the resulting table with the ANSI/ASQ Z1.4 switching rules.

This is the original procedure used to produce the ANSI/ASQ Z1.9 tables in 1980.
It should be noted that a few of the values are slightly off as the result of recomputation
over the years, but they are so slight as to be of little consequence in practical applica-
tion. The match with MIL-STD-105E is quite good as will be seen in the tables of dif-
ferences contained in Section E of the ANSI/ASQ Z1.9 tables. Figure 12.6 demonstrates
this change.
322 Acceptance Sampling in Quality Control

Acceptable quality levels (normal inspection)

Sample size Sample 0.04 0.065 0.10 0.15 0.25 0.40 0.65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
code letter size k k k k k k k k k k k k k k
B 3 1.12 0.958 0.765 0.566 0.341
C 4 1.45 1.34 1.17 1.01 0.814 0.617 0.393
D 5 1.65 1.53 1.40 1.24 1.07 0.874 0.675 0.455
E 7 2.00 1.88 1.75 1.62 1.50 1.33 1.15 0.955 0.755 0.536
F 10 2.24 2.11 1.98 1.84 1.72 1.58 1.41 1.23 1.03 0.828 0.611
G 15 2.64 2.53 2.42 2.32 2.20 2.06 1.91 1.79 1.65 1.47 1.30 1.09 0.866 0.664
H 20 2.69 2.58 2.47 2.36 2.24 2.11 1.96 1.82 1.69 1.51 1.33 1.12 0.917 0.695
I 25 2.72 2.61 2.50 2.40 2.26 2.14 1.98 1.85 1.72 1.53 1.35 1.14 0.936 0.712
J 30 2.73 2.61 2.51 2.41 2.28 2.15 2.00 1.86 1.73 1.55 1.36 1.15 0.946 0.723
KJ 35 2.77 2.65 2.54 2.45 2.31 2.18 2.03 1.89 1.76 1.57 1.39 1.18 0.969 0.745
L 40 2.77 2.66 2.55 2.44 2.31 2.18 2.03 1.89 1.76 1.58 1.39 1.18 0.971 0.746
MK 50 2.83 2.71 2.60 2.50 2.35 2.22 2.08 1.93 1.80 1.61 1.42 1.21 1.00 0.774
NL 75 2.90 2.77 2.66 2.55 2.41 2.27 2.12 1.98 1.84 1.65 1.46 1.24 1.03 0.804
OM 100 2.92 2.80 2.69 2.58 2.43 2.29 2.14 2.00 1.86 1.67 1.48 1.26 1.05 0.819
PN 150 2.96 2.84 2.73 2.61 2.47 2.33 2.18 2.03 1.89 1.70 1.51 1.29 1.07 0.841
QP 200 2.97 2.85 2.73 2.62 2.47 2.33 2.18 2.04 1.89 1.70 1.51 1.29 1.07 0.845
0.065 0.10 0.15 0.25 0.40 0.65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
Acceptable quality levels (tightened inspection)

FIGURE 12.6
Conversion of Table B.1 in MIL-STD-414 to create the corresponding table in ANSI/ASQ Z1.9.

MIL-STD-414 Derivatives
MIL-STD-414 was issued on June 11, 1957, and has not undergone any major changes since.
However, this classic standard was the precursor of several derivative standards, most
notably ANSI/ASQ Z1.9 and ISO 3951-1.

ANSI/ASQ Z1.9
A U.S. national standard, ANSI/ASQ Z1.9 represents an effort to unify variables and
attributes sampling systems by providing a reasonable match between a modified MIL-
STD-414 and MIL-STD-105. This was done using the Gascoigne technique. Other changes
included making the inspection levels coincide between the two standards and adopt-
ing the switching rules and lot size ranges of MIL-STD-105. Other editorial changes were
made as appropriate. ANSI/ASQ Z1.9, then, is a companion standard to the ANSI/ASQ
Z1.4 attributes standard. Given the lot size and AQL, it is possible to move between the two
standards with the same code letter and AQL.
The procedures and structure of ANSI/ASQ Z1.9 are essentially the same as for MIL-
STD-414. The excellent set of examples in MIL-STD-414 has been retained and will lead the
user through application of the Z1.9 standard. The ANSI/ASQ Z1.9 standard is an excellent
vehicle for in-house use and stands as the national standard to be employed internally to
the United States.

ASTM International Standard E2762

E2762 Standard Practice for Sampling a Stream of Product by Variables Indexed by AQL is an
ASTM standard, which contains and maintains the original MIL-STD-414 content as closely
Variables Sampling Schemes 323

as possible. This practice establishes lot or batch sampling plans and procedures for inspec-
tion by variables using MIL-STD-414 as a basis for sampling a steady stream of lots indexed
by AQL. This practice provides the sampling plans of MIL-STD-414 in ASTM format for
use by ASTM committees and others. It recognizes the continuing usage of MIL-STD-414
in industries supported by ASTM. Most of the original text in MIL-STD-414 is preserved in
Sections 6 through 9 of this practice. It is intended to provide a source for use in conjunction
with ASTM and other standards, which directly referenced MIL-STD-414. It is best used in
testing in a laboratory environment and with methodology in support of other standards.

ISO 3951-1
Part 1 of a set of five variables standards, ISO 3951-1 is the international version of MIL-
STD-414. Early versions were close to ANSI/ASQ Z1.9. In 2005, the standard underwent a
major revision, including adjustment of the tables to produce plans more closely matched to
the plans of ISO 2859-1. At that time, the range method was eliminated from the standard.
This standard is unique in its approach to variables plans in that it includes graphical
acceptance curves of the form shown in Chapter 10. The axes of the curves are converted
to (X, s) and the inspector simply plots X and s on the curve to determine if it is in the
region of acceptance or rejection. Given point A on the x-axis and point B on the y-axis, the
transformations are

s = a (U - L )

X = b ( U - L ) + L

A comparison of the procedure for ANSI/ASQ Z1.9 with ISO 3951-1 is shown in Table 12.2.
Note that ISO 3951-1 does not carry the M method used in ANSI/ASQ Z1.9 and uses the k
method essentially for single specification limits and the graphical technique for double
specification limits. ISO 3951-1 is best used in international trade.
ISO has also developed a series of schemes in support of the AQL system in ISO 3951-1
and patterned after the ISO 2859 series. These include the following:

ISO 3951-1 Specification for single sampling plans indexed by acceptance quality limit (AQL)
for lot-by-lot inspection of a single quality characteristic and a single AQL. This is the
ISO version of MIL-STD-414 and ANSI/ASQ Z1.9 and provides the basic tables and
subsidiary matter for the sampling system.
ISO 3951-2 General specification for single sampling plans indexed by acceptance quality limit
(AQL) for lot-by-lot inspection of independent quality characteristics. This is a complex
standard containing univariate and multivariate procedures addressing circum-
stances not covered by ISO 3951-1 for both the variability known and unknown.
The multivariate methods presented are for independent quality characteristics.
ISO 3951-3 Double sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot
inspection. This standard is complementary to the double sampling plans of ISO 2859-1
and addresses circumstances not covered there. It is quite complicated and includes
both univariate and multivariate methods for independent quality characteristics.
ISO 3951-4 Procedures for assessment of declared quality levels. The plans presented in
this standard have been matched to those of ISO 2859-4. It is the variables analog
of that standard. Note that this is a single test and does not involve a sampling
system as do the other parts of ISO 3951.
 324

TABLE 12.2
Procedure and Application of ANSI/ASQ Z1.9 and ISO 3951-1
Z1.9 (MIL-STD-414) ISO 3951-1

Double Specification

Step Section Form 1 Form 2 Section Single Specification Separate AQLs Combined AQL

Preparatory — Obtain k and n Obtain M and n Section 14 Obtain k and n from Obtain k and n Obtain appropriate
from from appropriate appropriate tables from acceptance curvea
appropriate tables appropriate
tables tables
U-X U-X U-X
Determine criteria Section B (s) zU = QU = Section 15 QU = Plot (s, X) and Reject if s > MSD =
s s s compare to f(U – L) otherwise,
X-L X-L X-L
zL = QL = QL = XU = U - kU s æ s X-Lö
s s s plot çç , ÷
XL = L + kL s è U - L U - L ÷ø
on diagram
U-X U-X
Section C (R) zU = QU = — — — —
R R
X-L X-L
zL = QL =
R R
c

Section D (σ) zU =
U-X
QU =
(U - X ) n Section 16 QU =
U-X
Compare X to Use separate AQL
s s s XU = U - kU s procedure
X-L ( X - L) n X-L
zL = QL = QL = XL = L + kLs
s s s
(Continued)
Acceptance Sampling in Quality Control
TABLE 12.2 (Continued)
Procedure and Application of ANSI/ASQ Z1.9 and ISO 3951-1
Variables Sampling Schemes

Z1.9 (MIL-STD-414) ISO 3951-1

Double Specification

Step Section Form 1 Form 2 Section Single Specification Separate AQLs Combined AQL
Estimation — — Enter table with n — — — —
and QU or QL to
get pU or pL
Action Single Accept if zU > Accept if pU < M or Single Accept if QU > k or Accept if X < —
specification k or zL > k pL < M specification QL > k XU or X > X L

Double Accept ifb zU > Accept if pU + pL < M Double Accept only if QU > k Separate AQL’s: Accept if point
specification k and zL > k specification or QL > k Accept only if plotted is inside
and s < MSD X L £ X £ XU diagram
or R < MAR
a Special procedure is used for sample size 3 or 4.
b Not official procedure.
c Scale factor, n = (n/(n - 1)) .
325
326 Acceptance Sampling in Quality Control

ISO 3951-5 Sequential sampling plans indexed by acceptance quality limit (AQL) for inspec-
tion by variables (known standard deviation). This standard presents variables sequen-
tial sampling plans matched to the attributes sequential plans of the ISO 2859-5
standard. It takes full advantage of the economics of sequential variables plans in
terms of minimal sample size.

As in the attributes plans, with the exception of ISO 3951-4, the ISO 3951 series is primar-
ily intended to be used with a continuing series of lots, utilizing the switching rules as
prescribed. The assumptions of variables sampling should be carefully considered in any
application of variables plans.

Further Considerations
An excellent description of the theory behind MIL-STD-414 has been given by Lieberman
and Resnikoff (1955) in the Journal of the American Statistical Association. Much of this
material was later presented in a detailed technical report on MIL-STD-414 published by
the Assistant Secretary of Defense (1958). These works give a detailed technical descrip-
tion of the background of the standard. A classic review of MIL-STD-414 was under-
taken by Kao (1971) and appeared in the Journal of Quality Technology. In a two-part series,
Duncan (1975) and Bender (1975) described the history and matching of MIL-STD-414 to
other national and international standards including MIL-STD-105D.

Software Applications
Excel
Use the Excel template Acceptance Sampling for Variables (ANSI_ASQC_Z19).xlsx to gen-
erate and evaluate a variables sampling plan based on the ANSI/ASQ Z1.9 standard.
Consider the following example, adapted from MIL-STD-414 (p. 69).
The specification for minimum electrical resistance of a certain electrical component
is 620 Ω. A lot of 100 items is submitted for inspection with an AQL = 1.0%. Suppose
the values of sample resistances are as follows: 643, 651, 619, 627, 658, 670, 673, 641, 638,
and 650.
Assume that the electrical resistances of this component have been shown to be normally
distributed. The sample size code letter table (Table T12.4) shows Code F to apply using
inspection level IV, which is used unless some other level is specified. The master tables
for normal and tightened inspection (Table T12.8) and the master table for reduced inspec-
tion (Table T12.9) give the following criteria for the plans involved in the MIL-STD-414
sampling scheme, Form 1:

Tightened: n = 10 , k = 1.84

Normal: n = 10 , k = 1.72

Reduced: n = 4, k = 1.34
Variables Sampling Schemes 327

The switching rules begin with normal inspection, and so the normal inspection plan
will be illustrated here. In application of the normal plan, we have

X = 647 s = 17.2

tL =
( X - L ) = 647 - 620 = 1.57
s 17.2

and since 1.57 < 1.72, the lot is rejected. The MSD is, of course, not used with single speci-
fication limits.
Using the Excel template Acceptance Sampling for Variables (ANSI_ASQC_Z19).xlsx, enter
the lot size of 100 in cell H2 as shown in Figure 12.7. Note that this lot size appears in cell
O12 next to the lot size range of 91–150. For a normal inspection plan at inspection
level II, enter these values in cells J7 and J8, respectively. The template determines the
proper sample size code letter is F and the required sample size n is 10.
Enter the lower specification limit of 620 into cell C15, and enter the sample data into
cells E5–E14. The spreadsheet computes the sample statistics from the entered data. In
addition, enter the AQL of 1% into cell I5 for the lower specification limit. To the right of
the cells shown in Figure 12.7 are lookup tables based on ANSI/ASQ Z1.9, which provide
the k, M, and F values needed depending on whether Form 1 or 2 is required.
Since this is a Form 1 situation, you can read the acceptability constant k from cell C20
and the lower quality index QL from cell C19. These values are 1.72 and 1.57, respectively,
which agree with calculations shown earlier. Finally, you can read the lot decision in cell
C24, which is to reject the lot.
The template also produces the corresponding OC and average outgoing quality (AOQ)
curves shown in Figures 12.8 and 12.9, respectively.

Minitab
Minitab cannot generate MIL-STD-414 (or equivalent) single variables sampling plans but
certainly can evaluate any given plan.

Statgraphics
Statgraphics has the ability to produce MIL-STD-414 (ANSI Z1.9/ISO 3951) single variables
sampling plans with OC curves for reduced, normal, and tightened inspection types for
the cases of standard deviation known or unknown. Note that the default inspection level
in Statgraphics is Level II (not Level IV as used in MIL-STD-414) because it is actually emu-
lating the ANSI/ASQ Z1.9 standard.
As a means of demonstrating how to generate variables inspection plans for all three
inspection types in the case of two specification limits (Form 2), we will revisit the ­earlier
example of electrical resistance of a certain electrical component. Stat­graphics will be used
to create plans based on the previous example described in the following texts.
The specification for electrical resistance of a certain electrical component is 650.0 = 30 Ω.
A lot of 100 items is submitted for inspection with AQL = 2.5% for the upper and AQL = 1%
for the lower specification limits. Suppose the values of sample resistances are as follows:
643, 651, 619, 627, 658, 670, 673, 641, 638, and 650.
Assume that the electrical resistances of this device have been shown to be nor-
mally distributed. The sample size code letter table (Table T12.4) shows Code F to apply
under  inspection level IV (level II in Statgraphics based on ANSI/ASQ Z1.9), which is
 328

FIGURE 12.7
Excel template Acceptance Sampling for Variables (ANSI_ASQC_Z19).xlsx sampling plan entry form.
Acceptance Sampling in Quality Control
Variables Sampling Schemes 329

Operating characteristic curve

100%

90%

80%
Probability of lot acceptance, Pa

70%

60%

50%

40%

30%

20%

10%

0%
0.000 0.050 0.100 0.150 0.200 0.250
Lot fraction nonconforming, p

FIGURE 12.8
OC curve for variables sampling plan with N = 100 and AQL = 1% for electrical resistance data.

Average outgoing quality curve

0.025

0.020
AOQ = P *Pa
Average outgoing quality, AOQ

AOQL = 0.0224
0.015

0.010

0.005

0.000
0.000 0.050 0.100 0.150 0.200 0.250
Lot incoming fraction nonconforming, p

FIGURE 12.9
AOQ curve with AOQL for variables sampling plan with N = 100 and AQL = 1% for electrical resistance data.
330 Acceptance Sampling in Quality Control

used unless some other level is specified. The master table for normal and tightened
inspection (Table T12.5) and the master table for reduced inspection (Table T12.6) give the
following criteria for the plans involved in the MIL-STD-414 sampling scheme.

1% AQL 2.5% AQL

Tightened n = 10, M = 2.17 n = 10, M = 4.77

Normal n = 10, M = 3.26 n = 10, M = 7.29
Reduced n = 4, M = 5.50 n = 4, M = 16.45

Since the scheme starts on normal inspection, we will illustrate application of the normal
plan. We find

X = 647 s = 17.2

In Statgraphics, generate the normal inspection variables plan for AQL = 1% for the lower
specification limit with SPC > Acceptance Sampling > Variables > MIL-STD-414 (ANSI
Z1.9/ISO 3951) and complete the dialog box as shown in Figure 12.10. We will use the
Variability Unknown option since we have an estimate based on our sample of s = 17.2 and
enter only the lower specification limit in Figure 12.11.
MIL-STD-414 (ANSI Z1.9/ISO 3951)
Variability: Unknown
Lot size: 91–150
Inspection level: II (default)

FIGURE 12.10
Statgraphics dialog box to generate an MIL-STD-414 (ANSI Z1.9/ISO 3951) variables plan for a lot size of 100 and
a 1% AQL for the lower specification limit.
Variables Sampling Schemes 331

FIGURE 12.11
Statgraphics dialog box to set the lower specification limit for an MIL-STD-414 (ANSI Z1.9/ISO 3951) variables plan.

AQL: 1.0%
Type of inspection: Normal
Sampling plan: (Code F)
Sample size: n = 10
Maximum allowable percent nonconforming: M = 3.270%

The StatAdvisor
The derived sampling plan requires taking a sample of 10 units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 3.27%, the lot should be rejected.

Next, generate the normal inspection variables plan for AQL = 2.5% for the upper speci-
fication limit with SPC > Acceptance Sampling > Variables > MIL-STD-414 (ANSI Z1.9/
ISO 3951) and complete the dialog box as shown in Figure 12.12. Again, use the Variability
Unknown option since we have an estimate based on our sample of s = 17.2, and enter only
the upper specification limit in Figure 12.13.
MIL-STD-414 (ANSI Z1.9/ISO 3951)
Variability: Unknown
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Normal
Sampling plan: (Code F)
332 Acceptance Sampling in Quality Control

FIGURE 12.12
Statgraphics dialog box to generate an MIL-STD-414 (ANSI Z1.9/ISO 3951) variables plan for a lot size of 100 and
a 2.5% AQL for the upper specification limit.

FIGURE 12.13
Statgraphics dialog box to set the upper specification limit for an MIL-STD-414 (ANSI Z1.9/ISO 3951) variables plan.
Variables Sampling Schemes 333

Sample size: n = 10
Maximum allowable percent nonconforming: M = 7.260%
StatAdvisor
The derived sampling plan requires taking a sample of 10 units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 7.26%, the lot should be rejected.

Following the procedure used to generate the plans for normal inspection, we can
c­ omplete the dialog boxes for each specification limit using the reduced and tightened
inspection types. The following results are obtained:
MIL-STD-414 (ANSI Z1.9/ISO 3951)
Variability: Unknown
Lot size: 91–150
Inspection level: II (default)
AQL: 1.0%
Type of inspection: Reduced
Sampling plan: (Code F)
Sample size: n = 4
Maximum allowable percent nonconforming: M = 5.460%
StatAdvisor
The derived sampling plan requires taking a sample of four units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 5.46%, the lot should be rejected.

MIL-STD-414 (ANSI Z1.9/ISO 3951)

Variability: Unknown
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Reduced
Sampling plan: (Code F)
Sample size: n = 4
Maximum allowable percent nonconforming: M = 16.410%
StatAdvisor
The derived sampling plan requires taking a sample of four units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 16.41%, the lot should be rejected.

MIL-STD-414 (ANSI Z1.9/ISO 3951)

Variability: Unknown
Lot size: 91–150
Inspection level: II (default)
AQL: 1.0%
Type of inspection: Tightened
Sampling plan: (Code F)
334 Acceptance Sampling in Quality Control

Sample size: n = 10
Maximum allowable percent nonconforming: M = 2.140%
StatAdvisor
The derived sampling plan requires taking a sample of 10 units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 2.14%, the lot should be rejected.

MIL-STD-414 (ANSI Z1.9/ISO 3951)

Variability: Unknown
Lot size: 91–150
Inspection level: II (default)
AQL: 2.5%
Type of inspection: Tightened
Sampling plan: (Code F)
Sample size: n = 10
Maximum allowable percent nonconforming: M = 4.720%
StatAdvisor
The derived sampling plan requires taking a sample of 10 units. From measurements
made on each unit, the mean and standard deviation are calculated. The program will cal-
culate the estimated percent of nonconforming units in the lot. If that estimate is greater
than 4.72%, the lot should be rejected.

The variables sampling plans generated by Statgraphics agreed closely with the table of
plans based on MIL-STD-414.

Problems
1. MIL-STD-414, 1.0% AQL is specified and a lot of 390 pieces is to be inspected. Find
the associated set of single-sided Form 1 normal, tightened, and reduced plans when
the standard deviation is unknown and estimated by s.
2. If the upper specification limit is 130, determine the acceptability of a lot for the plans
of Problem 1 if X = 110, s = 10.
3. MIL-STD-414, 1.0% AQL is specified and a lot of 390 pieces is to be inspected. Find
the associated set of two-sided Form 2 normal, tightened, and reduced plans when
the standard deviation is unknown and estimated by s.
4. If the upper and lower specification limits are 130 and 90, respectively, determine the
acceptance under the plans found in Problem 3 if X = 110, s = 10.
5. If Form 1 is to be used with the double specification limits of Problem 4, what is the
MSD? Would s = 10 pass the MSD?
6. What is the lot tolerance percent defective of the plan Code F, 0.4 AQL? What is its
indifference quality?
7. What action should be taken under Code G, 4.0 AQL normal inspection if 7 of the
10 lots have estimated percent defective greater than the AQL and the process ­average
of the last 10 lots exceeds the AQL? What would be the minimum possible process
Variables Sampling Schemes 335

average under these circumstances? Is it possible to switch to reduced inspection

under these conditions?
8. What MIL-STD-414 plan would roughly match MIL-STD-105D, Code K, 0.65 AQL?
9. Suppose range plans are substituted for the standard deviation plans of Problem 4.
The criteria for the normal range plan are n = 30, M = 2.81, and c = 2.353. Compute
QU when X = 110 and R = 23.53. If the corresponding pU = 1.88, should the lot be
accepted?
10. Suppose the standard deviation was known and the inspection of Problem 4 was
to be applied. The criteria for the normal standard deviation known plan are n = 9,
M = 2.59, and v = 1.061. If X = 110 and s = 10.61, should the lot be accepted? Use upper
tail percentages of the normal distribution as estimates of pU and pL.

References
American National Standards Institute, 2003, American National Standard: Sampling Procedures and
Tables for Inspection by Variables for Percent Nonconforming, ANSI/ASQC Standard Z1.9–2003,
American Society for Quality Control, Milwaukee, WI.
American Society for Testing and Materials, Standard Practice for Sampling a Stream of Product by
Variables Indexed by AQL, ASTM Standards E2762, Vol. 14.02, American Society for Testing and
Materials, West Conshohocken, PA.
Bender, A., 1975, Sampling by variables to control the fraction defective: Part II, Journal of Quality
Technology, 7(3): 139–143.
Duncan, A. J., 1955, The use of ranges in comparing variabilities, Industrial Quality Control, 11(5):
18–22.
Duncan, A. J., 1975, Sampling by variables to control the fraction defective: Part I, Journal of Quality
Technology, 7(1): 34–42.
Gascoigne, J. C., 1976, Future International Standards on Sampling by Variables, American Society for
Quality Control Technical Conference Transactions, Toronto, Ontario, Canada, pp. 472–478.
Juran, J. M. (Ed.), 1999, Quality Control Handbook, 5th ed., McGraw-Hill, New York, pp. 25.1–25.41.
Kao, J. H. K., 1971, MIL-STD-414 Sampling procedures and tables for inspection by variables for
percent defective, Journal of Quality Technology, 3(1): 28–37.
Lieberman, G. J. and G. J. Resnikoff, 1955, Sampling plans for inspection by variables, Journal of the
American Statistical Association, 50: 457–516.
Schilling, E. G., 1974, Variables sampling and MIL-STD-414, Quality Progress, 7(5): 16–20.
Schilling, E. G. and J. H. Sheesley, April 1984, The performance of ANSI=ASQ Z1.9-1980 under the
switching rules, Journal of Quality Technology, 16(2): 101–120.
United Kingdom Ministry of Defence, 1974, Sampling Procedure and Charts for Inspection by Variables
(Defence Standards 05-30), Ministry of Defence Directorate of Standardization, London, U.K.
United States Department of Defense, 1950, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105A), U.S. Government Printing Office, Washington, DC.
United States Department of Defense, 1957, Military Standard, Sampling Procedures and Tables for
Inspection by Variables for Percent Defective (MIL-STD-414), U.S. Government Printing Office,
Washington, DC.
United States Department of Defense, 1958, Mathematical and Statistical Principles Underlying MIL-
STD-414, Technical Report, Office of the Assistant Secretary of Defense (Supply and Logistics),
Washington, DC.
United States Department of Defense, 1963, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105D), U.S. Government Printing Office, Washington, DC.
13
Special Plans and Procedures

A variety of plans and procedures have been developed for special sampling situations
involving both measurements and attributes. Only a few of them can be shown here. Each
is tailored to do a specific job under prescribed circumstances. They range from a simpli-
fied variables approach involving no calculations to a more technically complicated combi-
nation of variables and attributes sampling in a so-called mixed plan. They provide useful
options in the application of acceptance sampling plans to unique sampling situations.

No-Calc Plans
Since variables plans for percent nonconforming usually assume an underlying normal
distribution of measurements, probability plots would seem to be a natural tool for accep-
tance sampling. Such plots can provide a visual check on the normality of the distribu-
tion involved, while at the same time affording an opportunity to estimate the fraction
nonconforming in the lot (see Chapter 3). Probability plots can also be used directly for lot
acceptance. Such a plan has been developed by Chernoff and Lieberman (1957). It assumes
underlying normality of individual measurements. Although its results are approximate,
it requires no calculations and can be used in inspection situations, where mathematical
calculation is out of the question. The authors of the plan point out (see the section “Lot
Plot Plans”) “No-Calc is not a replacement for the usual variables procedures when a con-
tract between two parties exists and calls for inspection by variables.” Nevertheless, it is
particularly useful for internal in-process acceptance inspections and the like.
The No-Calc procedure is matched to MIL-STD-414 (United States Department of
Defense, 1957). Plans are identified by code letter and AQL. The operating characteristic
(OC) curves of MIL-STD-414 approximate those of No-Calc and can be used to select a
plan. Sample sizes are, of course, limited to the MIL-STD-414 sequence, which appears in
the No-Calc tables. For a given sample size, the No-Calc procedure is as follows:

1. Plot the sample results on normal probability paper using the No-Calc plotting
positions of Appendix Table T13.1 when n ≤ 20; when n > 20, use the approximation

i - ( 1/2 )
p̂( i ) = ´ 100
n

2. If the points do not plot roughly in a straight line, discontinue the procedure on the
grounds that the underlying population may not be normal.
3. Estimate the underlying normal distribution by drawing a straight line through
the points.

337
338 Acceptance Sampling in Quality Control

4. Locate the specification limits on the x-axis and use the straight line to estimate the
percent nonconforming beyond the single or double specification limits. Call this
estimate p̂.
5. Obtain the critical value of p* from Appendix Table T13.2.
6. If p̂ ≤ p*, accept the lot; if p̂ > p*, reject the lot.

Clearly, p* plays the role of M in MIL-STD-414, while p̂ acts as pL, pU, or pT. To illustrate the
application of the No-Calc plan, consider the following example.
The specification for minimum electrical resistance of a certain electrical component is
620 Ω. A lot of 100 items is submitted for inspection with an AQL = 1.0%. A 10% limiting
quality of 15% is desired. Suppose values of sample resistances are as follows: 643, 651, 619,
627, 658, 670, 673, 641, 638, and 650 Ω.
A search through the OC curves of MIL-STD-414 shows Code F, 1.0% AQL is closest to the
specifications of the plan. Its 10% limiting quality is just about 15%, while Code E and Code
G differ substantially from that at 1.0% AQL. Reference to Appendix Table T13.2 shows that
a sample size of 10 should be taken with a critical value of p* = 3.88. Plotting positions are
obtained from Appendix Table T13.1 and associated with the observations as follows:

Order (i) X(i) P(i)

1 619 4.4
2 627 16.4
3 638 26.2
4 641 35.8
5 643 45.3
6 650 54.7
7 651 64.2
8 658 73.8
9 670 83.6
10 673 95.6

The probability plot appears as Figure 13.1. It estimates that 6% of the underlying distri-
bution is below the lower specification limit of 620 Ω. Since

pˆ = 6.0 > p* = 3.88

the lot is rejected. It is interesting to note that this is the same estimate obtained from the
probability plot of Chapter 3 and illustrates how good the approximation

i + ( 1/2 )
p̂( i ) =
n

is even with a reduced amount of data.

The No-Calc plan can easily be implemented in the shop by drawing a vertical decision
line at the specification limit (in the example, 620 Ω). On this line, mark the critical value
p* (3.88 here). Label the line below p* as “accept” and above p* as “reject.” Take the action
indicated by the intersection of the probability plot line with the decision line. Care should
be taken so that the decision line does not prejudice drawing the probability plot line.
Special Plans and Procedures 339

99

95
90
80
P

50

20
10
5

620 630 640 650 660 670 680

x

FIGURE 13.1
Probability plot.

Lot Plot Plans

Probably no acceptance sampling procedure has more intuitive appeal for inspectors than
the lot plot method, developed by Dorian Shainin at the Hamilton Standard Division of
United Aircraft Company. Shainin (1950) published an extensive introduction and descrip-
tion of the plan in Industrial Quality Control. The reader is well advised to study his paper
for the details of the method. Its wide acceptance attests to its appeal and value in practical
applications. Detailed examples of its use in various companies have also been given by
Shainin (1952).
Lot plot uses a constant sample size of 50 observations. It is based on the construction of
the histogram of the sample. The mean and the standard deviation are estimated and the
resulting “3σ” limits used as an acceptance criterion when compared to the specification
limits. It serves as a particularly useful tool in the introduction of statistical acceptance
sampling techniques and in applications where more sophisticated methods are inappro-
priate or not likely to be well received. As Shainin (1952) has pointed out, with the lot plot
method, “…it was possible to bring the method of analysis down to where anyone who
can read a micrometer can be taught in less than a week to analyze lot plots completely.”
The original procedure utilizes the Pearson–Tippett method to estimate the standard
deviation from 10 ranges of 5 obtained from subsamples of the 50 observations. We shall
present a variation of lot plot, due to Ashley (1952), which was applied at the Bendix
Aviation Corporation. This method calculates the standard deviation directly from the fre-
quency distribution itself, thus, avoiding use of the range. This preserves all the qualities
340 Acceptance Sampling in Quality Control

FIGURE 13.2
Example of completed modified lot plot form. (Reprinted from Ashley, R.L., Ind. Qual. Control, 8(5), 30, 1952.
With permission.)

intended by Shainin but leads to somewhat more rapid calculation, possible computeriza-
tion, and no need to order the observations into subgroups as taken.
A completed modified lot plot form appears as Figure 13.2. After completing the head-
ing, the form is filled in from left to right as follows:

1. A sample of 50 is taken.
2. The mean of the first five observations is used to locate the center of the distribu-
tion. Enter this value in the leftmost column next to the value of 0 in the second
column, suitably rounded to obtain a nice starting point.
3. Mark off the cells above and below the center value. Individual measurements (cell
width of 1) are desirable but not necessary. If cells of width other than 1 are to be
used, enter the cell midpoints above and below the middle cell. Shainin suggests a
cell width roughly equal to one-fourth the range of the first five observations.
4. Each observation is tallied by a check mark in the space provided. This will auto-
matically provide a histogram of the sample.
5. If the histogram appears to be obviously nonnormal, stop and investigate the cause.
6. Tally for each row is recorded in the F column. The numbers at the top of the grid
facilitate the count.
Special Plans and Procedures 341

7. The FX column is filled in as the product of the F and X values shown for each row.
8. The FX2 column is filled in as the product of the F and X2 values shown for each
row.
9. The sum of the FX column is recorded in the upper right box. When the sum is
multiplied by 0.02, the result is a coded X recorded in the second box.
10. The sum of the FX2 column is recorded in the third box on the right. When the
value is multiplied by 0.02, the result is recorded in the fourth box.
11. Table 13.1 is then used to estimate 3σ. The coded X (second box) is entered at the
top and .02ΣFX2 (fourth box) is entered at the side. The resulting closest tabulated
value estimates 3σ and is entered in the fifth box. If the table does not cover the
values obtained, the estimate of 3σ of the Xs can be calculated from the formula

3s = 3 .02SFX 2 - X 2

12. The cell width w is entered in the sixth box. If it is necessary to estimate the mean
and standard deviation units of the original measurement, use

mˆ = wX
w ( 3s )
sˆ = = w .02SFX 2 - X 2
3

To assess the acceptability of the lot, the upper lot limit (ULL) and lower lot limit (LLL) as
well as the specifications are drawn on the chart. To draw these limits

1. Using the X column (second and twenty-fourth columns), draw a horizontal

line at the coded X. This is an estimate of the mean of the distribution and can
be decoded simply by reading the corresponding value from the cell midpoints
recorded (first column).
2. Mark a distance 3σ (fifth box) in terms of the X column above and below the coded X.
These are the lot limits, LLL and ULL. They may be read in terms of the original
measurements simply by extending them to the cell midpoints (column 1) and
reading off the appropriate values.
3. Draw the specification limits on the chart in terms of the cell midpoints (column 1).
The acceptance criteria are as follows:
1. If the lot limits are within the specification limits, accept.
2. If the lot limits are outside the specification limits, perform the following:
a. Count the number of X spaces by which the lot limit exceeds the specifica-
tion limit. Call this E.
b. Compute

æ E ö
Z = 3ç1- ÷
è 3s ø

where the 3σ value is taken from the fifth box.

 342

TABLE 13.1
3σ Values for Lot Plot
X Values to Nearest Tenth
.02ΣFX 2 .0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5
.0
.5 2.1 2.1 2.0 1.9 1.7 1.5 1.1 0.3
1.0 3.0 3.0 2.9 2.9 2.7 2.6 2.4 2.1 1.8 1.3
1.5 3.7 3.7 3.6 3.6 3.5 3.4 3.2 3.0 2.8 2.5 2.1 1.6 0.7
2.0 4.2 4.2 4.2 4.1 4.1 4.0 3.8 3.7 3.5 3.3 3.0 2.7 2.2 1.7 0.6
2.5 4.7 4.7 4.7 4.7 4.6 4.5 4.4 4.3 4.1 3.9 3.7 3.4 3.1 2.7 2.2 1.5
3.0 5.2 5.2 5.2 5.1 5.1 5.0 4.9 4.8 4.6 4.4 4.2 4.0 3.7 3.4 3.1 2.6 2.0 1.0
3.5 5.6 5.6 5.6 5.5 5.5 5.4 5.3 5.2 5.1 4.9 4.7 4.5 4.3 4.0 3.7 3.4 2.9 2.3 1.5
4.0 6.0 6.0 6.0 5.9 5.9 5.8 5.7 5.6 5.5 5.4 5.2 5.0 4.8 4.6 4.3 4.0 3.6 3.2 2.6 1.9
4.5 6.4 6.4 6.3 6.3 6.2 6.2 6.1 6.0 5.9 5.8 5.6 5.4 5.2 5.0 4.8 4.5 4.2 3.8 3.4 2.8 2.1 0.9
5.0 6.7 6.7 6.7 6.6 6.6 6.5 6.5 6.4 6.3 6.1 6.0 5.8 5.7 5.5 5.2 5.0 4.7 4.4 4.0 3.5 3.0 2.3 1.2
5.5 7.0 7.0 7.0 7.0 6.9 6.9 6.8 6.7 6.6 6.5 6.4 6.2 6.0 5.9 5.6 5.4 5.1 4.8 4.5 4.1 3.7 3.1 2.4 1.4
6.0 7.3 7.3 7.3 7.3 7.3 7.2 7.1 7.0 6.9 6.8 6.7 6.6 6.4 6.2 6.0 5.8 5.6 5.3 5.0 4.6 4.2 3.8 3.2 2.5 1.5
6.5 7.6 7.6 7.6 7.6 7.6 7.5 7.4 7.4 7.3 7.2 7.0 6.9 6.7 6.6 6.4 6.2 6.0 5.7 5.4 5.1 4.7 4.3 3.9 3.3 2.6 1.5
(Countinued)
Acceptance Sampling in Quality Control
TABLE 13.1 (Countinued)
3σ Values for Lot Plot
X Values to Nearest Tenth
2
Special Plans and Procedures

.02ΣFX .0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5
7.0 7.9 7.9 7.9 7.9 7.8 7.8 7.7 7.7 7.6 7.5 7.3 7.2 7.1 6.9 6.7 6.5 6.3 6.1 5.8 5.5 5.2 4.8 4.4 3.9 3.3 2.6
7.5 8.2 8.2 8.2 8.2 8.1 8.1 8.0 7.9 7.9 7.8 7.6 7.5 7.4 7.2 7.1 6.9 6.7 6.4 6.2 5.9 5.6 5.3 4.9 4.5 4.0 3.4
8.0 8.2 8.1 8.0 7.9 7.8 7.7 7.5 7.4 7.2 7.0 6.8 6.5 6.3 6.0 5.7 5.3 5.0 4.5 4.0
8.5 8.2 8.1 8.0 7.8 7.7 7.5 7.3 7.1 6.9 6.6 6.4 6.1 5.7 5.4 5.0 4.5
9.0 8.3 8.1 8.0 7.8 7.6 7.4 7.2 7.0 6.7 6.4 6.1 5.8 5.4 5.0
9.5 8.2 8.1 7.9 7.7 7.5 7.3 7.0 6.8 6.5 6.2 5.8 5.4
10.0 8.2 8.0 7.8 7.6 7.3 7.1 6.8 6.6 6.2 5.8
10.5 8.3 8.1 7.9 7.6 7.4 7.1 6.8 6.5 6.2
11.0 8.2 7.9 7.7 7.4 7.2 6.9 6.5
11.5 8.2 8.0 7.7 7.5 7.2 6.9
12.0 8.3 8.0 7.8 7.5 7.2
12.5 8.3 8.1 7.8 7.5
13.0 8.3 8.1 7.8
13.5 8.3 8.1

Source: Reprinted from Ashley, R.L., Ind. Qual. Control, 8(5), 31, 1952. With permission.
343
344 Acceptance Sampling in Quality Control

c. Estimate the proportion of product out of specification as the upper normal

tail area corresponding to Z. This may be done graphically by the inspec-
tor simply by providing a normal probability paper having a straight line
drawn, which corresponds to the standard normal distribution. A suitable
table could also be provided.
d. If the estimated value is less than a predetermined allowable value, accept
the lot.
e. Otherwise, reject the lot.

The method suggested here for estimating the proportion out of specification has been
found by the authors to be accurate enough for most practical purposes. Clearly, lot plot
could be easily computerized but is presented here in its original form for historical rea-
sons and because of its popularity.

Narrow-Limit Gauging
The predominance of attributes-type data in an industry attests to the economic advan-
tages of collecting go/no-go data over recording specific variables data. Gauging is often
to be preferred over measurement. This is because it takes less skill to gauge properly, is
faster and less costly, and has become something of a tradition in certain industries. As put
by Ladany (1976):

Variables sampling plans have the known advantage, over sampling plans for attributes,
of requiring a much smaller sample size…This is due to the possibility of utilizing more
effectively quantitative data as opposed to qualitative data. The statistical advantage
may be out weighed by economic considerations, since the cost of inspecting a unit,
using a simple go/no-go gauge, is often much lower than the cost of determining the
exact value of the critical characteristic variable by a measuring instrument.

Narrow-limit sampling plans (sometimes called compressed limit plans) effectively bridge
the gap between variables and attributes procedures by utilizing go/no-go gauges setup
on the principles of variables inspection. Originated in England by Dudding and Jennett
(1944), they were introduced into the United States by Mace (1952). Ott and Mundel
(1954) did much to extend the theory and application of the procedure. The narrow-limit
plans were initially regarded as a process control device as evidenced by the title Quality
Control Chart Technique When Manufacturing to a Specification used by Dudding and Jennett.
Nevertheless, narrow-limit plans provide an excellent technique for acceptance sampling
in that they are based on the same assumptions as known standard deviation variables
plans for proportion nonconforming but require little calculation and are easier to use.
The basic idea is a simple one. Since the sample size required by an attributes plan is related
inversely to the size of the proportion nonconforming, it is required to detect that a pseudo-
specification, or narrow limit, is set inside the specification limits. The sampling plan is set up
on the number of items failing the narrow limit rather than the specification limit itself. Since the
relationship between the pseudo and the actual proportion nonconforming is strictly monotoni-
cally increasing, the one can be used to control the other. This is then done by using the narrow
limit. These plans assume the standard deviations to be known and the underlying distribution
of measurements to be normal. Of course, when the specification limits are more than 6σ apart,
individual narrow-limit plans can be applied on each side of double specification limits.
Special Plans and Procedures 345

Using the notation of Ott and Mundel (1954), narrow-limit gauge (NLG) plans are speci-
fied by three quantities:

1.
n is the sample size.
2.
c is the acceptance number for units allowed outside the NLG.
3.
t is the compression constant, the narrow limit is set inside the specification limit.

To implement a plan, follow these steps:

1. Check to be sure that the underlying distribution of measurements is consistently

normal using probability plots, tests of fit, control charts, etc., on past data.
2. For a single upper specification limit U or a lower specification limit L, set the NLG at
U - ts or L + ts
When double specification limits are at least 6σ apart, individual narrow-limit plans
can be applied to each of the specification limits separately.
3. Take a random sample of size n.
4. Gauge to the narrow limit. Items outside the narrow limit are treated as noncon-
forming to the NLG. Items inside the narrow limit are treated as conforming.
5. Accept if the number nonconforming to the NLG is less than or equal to c; other-
wise, reject.

It should be noted that changes in the criteria for acceptance affect the OC curve of the
narrow-limit gauging procedure in different ways:

n increased → plan tightened

c increased → plan loosened
t increased → plan tightened

A large value of t can lead to rejections even when p = 0. Ott and Mundel have found a
compression constant of
t = 1
to be very good in practice with moderately small sample sizes. OC curves for several
plans having t = 1 are shown in Figure 13.3.
The OC curve of a narrow-limit plan is relatively easy to compute. Figure 13.4 shows
diagrammatically the principle behind its computation.
Assuming an underlying normal distribution of measurements, each value of propor-
tion nonconforming p will be associated with a fixed position of the mean μ. If we let zγ to
be the standard normal deviation having an area γ in the upper tail, then the specification
limit will be a distance zpσ from the mean. Also, the NLG will be a distance tσ from the
specification limit or a distance zgσ from the mean. Hence, we have

zg = zp - t

The proportion of units outside the narrow-limit pg is the upper tail normal area cut off by
zg. The sampling plan will then be applied to a proportion pg when the proportion p is out
of the specification limit. The calculations are summarized in Table 13.2, which ­illustrates
346 Acceptance Sampling in Quality Control

1.00
t = 1.0
0.90
n=5
c=2
0.80
Pa = probability of acceptance
0.70

0.60
n = 15
c=4
0.50

0.40
n = 15
c=3
0.30

0.20 n = 15
c=2
0.10

0
1 2 3 4 5 6 7 8 9
β-Incoming percent defective

FIGURE 13.3
OC curves for NLG plans with t = 1. (Reprinted from Ott, E.R. and Mundel, A.B., Ind. Qual. Control, 10(5), 30,
1954. With permission.)

NLG SPEC

pg

zg σ fσ p

μ x
zg σ

FIGURE 13.4
Narrow-limit distribution.

TABLE 13.2
Calculation of Probability of Acceptance for n = 15, c = 2, t = 1

Proportion Out of σ Units, Proportion Out

Specification, p zp zg = z p − t of NLG, pg npg Poisson, Pa

0.005 2.580 1.580 0.057 0.86 0.94

0.03 1.880 0.880 0.189 2.84 0.46
0.08 1.405 0.405 0.341 5.12 0.12
Special Plans and Procedures 347

finding the probability of acceptance for the plan n = 15, c = 2, t = 1 using the Poisson prob-
abilities to approximate the binomial. Care should be taken to be sure that the Poisson
approximation applies. If not, the binomial distribution should be used directly. It will be
seen that these values are shown on the OC curve of Figure 13.3.
The following procedure may be utilized to derive NLG plans when the Poisson approx-
imation applies. Refer to Figure 13.4 for a diagrammatic representation of the procedure.
The following are given:

p1 as the producer’s quality level

p2 as the consumer’s quality level
α as the producer’s risk
β as the consumer’s risk
t as the compression constant

1. Determine zP1 and zP2.

2. Compute zg1 = zP1 − t and zg2 = zP2 − t.
3. Obtain upper tail areas pg1 and pg2.
4. Compute the operating ratio

p g2
R=
p g1

5. Determine standard acceptance sampling plan n, c with the risks α and β and oper-
ating ratio R.
6. The narrow-limit plan is specified as n, c, and t.

For example, suppose the following plan is desired:

p1 = .005, α = .05, t = 1.0

p2 = .08, β = .10

1. z.005 = 2.576 and z.08 = 1.405.
2. zg1 = 1.576 and zg2 = 0.405.
3. pg1 =  . 0575 and pg2 =  . 3427.
.3427
4. R = = 5.96.
.0575
5. Use of the table of unity factors gives c = 2, n = 14.2 ~ 15.
6. The plan is n = 15, c = 2, t = 1.

It can be confirmed that the desired characteristics were essentially obtained by reference
to Table 13.2, which was used to compute the OC curve of this plan. Use of the unity values
requires that the Poisson approximation to the binomial apply to both values of npg.
It is frequently desirable to obtain an optimum narrow-limit plan with regard to sam-
ple size. For p1, p2, α, and β, specified as before, Ladany (1976) has developed an iterative
procedure for the construction of such a plan. It utilizes a special nomograph based on
348 Acceptance Sampling in Quality Control

the Larson (1966) nomograph for the binomial distribution. The nomograph is shown in
Figure 13.5. It should be noted that in Ladany’s notation
t = DZ
Steps in the application of the procedure are as follows:

1. Connect p1 and p2 on the variables sampling plan axis (middle axis, right side) with
(1 − α) and β, respectively, on the probability axis (right half) using two straight
lines.
2. Read the corresponding σ known variables plan sample size, from the horizontal
axis on top, directly above the point of intersection of the two lines. This serves as
an extreme lower bound for the narrow-limit sample size.
3. Locate p1 and p2 on the binomial sampling plan axis (middle axis, left side) and con-
nect with (1 − α) and β, respectively, on the probability axis (right side). Using the
Larson nomograph, read the sample size n0 and acceptance number c0 from the grid.
This is the plan that would apply without the narrow limit. That is, when t = 0.
4. Select a trial value of t, say, t1. From p1, move down the slanted ΔZ (= t) axis a dis-
tance of t1. Read over horizontally to obtain pg1 on the binomial sampling plan axis
(middle axis, left side). Similarly, move down from p2 on the slanted ΔZ (= t) axis
a distance t and read over to get pg2. For example, moving from p2 = .08 a distance
t = 1.0 on the diagonal scale and reading across gives pg2 = .343, which is the same
value obtained in Table 13.2.
5. Connect pg1 and pg2 on the binomial sampling plan scale with 1 − α and β, respec-
tively, on the probability scale. The intersection of these two lines gives the value
of n and c, which will provide the desired risks with t = t1.
6. Select another value of t and determine the values of n and c for it as in steps 4
and 5.
7. Continue the iterative procedure until the last derived narrow-limit plan starts
to increase in sample size, with no indication that sample size may be further
reduced or until pg2 exceeds 0.50.

The σ known sample size provides a rough indication of how close the iterative procedure is to
optimum. Naturally, the variables sample size will never be reached; however, the NLG should
reduce the sample size by roughly 80% of the difference between attributes and variables.
It is best to keep a running table of the results of the iterations. Note that this table could
also be developed by changing t in the tabular method presented earlier. Ladany (1976)
gives the following example:

p1 = .02, α = .05
p2 = .08, β = .10

The nomograph for the example is shown in Figure 13.6. We see that

1. Initial lines are shown dotted from .02 on the probability scale to .95 and from .08
to .10 on the variables sampling plan scale. They cross at point 0.
2. The σ known sample size is n = 20.
3. Points .02 and .08 on the binomial axis are connected to .95 and .10 on the prob-
ability axis to give the plan n = 98, c = 4 when t = 0.
 Fraction defective, P1, P2
Binomial Variable Sample size of the variable sampling plan
sampling sampling

50
40
30
10
4
3

300
5

200
100
100
20
2
1
plan plan
0.01 nx
0.1 0
0.005 0 100
0.2 0
70
0.03 0
0.3 5 50
0.2 0.1 0
0.4
40
10 0
0.3 0.2 0.03 30
0.5
0.4 0.3
0
0.04 20 20
0.6 0.01
0.5 0.2 0
0.4 30 14 0.001
0.7 0.05
0
0.6 0.3 10
Special Plans and Procedures

0.5 0.06 40
0.2
0.8

S
0.4 70 0.005

currences (c)
0.6 30

am
0.7 0.3 0.07
0.2 0.01
0.9

ple
0.08
40
0.5 0.02

siz
0.8 0.7 30

e
0.4 0.3 0.2 0.09 70
1.0
0.1
(n)
30
0.6 0.00

Δz
0.9 0.8 0.10
0.5 0.4 0.3 0.02
1.1

Number of oc
0.2
20
1.0 0.7 0.1 100 0.10
0.9
0.6 0.5 0.4
1.2 0.3
0.8 0.2 0.1
10 0.20
1.1 1.0 0.7 0.6 0.5 140
1.3 0.4 0.15
0.30
1.2 0.9 0.3 0.2 0.03 8
1.1 0.8 0.7 0.6 0.40
0.1
1.4
1.3 0.5 0.50
1.0 0.4 0.3
1.2 0.9 0.8 0.7 200
0.2 0.20 2 0.60
1.5 0.6
1.4 0.04 0
1.3 1.1 0.5 0.4 0.70
1.0 0.9 0.8 0.1
0.3
1.6 0.7 0.80
1.5 1.2 0.6 0.5 0.25
1.4 1.1 1.0 0.9 0.4 0.2 0.05
1.7 0.8 0.90
1.3
1.6 0.7 0.6 0.1
1.5 1.2 1.1 1.0 0.5 0.3 0.95
1.8 0.30 0.06
0.9
1.4 0.8 0.7 0.2 2
1.7 1.6
1.3 1.2 1.1 0.6 0.4 0.98
(c)

1.9 0.1
1.8 1.0 0.07 3 0.99
1.5 0.9 0.8 0.3 0.35
1.7
nces

1.4 1.2
Probability of ε or fewer occurrences in η trials, β, 1–α

1.3 0.7 0.5 4 0.995

0.2
1.1 0.08
urre

0.1 6
1.6 1.0 0.9 0.4
1.8 1.5 1.4 1.3 0.6 0.40
Occ

0.8 0.999
0.3 0.09 7
1.2 0.2
1.7 1.1 1.0 0.5
1.6 1.5 1.4 0.7 0.1 9
0.9 0.10
0.4 0.45
1.3 1.2 0.6 0.3
1.7 1.6 1.5 1.1 0.2 0.11
1.0 0.8 0.5 0.1
0.50 0.12

FIGURE 13.5
Ladany nomograph of narrow-limit gauging sampling plans. (Reprinted from Ladany, S.P., J. Qual. Technol., 8(4), 227, 1976. With permission.)
349
 350
Fraction defective, P1, P2
Binomial Variable Sample size of the variable sampling plan
sampling sampling

200
100
50
30
20
10
4
3
2

300
100
40
5
1
plan plan n
x
0.1
0.005 0 1000
0.2 0
70
0.3 0.03 0
5 50
0.2 0
0.4 0.1 40
10 0
0.3 0.2 0.03 30
0.5

∆Z = 0.30
0.4 0.3 0
0.6 0.04 20 20

(c)
0.01 0
0.2

s
Sam
0.5 0.4
0.06 30 10
0.7 0.001

ple

∆Z = 0.650
0
0.3 10

siz
0.6 0.5
0.2 0.06 40
0.8

e (n

∆Z = 0.950
)
0.7 0.6 0.4 0.07 70 0.05

occurrence
0.3 0.2 50
0.9 30 0.01
0.5 0.08
0.8 0.7 0.2 40 0.02
0.4 0.3 0.09
1.0 70

∆Z = 1.215
0.1 30
0.9 0.6 0.10

Number of
∆z 1.1 0.8 0.5 0.4 0.3 0.02 0.05

∆Z = 1.405
0.2 100 20
1.0 0.9 0.7 0.1 0.10

∆Z = 0.300
0.6 0.5 0.4
1.2
0.3
1.1 1.0 0.8 0.2 0.1 10 0.20
0.7 0.6 0.5 140
1.3 0.15
0.4 0.30
1.2 0.9 0.3 0.2 0.03
1.1 5

∆Z = 0.650
0.8 0.7 0.6 0.1 0.40
1.4
0.5
1.0 0.4 0.3 0.50

∆Z = 0.950
1.3 1.2
0.9 0.8 0.7 0.2 0.20 200 2
1.5 0.60
0.6 0.04
1.4 1.1 0.5 0.4 0
1.3 1.0 0.9 0.8 0.3 0.1
0.70
1.6

∆Z = 1.215
1.2 0.7 0.80
1.5 0.6 0.5 0.25
1.4 1.1 1.0 0.9 0.2
1.7 0.4 0.05
0.8 0.90

∆Z = 1.405
1.6 1.3 0.7 0.6 0.1 1
1.5 1.2 1.1 1.0
1.8 0.5 0.3
0.30 0.06 0.95
0.9
1.7 1.4 0.8 0.7 0.2
1.3 1.2 1.1 2
1.6 0.6 0.4
s (c)

1.9 0.1 0.98

1.0 0.07 3
1.5 0.9 0.3 0.35
nce

1.8 0.8 0.99

1.7 1.4 1.3 1.2 0.7 0.5 0.2 4
Probability of ε or fewer occurrences in η trials, β, 1–α

0.995
urre

1.1 0.1 0.08

1.6 1.0 0.9 0.4 5
1.8 1.5
Occ

1.4 1.3 0.8 0.6 0.40

0.3 0.09 7 0.999
1.2 1.1 0.2
1.7 1.0 0.5
1.6 1.5 1.4 0.7 0.1
0.9 0.10 9
0.4 0.45
1.3 0.3
1.2 0.6
1.7 1.6 1.5 1.1 0.2 0.11
1.0 0.8 0.5 0.1
0.50 0.12

FIGURE 13.6
Use of the Ladany nomograph of narrow-limit gauging sampling plans for solution of example. (Reprinted from Ladany, S.P., J. Qual. Technol., 8(4), 229, 1976.
With permission.)
Acceptance Sampling in Quality Control
Special Plans and Procedures 351

TABLE 13.3
Iterative Use of Ladany Nomograph
Point Number
Iteration t pg1 pg 2 n c in Figure 13.6

0 0.000 0.020 0.080 98 4 1

1 0.300 0.040 0.135 67 5 2
2 0.650 0.080 0.225 48 7 3
3 0.950 0.135 0.325 38 8 4
4 1.215 0.200 0.425 33 10 5
5 1.405 0.258 0.500 31 11 6

4. A trial value of t = 0.3 is selected. Moving from .02 on the binomial axis down the
diagonal ΔZ (= t) axis, a distance 0.3 and reading over gives pg1 = .04. Similarly,
moving from .08 on the binomial axis down the diagonal ΔZ (= t) axis, a distance
0.3 and reading over gives pg2 = .135.
5. Connecting pg1 and pg2 with α and β on the probability axis and reading the Larson
grid gives n = 67, c = 5 as the plan when t = 0.3.
6. The next value of t selected is t = 0.65 and the procedure starts again.
7. The nomograph is used iteratively to produce Table 13.3. The procedure stops at
t = 1.405 since at that value pg = .50.
2

Hence, the optimum narrow-limit plan for these conditions is n = 31, c = 11, t = 1.405.
Use of the Larson nomograph constrains the Ladany procedure to values of pg ≤ .50.
Schilling and Sommers (1981) have computed tables of optimal narrow-limit plans based
on the binomial distribution through an iterative procedure not subject to this constraint.
Appendix Table T13.3 shows narrow-limit plans, which have minimum sample size tabu-
lated by producer’s quality level p1 and consumer’s quality level p2 for fixed α = .05 and
β = .10. Also shown are matched binomial attributes plans and the single variables plans,
which appear in Appendix Table T10.2. All the plans in Appendix Tables T10.2 and T13.3
are matched and tabulated using the same values of p1 and p2. In assessing narrow-limit
plans, comparison should be made with known standard deviation variables plans since
the standard deviation is assumed known for both procedures.
As an example of the use of Appendix Table T13.3, consider the example used with the
Ladany nomograph:

p1 = .02, α = .05
p2 = .08, β = .10

These specifications result in the following plans:

Attributes: n = 97, c = 4
Narrow limit: n = 31, c =15, t = 1.69
Variables (σ known): n = 21, k = 1.69
Variables (σ unknown): n = 50, k = 1.69
352 Acceptance Sampling in Quality Control

This narrow-limit plan differs slightly from that given by Ladany since, for the plan devel-
oped from the nomograph, the α and β risks are not held exactly because of the constraint
pg2 ≤ .050. For the Ladany plan, α = .079 and β = .075, whereas using the Schilling–Sommers
table, α = .052 and β = .101.
In this example, the narrow-limit plan affects a two-thirds reduction in sample size rela-
tive to attributes compared to an 80% reduction using the variables plan. Advantages of
narrow-limit plans over variables are

1. No calculations for the inspector

2. Ease and accuracy in collecting the data
3. Ease of understanding and use

Of course, less information is generated by the narrow-limit plans for possible feedback in
acceptance control.
Schilling and Sommers (1981) found that a simple heuristic approximation can be used
to develop an optimal narrow-limit plan from the known standard deviation variables
plan having the same p1, p2, α, and β. If the variables plan has sample size nv and acceptance
constant k, the parameters

n = 15nv , t = k
c = .75nv - .67

provide an excellent approximation to the optimal narrow-limit plan. This can be con-
firmed from the results of the preceding example. Using this procedure, an approximation
of the known standard deviation plan n = 21, k = 1.69 is

n = 31.5 ~ 32, t = 1.69

c = 15.08 ~ 15

which is very close to the optimal narrow-limit plan.

In the tables of optimal narrow-limit plans matching the MIL-STD-105E (United States
Department of Defense 1989), normal, tightened, and reduced tables are also presented
by Schilling and Sommers (1981). This allows use of these narrow-limit plans as substi-
tutes for the attributes plans given in the standard when the assumptions of narrow-limit
gauging are met. Use of the MIL-STD-105E AQL system and its switching rules allow for
significant reductions in scheme sample size.
The tables of optimal narrow-limit plans matching MIL-STD-105E are given here as
follows:

Appendix Table T13.4: 

Tightened inspection optimal narrow-limit plans for
MIL-STD-105E
Appendix Table T13.5: 
Normal inspection optimal narrow-limit plans for
MIL-STD-105E
Appendix Table T13.6: 
Reduced inspection optimal narrow-limit plans for
MIL-STD-105E

The OC curves of the resulting plans closely follow those of the counterpart attributes
plans from MIL-STD-105E. Thus, when substituted for the attributes plans, the OCs and
Special Plans and Procedures 353

other measures of the narrow-limit plans are essentially the same as those given in that
standard. The following tables from MIL-STD-105E can be used directly to assess their
properties.

MIL-STD-105E Table V-A: Average outgoing quality (AOQ) limit factors for normal
inspection
MIL-STD-105E Table V-B: AOQ limit factors for tightened inspection
MIL-STD-105E Table VI-A: Limiting quality for which Pa = 10%
MIL-STD-105E Table VII-A: Limiting quality for which Pa = 5%
MIL-STD-105E Table X-A: Tables for sample size code letter

The MIL-STD-105E average sample size Table IX is not represented among these tables
since average sample sizes using narrow-limit plans will be much less than those given
in MIL-STD-105E, and further, the narrow-limit plans shown are for single sampling only.
When the AQL sampling scheme, which MIL-STD-105E represents, is properly used (with
the switching rules), the average sample number (ASN) for the overall scheme using
the narrow-limit plans can be computed. This has been tabulated using the approach of
Schilling and Sheesley (1978) for the overall tightened–normal–reduced scheme except
that the limit numbers for switching to reduced inspection were not utilized in the tabula-
tion. The resulting average sample sizes are shown in Appendix Table T13.7 for the case
when the process is running at the AQL. Probabilities of acceptance for the scheme at the
AQL are also shown in Appendix Table T13.7.
When the MIL-STD-105E system is applied using narrow-limit plans, the switching rules
and other procedures may be used directly. Use of the limit numbers in switching to reduced
inspection poses a problem, however, in that gauging is to the narrow limit and not to the
specification limit. Accordingly, units not conforming to the narrow limit would have to be
regauged to determine the number of defectives (or nonconformances to the specification
limit) in the sample to compare to the limit numbers. Also, the sample sizes are reduced
to such an extent by using narrow-limit plans that would take considerably more than 10
lots to accumulate a sample large enough to use the limit numbers for reduced inspection
in Table VIII of MIL-STD-105E. It is recommended that the limit numbers be dropped from
the switching procedure. As stated by Schilling and Sheesley (1978), “The effect of the limit
numbers for reduced inspection on the operating characteristics is minimal. Yet they serve
as an impediment to easy use of the switching rules.”
Use of the switching rules with narrow-limit plans can result in a significant decrease in
average sample size. For example, with Code M, 1.5% AQL, the sample size for attributes
plans drops from 315 for the normal plan alone to 268 for the scheme with the switching
rules. When narrow-limit plans are substituted in the scheme, the average sample size drops
even further from 79 for the normal plan alone to 50.5 when the switching rules are used.
As an example of the use of narrow-limit plans in the MIL-STD-105E system, consider
the plan Code F, 2.5% AQL. A comparison of MIL-STD-105E attributes plans with their nar-
row-limit counterparts is shown in Table 13.4. Here, the scheme average sample size is 21.5
at the AQL using the attributes plans and 8.6 when the narrow-limit plans are substituted.
It should be noted that the acceptance criteria for reduced plans under MIL-STD-105E
show a gap between the acceptance and rejection numbers. Sample results falling in this
gap initiate a switch back to normal inspection although the lot itself is accepted under
the reduced plan. When the tables for the narrow-limit plans were prepared, the plan at
the attributes rejection number was matched at Pa = .95 and Pa = .10 and made optimum.
354 Acceptance Sampling in Quality Control

TABLE 13.4
Narrow-Limit Plans Substituted for Attributes Plans in MIL-STD-105E,
Code F, 2.5 AQL
Attributes as Given Narrow-Limit Counterparts
Normal n = 20, Ac = 1, Re = 2 n = 9, Ac = 4, Re = 5, t = 1.43
Tightened n = 32, Ac = 1, Re = 2 n = 11, Ac = 5, Re = 6, t = 1.67
Reduced n = 8, Ac = 0, Re = 2 n = 6, Ac = 1, Re = 4, t = 1.07

The plan for the corresponding attributes acceptance number was then matched as closely
as possible at Pa = .10 using the sample size, n, and compression constant, t, from the plan
derived from the rejection number.
The assumption of normality upon which the narrow-limit plans presented are based
is an important consideration in application. Preliminary investigation by Schilling and
Sommers (1981) showed increasing sensitivity to the assumption with small p (large t). The
risks may differ considerably from those specified by the plan depending on the degree of
nonnormality. The standard deviation must, of course, be known and stable.
As an extreme illustration, suppose the plan n = 31, c = 15, t = 1.69 was set up to be used
with a standard normal distribution of product. This implies p1 = .02 has .95 probability of
acceptance and p2 = .08 has .10 probability of acceptance. If the distribution subsequently
changed to that of a t-distribution with one degree of freedom (i.e., the symmetrical thick-
tailed Cauchy distribution) with an appropriate location parameter and an interquartile
range the same as the assumed normal distribution, p1 = .103 would have .95 probability
of acceptance while p2 = .139 would have .10 probability of acceptance. Thus, it is very
important that the normal assumption be verified and monitored in the use of narrow-
limit plans.
With variables sampling, the sampling data currently obtained could be used to set up
control charts for checking on known variability and the continued validity of the normal-
ity assumption. Control charts using gauging techniques have been discussed by Ott and
Mundel (1954) and Stevens (1948).
Narrow-limit plans provide an excellent vehicle for sample size reduction when prop-
erly used in applications in which a normal distribution is assumed and where σ has been
accurately estimated. Their use with the MIL-STD-105E scheme switching rules can lead to
still further reductions in sample size and utilization of that standard in situations in which
the attributes sample sizes required by the standard would be prohibitive. They provide a
useful and viable alternative in a continuing effort to attain maximum quality at minimum
costs. In the words of Ott and Mundel (1954), “The advantages which are inherent in a
program of quality control require an appreciation of its philosophy, an understanding of
its techniques, and provision for competent management of the program.” Used in such an
environment, narrow-limit plans are an excellent tool for quality assessment and control.

Mixed Variables: Attributes Plan

The choice between acceptance sampling by attributes and variables has commonly been
considered a first step in the application of sampling plans to specific problems in an
industry. The dichotomy is more apparent than real, however, since other alternatives exist
Special Plans and Procedures 355

in the combination of both attributes and variables results to determine the disposition of
the lot. One such procedure is the so-called mixed variables–attributes sampling plan. It is,
in essence, a double-sampling procedure involving variables inspection of the first sample
and subsequent attributes inspection if the variables inspection of the first sample does not
lead to acceptance.
As early as 1932, Dodge (1932) suggested that variables criteria be used in the first stage
(only) of a double-sampling plan “…for judging the results of a first sample and for deter-
mining when a second, substantially large sample should be inspected before rejecting the
lot.” Such procedures are now called mixed or variables–attributes sampling plans. The
double-sampling feature distinguishes these plans from single-sampling plans using both
variables and attributes criteria as proposed by Woods (1960) and Kao (1966).
Mixed variables–attributes sampling differs from the ordinary double-sampling proce-
dure in the sense that only acceptance can take place as a result of the application of the
variables plan to the first sample. If acceptance is not indicated, a second sample is drawn,
acceptance or rejection then being determined on an attributes basis. Use of variables on
the first sample with attributes on the second sample combines the economy of variables
for quick acceptance on the first sample with the broad nonparametric protection of attri-
butes sampling when a questionable lot requires a second sample. Schematically, the pro-
cedure is shown in Figure 13.7.
Mixed plans are of two types, so-called independent and dependent plans. Independent
mixed plans do not incorporate first sample results in the assessment of the second sample,
that is, decisions on the two samples are kept independent. Dependent mixed plans com-
bine the results of the first and second samples in making a decision if a second sample
is necessary; thus, the second sample decision is dependent on the first sample results. In
describing mixed variables–attributes plans, Bowker and Goode (1952, p. 8) indicate that
“Under this procedure, a sample is drawn and inspected on a variables basis…if the action
indicated is rejection an additional sample is drawn. This additional sample is inspected
on an attribute basis, and the final decision concerning disposal of the lot is made on the
basis of the attribute plan.”
Their discussion of mixed plan is, for the most part, limited to the independent case;
that is, to plans in which the attributes procedure, when called for, is applied to the results
of the second sample only. This keeps the probabilities of acceptance of the variables and
attributes components of the plan independent. Schilling (1966) has provided procedures
for deriving independent mixed plans given two points on the OC curve. In contrast, the
so-called dependent mixed plan is one in which attributes data arising from both the first

Accept

Take Pass Accept

Variables
sample Ι test Fail Pass
Take Attributes
sample ΙΙ test Fail
Reject

FIGURE 13.7
Operation of a mixed plan. (From Schilling, E.G., Mixed variables-attributes sampling, the independent case,
Transactions of the 18th Annual Conference on Quality Control at Rutgers, The State University, New Brunswick, NJ,
1966, p. 83.)
356 Acceptance Sampling in Quality Control

and second samples are combined for testing when the attributes procedure is employed.
This makes the probabilities of acceptance of the variables and attributes parts of the plan
dependent. Dependent plans have been examined by Gregory and Resnikoff (1955), Savage
(1955), and Schilling and Dodge (1969).
Independent mixed plans maintain stochastic independence between the probabilities of
the variables and attributes constituents of the procedure. Bowker and Goode (1952) sug-
gest that independent plans have conventionally been carried out as follows:

1. Obtain first sample.

2. Test the first sample against a given variables-acceptance criterion.
a. Accept if the test meets the variables criterion.
b. Resample if the test fails to meet the variables criterion.
3. Obtain a second sample if necessary (per 2b).
4. Test the second sample (only) against a given attributes criterion and accept or
reject as indicated by the test.

Dependent mixed plans are those in which the probabilities of the variables and attributes
constituents of the procedure are made dependent. The dependent procedure, as proposed
by Savage (1955), can be summarized as follows:

1. Obtain first sample.

2. Test the first sample against a given variables-acceptance criterion.
a. Accept if the test meets the variables criterion.
b. If the test fails to meet the variables criterion,
i. Reject if the number nonconforming in the first sample exceeds a given
attributes criterion.
ii. Otherwise resample.
3. Obtain a second sample if necessary (per 2bii).
4. Test the results for the first and second samples taken together against the given
attributes criterion and accept or reject as indicated by the test.

Note that this procedure can be generalized by providing for the use of different attributes
criteria in steps 2 and 4. Such a generalized dependent mixed plan has been presented by
Schilling and Dodge (1967a).
The dependent plan provides the optimal procedure in terms of the size of ASN associ-
ated with the plan. Attention will be directed here to mixed plans for the case of single
specification limit, known standard deviation, when a normal distribution of product is
assumed. Gregory and Resnikoff (1955) have examined the case of dependent plans with
standard deviation unknown, while Bowker and Goode (1952) have provided an approxi-
mation useful in estimating the OC curve of such plans. Adams and Mirkhani (1976) have
derived an approach to standard deviation unknown when c = 0 and examined the effect
of nonnormality on combined variables–attributes plans.

Advantages and Disadvantages of Mixed Plans

The assumption of normality inherent in most variables-acceptance procedures has proved
both their strength and their undoing. Perturbations in the production process or screening
Special Plans and Procedures 357

of nonconforming product may make otherwise normally distributed product anything

but normal. Whatever the potential source of nonnormality, the possibility of submission
of such product to standard variables plans is a serious consideration weighing against
their use except under conditions where normality is well assured. Nonetheless, the reduc-
tion in sample size attendant with variables plans makes them particularly inviting.
The mixed variables–attributes plan achieves some of the reduction of sample size asso-
ciated with a variables plan without some of the related disadvantages. The mixed proce-
dure appeals to the psychology of inspectors by giving a questionable lot a second chance.
In rejecting lots, it is also often a decided psychological or legal advantage to show actual
defectives to the producer, a feature that can be had only by rejecting on an attributes
basis. Truncated and nonnormal distributions cannot be rejected for poor variables results
alone, but only on the basis of defective or nonconforming units found in the attributes
sample. Furthermore, with regard to acceptance–rejection decisions, the effect of changes
in shape of distribution can be minimized by accepting only on variables evidence so
good as to be practically beyond question for most distributions which might reasonably
be presented to the plan. Thus, mixed plans provide a worthwhile alternative to variables
plans used alone.
The principal advantage of a variables–attributes plan over attributes alone is a reduc-
tion in sample size for the same protection. The variables aspect of the mixed plan also
allows for a far more careful analysis of the distribution of product presented to the plan
than would be possible with attributes inspection alone. Variables control charts kept on
these data can provide information on the variability and stability of product from lot to
lot. Control charts should normally be used in conjunction with acceptance sampling pro-
cedures involving variables inspection.
With small first samples, the mixed plan provides an excellent form of surveillance
inspection on product, which is generally expected to be of good quality but which may,
at times, show degradation. A small variables first sample can be employed to accept at
relatively low values of proportion nonconforming, and the second attributes sample is
then used to provide a definitive criterion for disposition of the lot if it is not accepted on
the first sample.
Unfortunately, mixed plans do not provide the same protection against nonnormality for
acceptance as they do for rejection since product is accepted at the first stage of the plan
on a variables basis. It is possible, however, to minimize this disadvantage for product
well within specification by designing the plan in such a way as to accept on a variables
basis only product with distribution located far enough from the specification limit so that
reasonable changes in the shape of the distribution will not cause appreciable changes in
proportion nonconforming. In this way, a tight variables criterion could be employed to
minimize the effect of changes in shape of distribution on the OC curve of the plan (see
Schilling 1967).
In application, it is also conceivable that mixed plans might be more difficult to adminis-
ter either than variables plans or than attributes plans alone. As with all plans using vari-
ables criteria, a separate mixed plan must be developed for each characteristic to which it
is applied. Any increase in complexity would, however, probably be compensated for by
the advantages of the mixed procedure.

Generalized Mixed Dependent Procedure

Given an upper specification limit, the inspection procedure for the application of a sin-
gle specification limit U, known standard deviation σ, dependent mixed plan has been
358 Acceptance Sampling in Quality Control

generalized by Schilling and Dodge (1969) by allowing for two acceptance numbers.
Symmetry obviates the necessity for parallel consideration of a lower specification limit.
The first acceptance number c1 is applied to the attributes results of the first sample after
rejection by variables and before a second sample is taken. The second acceptance number
c2 is applied to the combined first and second sample attributes results. As a special case,
the two acceptance numbers may be made the same; this is the plan proposed by Savage
(1955). Providing for the use of different acceptance numbers increases the flexibility and
potential of the dependent mixed plan. Of the several methods of specifying the variables
constituent of known standard deviation variables plans, the X method involving designa-
tion by sample size n1 and acceptance limit on the sample average A is used here since it
simplifies the notation somewhat. Note that A = U − kσ for upper specification limit and
standard variables-acceptance factor k.
Let

N be the lot size

n1 be the first sample size
n1 be the second sample size
A be the acceptance limit on sample mean (X)
c1 be the attributes acceptance number on first sample
c1 be the attributes acceptance number on second sample

Then, the generalized plan would be carried out in the following manner:

1. Determine the parameters of the mixed plan: n1, n2, A, c1, and c2.
2. Take a random sample of n1 from the lot.
3. If the sample average X ≤ A, accept the lot.
4. If the sample average X > A, examine the first sample for the number of defectives
d1 therein.
5. If d1 > c1, reject the lot.
6. If d1 ≤ c1, take a second random sample of n2 from the lot and determine the num-
ber of defectives d2 therein.
7. If in the combined sample of n = n1 + n2, the total number of defectives d = d1 + d2
is such that d ≤ c2, accept the lot.
8. If d > c2, reject the lot.

When semicurtailed inspection is employed, the procedure remains the same, except that,
if c2 is exceeded at any time during the inspection of the second sample, inspection is
stopped at once and the lot rejected.

Measures: Independent Mixed Plan

The four principal curves that describe the properties of an acceptance sampling plan for
various proportions nonconforming are the OC curve, the ASN curve, the average total
inspection (ATI) curve, and the AOQ curve. The operation of mixed plans cannot be prop-
erly assessed until these curves, for given values of the true proportion nonconforming, are
defined. In ­particular, attention will be directed here to Type B OC curves (i.e., sampling
from a process).
Special Plans and Procedures 359

Let

n1 be the first sample size

n2 be the second sample size
V′ be the probability of acceptance under variables plan (with sample size n1)
A′ be the probability of acceptance under attributes plan (with sample size n2)
p be the proportion nonconforming in process

Then, the probability of acceptance and other measures of independent mixed plans can be
developed by analogy to attributes sampling (Schilling 1966) for a lot of size N as

Pa = V ¢ + ( 1 - V ¢ ) A¢
ASN = n1 + ( 1 - V ¢ ) n2
ASN c = n1 + ( 1 - V ¢ ) ASN c*
AOQ  pPa
ATI = n1V ¢ + ( n1 + n2 ) ( 1 - V ¢ ) A¢ + ( N ) ( 1 - V ¢ ) ( 1 - A¢ )

where ASN *c is the ASN under semicurtailed inspection for the attributes plan.
It is important to note that these equations for independent plans hold whatever the
nature of the variables or attributes sampling plans involved. Any variables plan (using
range or standard deviation) can be combined with any attributes plan (single or multiple)
using the independent procedure, provided, of course, that the underlying assumptions
of the two plans are appropriate to the situation to which the mixed plan is to be applied.
Also, the probabilities of acceptance, V′ and A′, are usually readily available since they can
be read directly from the OC curves of the variables and attributes plans used.
The assumption of a known underlying distribution inherent in variables sampling
would seem to indicate sufficient knowledge of the underlying process to allow use of
known standard deviation variables plans in most applications. The possibility of a pro-
cess generating product with a distribution of constant shape, but frequent changes in
variability suggest that unknown standard deviation plans may sometimes be in order.
The appropriate selection should, of course, be made subsequent to investigation of the
stability of the distribution from lot to lot as revealed by a control chart and by examina-
tions of the shape of the distribution and its constancy. As with variables plans, mixed
plans should not be used “in the blind” with product of unknown history. Unknown stan-
dard deviation plans are easily derived and measures determined for the independent case
using this procedure. A method for assessing the operating measures of unknown stan-
dard deviation dependent mixed plans when c = 0 is given by Adams and Mirkhani (1976).
For example, consider the following independent variables–attributes plan:

Variables: n1 = 7 k = 1.44
Attributes: n2 = 20 c = 1

The probability of acceptance has previously been calculated for the two constituents of
the independent mixed plan (see Chapters 5 and 10). For example, when p = .18, it was
found for a lot size of 120 that
360 Acceptance Sampling in Quality Control

Measure Variables Plan Attributes Plan

Probability of acceptance V′ = .08 A′ = .10
ASN 7 20
ASN (semicurtailed) — 10.44
AOQ (approximate) 0.014 0.018
ATI (N = 120) 111.0 109.8
AOQ limit 0.036 0.036

So for the independent mixed plan,

Pa = .08 + ( 1 - .08 ) .10 = 0.172
ASN = 7 + ( 1 - .08 ) 20 = 25.4
ASN c = 7 + ( 1 - .08 ) 10.44 = 16.6
AOQ = pPa = .18 ( .172 ) = 0.031
ATI = 7 ( .08 ) + ( 7 + 20 ) ( 1 - .08 ) ( .10 ) + ( 120 ) ( 1 - .08 ) ( 1 - .10 ) = 102.4

Thus, although probability of acceptance and ASNs is higher for this proportion noncon-
forming, the other measures AOQ and ATI are improved over the attributes plan taken alone.

Measures: Dependent Mixed Plan

Formulas for the measures of the generalized dependent mixed plan as given by Schilling
and Dodge (1969) are shown in Table 13.5, where

P(Y) is the probability of Y

Pn(Y, W) is the probability of Y and W in a sample of n
Pn(Y|W) is the probability of Y given W in a sample of n
P(i, n) is the probability of i defectives in a sample of n
p is the population (process) fraction defective

TABLE 13.5
Formulas for Measures of Dependent Mixed Plans
Measure Formulas

( ) å å ( )
c1 c2 - i
Pa Pa = P X £ A + Pn1 i , X > A P ( j , n2 )
i =0 j =0

å ( )
c1
ASN ASN = n1 + n2 Pn1 i , X > A
i =0

éc -i+1 ù
å ( ) å å
c1 n2 +1 c2 - i
ASNc ASN c = n1 + Pn1 i , X > A ê 2 P ( k , n2 + 1) + n2 P ( j , n2 )ú
i =0
ë p k = c2 - i + 2 j =0
û

å ( )
Pn1 i , X > A + ( N - n1 - n2 ) æç 1 - Pa - å ( )
Pn1 i , X > A ö÷
n1 n1
ATI ATI = ASN + ( N - n1 )
i =c1 +1 è i =c1 +1 ø

AOQ AOQ =
p
N êë( 1 )
éP X £ A ( N - n ) + P - P X £ A
a ( ( )) ( N - n 1 - n2 ) ù
úû

Source: Reprinted from Schilling, E.G. and Dodge, H.F., Technometrics, 11(2), 344, 1969. With permission.
Note: Except for ASN, all formulas are the same with or without curtailed inspection.
Special Plans and Procedures 361

Since σ is assumed known, it is possible to evaluate the expressions shown in Table 13.5
using tables of Pn(i, X > A) for a standard normal universe, that is, μ = 0, σ = 1. Such val-
ues are given in Appendix Table T13.8 for first sample size n1 = 5. To accomplish this, the
value of Pn(i, X > A) for a particular application can be found by transforming the variates
involved to standard normal deviates by the use of the familiar z-transformation. This
expresses the departure of given values from the population mean in units of the (known)
standard deviation. Thus, an upper specification limit U is expressed as zU, where
U -m
zU =
s
and μ is the population mean of a normal distribution such that fraction defective p of the
said distribution exceeds the upper specification limit U (see Figure 13.8). Thus,

( )
Pn = i , X > A = Pn ( i , z > z A )

where z and zA are standard normal deviates such that

X -m A-m
z= zA =
s s
The tables in the appendix are entered with these values for the mean and the acceptance
limit.
The values shown in the appendix were calculated using the method given in Schilling
and Dodge (1966). Similar tables for sample sizes 4–10 are presented in Schilling and Dodge
(1967b) and for sample sizes 11–20 when c = 0 in Schilling and Dodge (1967c).
Figure 13.8 shows the relationship of k and A for a given distribution of product with
mean μ associated with fraction nonconforming p. It also displays the role of the trans-
formed variables zA and zU.
Mixed plans have been discussed in terms of the X method mentioned earlier since this
simplifies the notation somewhat. Variables plans specified in terms of the other methods
can be converted to the X method using the k method
A = U - k s

A U
Probability of
acceptance on kσ Distribution of
variables sample
sample means

Distribution of
individuals
P
Original units
μ A U x of measurement
kσ
Transformed
O zA zU z z units
k

FIGURE 13.8
Relationship of k and A. (Reprinted from Schilling, E.G. and Dodge, H.F., Technometrics, 11(2), 346, 1969. With
permission.)
362 Acceptance Sampling in Quality Control

or the M method
n -1
A =U- k s,
n
k such that
¥
1 -t2/2 M

ò
k
2p
e dt =
100

in the notation of MIL-STD-414, respectively.

In combining any two variables and attributes plans in a dependent mixed plan, the
formulas of Table 13.5 define the probability of acceptance, or OC curve, and associated
measures of the combined plan. Note that the formulas simplify greatly when c1 = 0.
To illustrate the inspection procedure to be followed and the methods to be used in deter-
mining the properties of a mixed-acceptance sampling plan, consider the following example.
Suppose the plan

n1 = 5, k = 1.5
n2 = 20, c1 = 1, c2 = 2

is to be applied to the lot-by-lot acceptance inspection of a particular kind of device. The

characteristic to be inspected is the operating temperature of the device, for which there
is a specified upper limit of 209.0°F. For this characteristic, the standard deviation of the
process is known to be 4.0°F, based on past experience substantiated by a control chart.
What inspection procedure should be followed and what are the properties of this plan?
For this example,
U = 209.0°F
s = 4.0°F
A = U - ks = 209.0 - 1.5 ( 4.0 ) = 203.0°F

The procedure would be carried out as follows:

Step Results

1. Determine parameters of plan. n1 = 5, n2 = 20, A = 203.0, c1 = 1, c2 = 2.

2. Take sample of n1 = 5 from the lot. First sample results: 205, 202, 208, 198, 207.
3. If X ≤ A, accept the lot. X = 204; not ≤ A = 203.0, so go to the next step
4. If X > A, examine first sample for number of defectives No sample value > U = 209.0, so d1 = 0.
d1 therein.
5. If d1 > c1, reject the lot. d1 = 0, not > c1 = 1, so go to the next step.
6. If d1 ≤ c1, take second sample of n2 = 20 and determine Second sample results: 3 nonconforming in
number nonconforming or defective d2 therein. n2 = 20, so d2 = 3.
7. If in combined sample, total next nonconforming or d = d1 + d2 = 0 + 3 = 3, not ≤ c2 =2, so go to next
defective d = d1 + d2 ≤ c2, accept the lot. step.
8. If d > c2, reject the lot. d = 3 > c2 = 2; reject the lot.

Suppose the probability of acceptance Pa and associated measures are to be calculated for
fraction nonconforming p = .02. Then, since the distribution is normal, p = .02 implies the
Special Plans and Procedures 363

μ A U

kσ = 6.0

p= 0.02

Original units
x

200.8

209.0
of measurement

203.0
Transformed
z

0.55
0
units

2.05
k = 1.5

FIGURE 13.9
Distribution of individuals when p = .02, known σ = 4.0. (Reprinted from Schilling, E.G. and Dodge, H.F.,
Technometrics, 11(2), 341, 1969. With permission.)

distribution of individuals will be as indicated in Figure 13.9, and from a normal probabil-
ity table, we find zU = 2.05 for p = .02. Thus,

z A = zU - k = 2.05 - 1.5 = 0.55

Then, the following are the probability of acceptance and associated measures of the plan
given earlier:

1. Probability of acceptance (at p = .02)

c1 c2 - i

(
Pa = P X £ A + ) ååP
i =0 j =0
n1 ( i, X > A ) P( j; n ) 2

2 1

( )
= P z £ n1 z A + P5 ( 0, z > z A ) å P ( j ; 20 ) + P5 (1, z > z A ) åP ( j; 20 )
j =0 j =0
2 1

( )
= P z £ 5 ( 0.55 ) + P5 ( 0, z > 0.55 ) å P( j ; 20) + P5 ( 1, z > 0.55 ) å ( j; 20 )
j =0 j =0
= .8907 + .0693 ( .9929 ) + .037 ( .9401)
= .9943

2. ASN (at p = .02)

c1

ASN = n1 + n2 åP
i =0
n1 ( i, X > A )
1

= 5 + 20 åP ( i, z > z
i =0
5 A )
1

= 5 + 20 åP ( i, z > 0.55)
i =0
5

= 5 + 20 [.0693 + .037 ]
= 7.126
364 Acceptance Sampling in Quality Control

3. ASN under semicurtailed inspection (at p = .02)

c1 é c - i + 1 n2 +1 c2 - i ù
ASN c = n1 + å (
Pn1 i , X > A ê 2
ê
)
p å å
P ( k ; n2 + 1) + n2 P ( j ; n2 ) ú
ú
i =0 ë k = c2 - i + 2 j =0 û
1 é 2 - i + 1 20 +1 2-i ù
= 5+ å
i =0
P5 ( i , z > 0.55 ) ê å
ê .02 k = 2 - i + 2
ë
å
P ( k ; 20 + 1) + 20 P ( j ; 20 ) ú
j =0
ú
û
é 3 - 0 21 2-0 ù
= 5 + .0693 ê
ê .02 k = 4 - 0 å å
P ( k ; 21) + 20 P ( j ; 20 ) ú
ú
ë j =0 û
é3 -1 21 2 - 1 ù
+ .037 ê
ê .02 k = 4 -1
ë
å j =0
å
P ( k ; 21) + 20 P ( j ; 20 ) ú
ú
û
= 5 + .0693 éë150 ( .0007 ) + 20 ( .9929 ) ùû + .037 éë100 ( .0081) + 20 ( .9401) ùû
= 7.109

4. ATI for lot size N = 1000 (at p = .02)

n1 æ n1 ö
ATI = ASN + ( N - n1 )
i = c1 + 1
å ( )
Pn1 i , X > A + ( N - n1 - n2 ) ç 1 - Pa -
ç
è i = c1 + 1
å (
Pn1 i , X > A ÷ ) ÷
ø
5
æ 5
ö
= 7.126 + ( 1000 - 5 ) å
P5 ( i , z > 0.55 ) + ( 1000 - 5 - 20 ) ç 1 - .9943 -
ç
è
å P5 ( i , z > 0.55 ) ÷
÷
ø
i = 1+ 1 i = 1+ 1

but

5 1

åP ( i, X > A ) = P ( X > A ) - åP ( i, X > A )

i=2
5
i =0
5

5 1

å (
P5 ( i , z > 0.55 ) = P z > 5 ( 0.55 ) - ) åP ( i, z > 0.55)
5
i=2 i =0
= .1093 - ( .0693 + .0370 )
= .0030

so

ATI = 7.126 + 995 ( .003 ) + 975 ( 1 - .9943 - .003 )

= 7.126 + 2.985 + 2.632
= 12.743
Special Plans and Procedures 365

5. AOQ for lot size N = 1000 (at p = .02)

AOQ =
p é
Në
( ) ( ( ))
P X £ A ( N - n1 ) + Pa - P X £ A ( N - n1 - n2 ) ù
û

=
.02 é
1000 êë
( ) ( (
P z £ n1 z A ( 1000 - 5 ) + .9943 - P z £ n1 z A)) (1000 - 5 - 20 )ùúû
= .00002 éë.8907 ( 995 ) + ( .9943 - .8907 )( 975 ) ùû

= .0197

MIL-STD-414 Dependent Mixed Plans

Dependent mixed variables–attributes plans are, in fact, specified for use in MIL-STD-414
in paragraphs A9.2.2 to A9.4.2 reproduced as follows:

A9.2.2 Mixed variables–attributes inspection. Mixed variables–attributes inspection is

inspection of a sample by attributes, in addition to inspection by variables already
made of a previous sample, before a decision as to acceptability or rejectability of
a lot can be made.
A9.3 Selection of sampling plans. The mixed variables–attributes sampling plan shall be
selected in accordance with the following:
A9.3.1 Select the variables sampling plan in accordance with Section B, C, or D.
A9.3.2 Select the attributes sampling plan from MIL-STD-105 using a single sam-
pling plan and tightened inspection. The same AQL value(s) shall be used for
the attributes sampling plan as used for the variables plan of paragraph A9.3.1.
( Additional sample items may be drawn, as necessary, to satisfy the requirements
for sample size of the attributes sampling plan. Count as a defective each sample
item falling outside of specification limit[s].)
A9.4 Determination of acceptability. A lot meets the acceptability criterion if one of the
following conditions is satisfied:
Condition A. The lot complies with the appropriate variables acceptability criterion
of Section B, C, or D.
Condition B. The lot complies with the acceptability criterion of MIL-STD-105.
A9.4.1 If Condition A is not satisfied, proceed in accordance with the attributes
sampling plan to meet Condition B.
A9.4.2 If Condition B is not satisfied, the lot does not meet the acceptability
criterion.

To illustrate the method for determining the OC curve of a combination of two such plans,
suppose the following two plans are combined after the manner of MIL-STD-414:
MIL - STD - 414, Code F ( AQL = 4.0 ) : n = 5, k = 1.20
MIL-STD-105D (United States Department of Defense, 1963), Code F (AQL = 4.0
­tightened): n = 20, c = 1
366 Acceptance Sampling in Quality Control

Note that in combining these published plans in the manner of MIL-STD-414, the second
sample size is n2 = 15 in the calculations since five units are contributed by the first sample
to the attributes determination.
Let c1 = c2 = 1. The combined Type B OC curve would be derived as follows:

1. The formula is
1 1- i

(
Pa = P X £ A + ) ååP ( i, X > A ) P ( j; 15)
i =0 j =0
5

2. Computation then proceeds in the same manner as the previous example. For
example, to obtain the probability of acceptance when p = .05,
1

( )
Pa = P z £ n1 z A + P5 ( 0, z > z A ) åP ( j; 15) + P (1, z > z 5 A ) P ( 0; 15 )
j =0
1

( )
= P z £ 5 ( 0.44 ) + P5 ( 0, z > 0.44 ) åP ( j; 15) + P (1, z > 0.44 ) P ( 0; 15)
5
j =0

= .8374 + .0649 ( .8290 ) + .079 ( .4633 )

= .928

Comparison of Independent and Dependent Mixed Plans

A comparison can be made between independent and dependent plans that have essen-
tially the same OC curve. A criterion for comparison is the ASN of the two plans. The prob-
ability of acceptance and ASN of an independent mixed plan can be calculated as
c2

(
Pa = Pn1 X £ A + Pn1 X > A ) ( ) åP ( j; n ) 2
j =0

ASN = n1 + n2Pn1 X > A ( )

Now, it can be shown (Schilling and Dodge 1967a) that if the two plans have the same first
stage variables plan and attributes acceptance number c2 (where for the dependent plan
c1 ≤ c2), the second sample size of the independent plan will be greater than that of the
dependent plan.
Therefore, for the same probability of acceptance, that is, the same OC curve, the inde-
pendent plan requires a larger second sample size. But even if the second sample size of
the dependent plan is kept the same as that of the independent plan, the ASN of the depen-
dent plan will be lower since

ASN (Independent) ≥ ASN (Dependent)

c1

(
n1 + P X > A n2 ) ≥ n1 + n2 åP ( i, X > A )
n1
i =0

c1

(
P X³A ) ≥ åP ( i, X > A )
i =0
n1
Special Plans and Procedures 367

Thus, the dependent plan is superior to the independent plan in terms of the same pro-
tection with a smaller sample size.
The difference in ASN can become quite large if particularly bad quality is submitted to the
plan and if, as seems customary, the independent plan has no provision for rejection on an
attributes basis immediately after taking the first sample and before taking the second sample.
Thus, in the event of poor quality the attributes plan is utilized to a greater extent in the inde-
pendent scheme than in the dependent procedure with further possible increase in the ASN.
As an example of the superiority of dependent plans, consider the following:

n1 = 5, k = 2, n2 = 20, c1 = c2 = 0

The probability of acceptance and ASNs was calculated for the specified mixed plan,
assuming it to be carried out in dependent and independent forms. A comparison of the
results for the dependent and independent procedures is shown in Table 13.6.

Comparison of Mixed with Other Type Plans

As an indication of the relative merit of mixed plans, variables plans and single- and dou-
ble-sampling attributes plans were matched as closely as possible by Schilling and Dodge
(1967a) to the same dependent mixed plan

n1 = 5, k = 2,
n2 = 20, c1 = c2 = 0

at the two points

p1 = 0.008, Pa = 0.953
p2 = 0.107 , Pa = 0.098

which lie on the OC curve of the mixed plan. Because of inherent differences in the shape
of the various OC curves, exact matches could not be obtained; however, all the plans
obtained show probability of acceptance within ±0.015 of the mixed plan at these points.
The results are shown in Table 13.7.

TABLE 13.6
Comparison of Pa and Average Sample Number for a Specified Mixed Plan
Applied in Dependent and Independent Forms: n1 = 5, n2 = 20, k = 2, c1 = c2 = 0
Dependent Independent
p Pa ASN Pa ASN

0.005 0.980 6.7 0.991 6.9

0.010 0.931 8.9 0.958 9.6
0.020 0.794 12.5 0.849 14.1
0.050 0.415 16.4 0.493 20.8
0.100 0.119 15.8 0.169 23.9
0.150 0.032 13.6 0.054 24.7
0.200 0.008 11.5 0.016 24.9
368 Acceptance Sampling in Quality Control

TABLE 13.7
Comparison of Various Plans to Match p1 = .008, Pa = .953; p2 = .107, Pa = .098
Probability of Acceptance ASN
Plan Criteria p = .008 p = .107 p = .008 p = .107

Dependent mixed n1 = 5, k = 2 0.953 0.098 8.1 15.5

n2 = 20, c1 = c2 = 0
Variables n = 6, k = 1.75 0.947 0.106 6 6
Attributes (single) n = 37, c = 1 0.965 0.083 37 37
Attributes (double) n1 = 21, c1 = 0 0.947 0.095 27 30.8
n1 = 42, c2 = 1

Comparison of the ASN at these points for various plans gives a rough indication of
the advantages of mixed plans against either single- or double-sampling attributes plans.
Also, it would appear that for low-percent nonconforming the ASN for the mixed plan
approaches that of the variables plan as illustrated in the following tabulation:

Probability of Acceptance ASN

p~0 p = .005 p = .01 p~0 p = .005 p = .01

Dependent mixed 1.0 .980 .931 5 6.7 8.9

Variables 1.0 .979 .922 6 6 6
Single attributes 1.0 .985 .947 37 37 37
Double attributes 1.0 .977 .922 21 25 28.2

This is reasonable since if perfect product (within the constraint of the assumption of nor-
mality) were submitted to both plans, it would be accepted on the first stage of the mixed
procedure, resulting in an ASN of 5 compared to the variables ASN of 6.

Problems
1. A Code H, 1.0 AQL, MIL-STD-414 plan is to be used in in-process inspection. If a
No-Calc plan is to be substituted, what is the plotting position of the largest value in
the sample? If the largest value is exactly at the upper specification limit, should the
lot be accepted?
2. A modified lot plot form was drawn up for 50 observations of coating weight of
instrument pins. It was found that m̂ = 15 mg and ŝ = 1 mg. The lot limit exceeded the
specification limit by 1 space. Estimate the fraction nonconforming. Cell width is 1.
Should the lot be accepted if it is important to have less than .01 nonconforming?
3. The plan n = 10, c = 0, t = 1.5 is applied to an upper specification limit U = 110. The
standard deviation is known to be σ = 6. If the largest value in the sample is 102,
should the lot be accepted?
Special Plans and Procedures 369

4. Sketch the OC curve of the narrow-limit plan n = 5, t = 1.92, c = 2 through the points for
p = .0025, .034, and .109. Compare it to MIL-STD-105E, Code F, .65 AQL. Use Poisson
and binomial probabilities. Why are the two closer using binomial probabilities?
5. Use the Ladany nomograph to obtain an optimum plan when p1 = .03, p2 = .12, α = .05,
β = .010.
6. Find tightened, normal, and reduced optimum narrow-limit plans that match those
for the MIL-STD-105E, Code H, 1.0 AQL scheme.
7. Suppose MIL-STD-414 normal and MIL-STD-105D tightened unknown standard
deviation plans for Code F, 4.0 AQL, are to be combined in a mixed sampling proce-
dure. The plans are
MIL-STD-414: n = 10, k = 1.23
MIL-STD-105E: n = 20, Ac = 1, Re = 2
Calculate the following measures for p = .04 when the plans are combined to form an
independent mixed plan:
a. Probability of acceptance
b. ASN
c. AOQ
8. To conform to the procedure for combining mixed plans recommended in MIL-STD-414,
a dependent mixed plan should be used. Compute the measures of Problem 7 for
p = .05 when the Code G, 0.65 AQL, known standard deviation plan n = 5, k = 1.88, from
MIL-STD-414 is combined with the corresponding tightened plan from MIL-STD-105E,
n = 32, c = 0, in a dependent mixed procedure. Note that for the MIL-STD-414 method,
the total combined second sample size would be 32 so that n2 = 27.
9. Using the relationship developed by Campbell (1923)

2
np0 = c +
3

compute the indifference quality level for n = 100, c = 0, 1, 2, 3, 4, 5. Compare with the
values obtained from the Schilling–Johnson table.
10. Devise a narrow-limit plan with p1 = .01, p2 = .06, α = .05, and β = .10. What are the
parameters of a matching single-sampling attributes plan?

References
Adams, R. M. and K. Mirkhani, 1976, Combined variables/attributes plans—Sigma unknown,
Transactions of the 29th Annual Technical Conference of the American Society for Quality Control,
Toronto, Ontario, Canada, pp. 292–300.
Ashley, R. L., 1952, Modification of the lot plot method, Industrial Quality Control, 8(5): 30–31.
Bowker, A. H. and H. P. Goode, 1952, Sampling Inspection by Variables, McGraw-Hill, New York.
Campbell, G. A., 1923, Probability curves showing Poisson exponential limit, Bell System Technical
Journal, 2(1): 95–113.
Chernoff, H. and G. J. Lieberman, 1957, Sampling inspection by variables with no calculations,
Industrial Quality Control, 13(7): 5–7.
370 Acceptance Sampling in Quality Control

Dodge, H. F., 1932, Statistical control in sampling inspection, American Machinist, October: 1085–1088;
November: 1129–1131.
Dudding, B. P. and W. J. Jennett, 1944, Control Chart Technique When Manufacturing to a Specification,
British Standards Institution, London, U.K.
Gregory, G. and G. J. Resnikoff, 1955, Some notes on mixed variables and attributes sampling plans,
Technical Report No. 10, Applied Mathematics and Statistics Laboratory, Stanford University,
Stanford, CA.
Kao, J. H. K., 1966, Single-sample Attri-Vari plans for item-variability in percent defective, American
Society for Quality Control Annual Technical Conference Transactions, New York, pp. 743–758.
Ladany, S. P., 1976, Determination of optimal compressed limit gaging sampling plans, Journal of
Quality Technology, 8(4): 225–231.
Larson, H. R., 1966, A nomograph of the cumulative binomial distribution, Industrial Quality Control,
23(6): 270–278.
Mace, A. E., 1952, The use of limit gauges in process control, Industrial Quality Control, 8(4): 24–31.
Ott, E. R. and A. B. Mundel, 1954, Narrow limit gaging, Industrial Quality Control, 10(5): 2–9.
Savage, I. R., 1955, Mixed variables and attributes plans: The exponential case, Technical Report
No. 23, Applied Mathematics and Statistics Laboratory, Stanford University, Stanford, CA.
Schilling, E. G., 1966, Mixed variables-attributes sampling, the independent case, Transactions of the
18th Annual Conference on Quality Control, Rutgers, The State University, New Brunswick, NJ,
pp. 82–89.
Schilling, E. G., 1967, A general method for determining the operating characteristics of mixed
­variables- attributes sampling plans, single sided specification, standard deviation known,
PhD dissertation, Rutgers, The State University, New Brunswick, NJ.
Schilling, E. G. and H. F. Dodge, 1966, On some joint probabilities useful in mixed acceptance sampling,
Technical Report No. N-26, Rutgers, The State University Statistics Center, New Brunswick, NJ.
Schilling, E. G. and H. F. Dodge, 1967a, Dependent mixed acceptance sampling plans and their
evaluation, Technical Report No. N-27, Rutgers, The State University Statistics Center,
New Brunswick, NJ.
Schilling, E. G. and H. F. Dodge, 1967b, Tables of joint probabilities useful in evaluating mixed
acceptance sampling plans, Technical Report No. N-28, Rutgers, The State University Statistics
Center, New Brunswick, NJ.
Schilling, E. G. and H. F. Dodge, 1967c, Supplement to tables of joint probabilities, Technical Report
No. N-29, Rutgers, The State University Statistics Center, New Brunswick, NJ.
Schilling, E. G. and H. F. Dodge, 1969, Procedures and tables for evaluating dependent mixed accep-
tance sampling plans, Technometrics, 11(2): 341–372.
Schilling, E. G. and J. H. Sheesley, 1978, The performance of MIL-STD-105D under the switching
rules, Journal of Quality Technology, Part 1, 10(2): 76–83; Part 2, 10(3): 104–124.
Schilling, E. G. and D. J. Sommers, 1981, Two-point optimal narrow limit plans with applications to
MIL- STD-105D, Journal of Quality Technology, 13(2): 83–92.
Shainin, D., 1950, The Hamilton standard lot plot method of acceptance sampling by variables,
Industrial Quality Control, 7(1): 15–34.
Shainin, D., 1952, Recent lot plot experiences around the country, Industrial Quality Control, 8(5): 20–29.
Stevens, W. L., 1948, Control by gauging, Journal of the Royal Statistical Society Series B, 10(1): 54–108.
United States Department of Defense, 1957, Military Standard, Sampling Procedures and Tables for
Inspection by Variables for Percent Defective (MIL-STD-414), U.S. Government Printing Office,
Washington, DC.
United States Department of Defense, 1963, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105D), U.S. Government Printing Office, Washington, DC.
United States Department of Defense, 1989, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
Woods, W. M., 1960, Variables inspection procedures which guarantee acceptance of perfectly
screened lots, Technical Report No. 47, Applied Mathematics and Statistics Laboratory, Stanford
University, Stanford, CA.
14
Series of Lots:
Rectification Schemes

Although it is impossible to inspect quality into the product, it is possible to use 100%
inspection or screening operations in such a way as to ensure with known probability that
levels of quality in lots outgoing from an inspection station will not, either individually
or on the average, exceed certain levels. Often this is done with minimum average total
inspection (ATI). The following schemes utilize this concept:

Lot tolerance percent defective (LTPD) schemes: Specify LTPD protection on each lot.
Assuming screening of rejected lots, the sampling plan is selected to make ATI a
minimum at a projected process average level of percent defective.
Average outgoing quality limit (AOQL) schemes: Specify AOQL protection for the lots.
Assuming screening of rejected lots, the sampling plan is selected to make ATI a
minimum at the projected process average level of percent defective.

In rectification schemes, a 100% inspection of rejected lots with replacement of defective

units with good ones, or equivalent screening, is assumed.
The LTPD plans are useful when the producer desires LTPD protection on individual
lots with the intention of screening rejected material in a large sequence of lots. The pur-
pose is to minimize the total amount of inspection, including the screening, by using a
plan with the lowest possible ATI. This is particularly suitable for internal in-process sam-
pling where the costs of screening can be borne by the component responsible for produc-
ing the product. Also, on final inspection these plans provide LTPD protection for the
consumer while the producer, who may do the screening, keeps the overall testing costs
to a minimum.
The AOQL plans are concerned with the series of lots as a whole. They do not focus on
individual lots but guarantee that the average outgoing quality (AOQ) will not exceed
a certain specified amount. Thus, the consumer is assured that the average quality level
received will not exceed the AOQL in the long run. The producer, or the inspection agency,
achieves minimum inspection. Again, these plans are useful in in-process or incoming
inspections to guarantee outgoing quality levels regardless of the quality coming into the
inspection station.
Note that a specification of LTPD is much more severe than that of AOQL. When operat-
ing at the LTPD level of percent defective, the AOQ will be about 10% of the LTPD speci-
fied. Thus, a 4% LTPD is much tighter than a 4% AOQL. With plans of this type, the actual
AOQ should generally be less than half to two-thirds the value of the AOQL. A 1.5% AOQL
implies that the producer must hold about a 1.0% process level to avoid excessive rejec-
tions and consequent screening.
Special procedures have been developed for application lot by lot to provide LTPD or
AOQL protection. Like the AQL in the MIL-STD-105E or MIL STD-414 AQL schemes,

371
372 Acceptance Sampling in Quality Control

the AOQL or LTPD becomes the index for such rectification schemes. Two of these, the
Dodge–Romig scheme and the Anscombe rectifying inspection scheme, will be presented
here. The selection of a simple AOQL plan will be discussed first.

Single-Sampling AOQL Plan

It is sometimes desirable to quickly select and compare AOQL plans without regard to
process average levels or minimization of inspection. For a single-sampling attributes
plan, the AOQL can easily be determined from a graph developed by Altman (1954).

100
90
80
70 Relationship of sample size and
60 average outgoing quality limit
50 (AOQL)–single sampling (attributes)
40

30

20

10
9
8
7
6 c
c =6
5 c= =5
c= 4
AOQL (%)

4 Ac c= 3
ce 2
3 pt c=
an
ce 1
2 no
c=
1.5
0

1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3

0.2
0.15

0.1
1 2 3 4 5 6 7 8 9 10 15 20 30 40 50 70 90 150 200 300 500 700 900
60 80 100 400 600 800 1000
Sample size

FIGURE 14.1
Altman AOQL diagram. (Reprinted from Altman, I.B., Ind. Qual. Control, 10(4), 30, 1954. With permission.)
Series of Lots 373

The Altman diagram, which assumes sample size to be small relative to lot size (<10%),
is shown in Figure 14.1. It is based on the Poisson distribution. The diagram allows the
comparison of n, c, and AOQL for various plans to achieve the combinations desired. For
specified AOQL and sample size, the diagram gives the appropriate acceptance number, c.
For example, suppose limitations on inspection staff are such that a sample size of about
20 is deemed feasible while an AOQL of 4% is desired. Lot size is large. Cross-reference of
these two criteria on the graph indicates that an acceptance number of c = 1 is appropriate.
The plan becomes n = 20, c = 1.

Dodge–Romig Sampling Scheme

Rectification schemes stand among the earliest examples of sampling schemes as such.
They precede the development of AQL schemes by well over a decade. The celebrated
Dodge–Romig (Dodge and Romig 1941) tables are an excellent example of such early
efforts. Developed by Harold F. Dodge and Harry G. Romig at Bell Telephone Laboratories
in the late 1920s through the 1930s, they were first published in the Bell System Technical
Journal in the early 1940s. The tables were published in book form in 1944 with the revised
second edition coming forth in 1959.
There are two sets of tables:

1. LTPD single- and double-sampling plans that minimize ATI for values of LTPD =
0.5%, 1.0%, 2.0%, 3.0%, 4.0%, 5.0%, 7.0%, 10.0%
2. AOQL single- and double-sampling plans that minimize ATI for values of AOQL =
0.1%, 0.25%, 0.5%, 0.75%, 1.0%, 1.5%, 2.0%, 2.5%, 3.0%, 4.0%, 5.0%, 7.0%, 10.0%

The LTPD tables are set up to minimize ATI based on Type B probabilities while maintaining
LTPD protection (shown as pt%) determined from Type A probabilities since the lot size is
specified. The AOQL tables utilize Type B probabilities in determining both ATI and AOQL.
Thus, the same plan may appear in different lot size ranges in the AOQL and LTPD tables. The
disparity represents the difference in use of the scheme in protecting individual lots (Type A)
or providing protection on the process producing the lots (Type B). The Type A probability
used applies to the middle of the lot size range shown and so is exact for that value only.
LTPDs are always calculated using Type A probabilities in the Dodge–Romig scheme.
Both sets of tables require knowledge of the process average percent defective to achieve
an optimum plan to minimize ATI. This implies that the producer, or the incoming inspec-
tion station, must keep adequate records and control charts to properly assess the process
average.
The tables are set up such that if the process average is not known, they can be entered
at the highest level of process average percent defective shown in the table. In such a
­situation, the protection desired will be guaranteed with a somewhat less than optimum
plan until the necessary information can be developed.
In general, with rectification sampling plans, larger lot sizes result in less overall inspec-
tion. Too large a lot size, of course, may preclude effective random sampling. Also, such
plans may actually provide an incentive to improve the quality by forcing the producer to
incur screening costs on lots of poor quality.
374 Acceptance Sampling in Quality Control

It is important to note that the Dodge–Romig tables contain an appendix with an excel-
lent collection of binomial operating characteristic (OC) curves for some of the most
­commonly used plans, including

c = 0 , n = 2 ( 1) 15, 16 ( 2 ) 34 , 35 ( 5 ) 50, 50 ( 10 ) 100, 100 ( 20 ) 200, 200 ( 50 ) 500

c = 1, n = 3 ( 1) 20, 20 ( 2 ) 50, 50 ( 5 ) 95, 90 ( 10 ) 200, 200 ( 20 ) 500

c = 2, n = 5 ( 1) 20, 20 ( 2 ) 36, 30 ( 5 ) 100, 100 ( 10 ) 160, 160 ( 20 ) 280, 250 ( 50 ) 500

c = 3, n = 8 ( 1) 19, 20 ( 2 ) 36, 35 ( 5 ) 75, 70 ( 10 ) 200, 200 ( 20 ) 300, 300 ( 50 ) 500

where X(Y)Z indicates that the curves start at X and progress in increments of Y up to Z.
OC curves of the AOQL single- and double-sampling plans specified in the tables are also
given.

Operation
All the plans contained in the Dodge–Romig tables assume 100% inspection of rejected
lots. For plans indexed by LTPD or AOQL, the operation of the scheme is as indicated in
Figure 14.2.
The Dodge–Romig work is more than just tables. It describes the mathematical develop-
ment behind the plans presented together with much practical material on the application
of sampling plans and the meaning of OC curves and other measures. The content is struc-
tured as indicated in Figure 14.3.

Selection
The tables are indexed in two sets by LTPD or AOQL, respectively. Plans are selected on the
basis of lot size and process average. The sets of tables are entered at the specific value of
LTPD or AOQL and the plan determined by cross-referencing process average and lot size.
If the process average is not known, the largest value of process average appearing in the
table is used until adequate information can be developed. Plans are available for single or

Flow of
lots

Apply
sampling Screen lot
plan and ship

Accept Reject

FIGURE 14.2
Operation of Dodge–Romig plans.
Series of Lots 375

Ch1 Ch3 Ch4

Ch2
Intoduction LTPD and AOQL Application Description and
Construction
to sampling sampling of double meaning of OC
of tables
methodology sampling curves

Index Appendix

App 3
OC curves OC curves of
of selected LTPD AOQL
tabulated single tables tables
AOQL plans sampling
plans
Double sampling

Double sampling

Double sampling
single sampling

single sampling

single sampling
App 1

App 2

App 4

App 5

App 6

App 7
FIGURE 14.3
Structure of Dodge–Romig tables.

double sampling. The tables give the sample sizes and acceptance criteria and also show
the value of the measure not specified, that is, AOQL if LTPD is specified and vice versa.
A check sequence for the selection of a Dodge–Romig plan is given in Figure 14.4.
Examples of the Dodge–Romig (Dodge and Romig 1959) tables and OC curves are shown
here as follows:

Table Dodge–Romig Table Content

Table 14.1 Appendix 6 4% AOQL single-sampling plans

Table 14.2 Appendix 7 4% AOQL double-sampling plans
Table 14.3 Appendix 4 4% LTPD single-sampling plans
Table 14.4 Appendix 5 4% LTPD double-sampling plans

To exemplify the use of these tables and figures, suppose a plan is desired having 4%
AOQL with lot size 250 and process average percent defective 1.6%.
Tables 14.1 and 14.2 give the following plans that will guarantee AOQ <4% defective
with minimum ATI.
AOQL single sampling (Table 14.1)

n = 20, c = 1, LTPD = 19%

376 Acceptance Sampling in Quality Control

Start

Estimate process
average
(or use the highest values in tables)

Determine lot size

Specify protection
desired

LTPD AOQL

Select Select
plan plan
Appendix 4

Appendix 5

Appendix 6

Appendix 7
sampling

sampling

sampling

sampling
Double

Double
Single

Single

Apply
plan

Screen
rejected
lots

FIGURE 14.4
Check sequence for selecting Dodge–Romig plan.

AOQL double sampling (Table 14.2)

Sample Size Cumulative Sample Size (n1 + n 2) Acceptance Number Rejection Number

n1 = 16 16 c1 = 0 c2 + 1 = 3
n2 = 18 34 c2 = 2 c2 + 1 = 3

LTPD = 17.4%
The rejection number for both samples in double sampling is always 1 more than the
­acceptance number for the second sample, c2. Also, the second size is not kept at a constant
ratio of the first sample size. It varies. In MIL-STD-105E, n1 = n2.
LTPD plans afford more protection on individual lots and so require larger sample sizes.
If 4% LTPD were specified (rather than 4% AOQL) with lot size 250 and process average
percent defective 1.6%, we would have from Tables 14.3 and 14.4 the following:

LTPD single sampling (Table 14.3)

n = 85, c = 1, AOQL = 0.71%

TABLE 14.1
Dodge–Romig Single-Sampling Table for Average Outgoing Quality Limit = 4.0%
Series of Lots

Process Average
0%–0.08% 0.09%–0.80% 0.81%–1.60% 1.61%–2.40% 2.41%–3.20% 3.21%–4.00%
Lot Size n c p t% n c p t% n c p t% n c p t% n c p t% n c p t%

1–10 All 0 — All 0 — All 0 — All 0 — All 0 — All 0 —

11–50 8 0 23.0 8 0 23.0 8 0 23.0 8 0 23.0 8 0 23.0 8 0 23.0
51–100 8 0 24.0 8 0 24.0 8 0 24.0 8 0 24.0 17 1 21.5 17 1 21.5
101–200 9 0 22.0 9 0 22.0 19 1 20.0 19 1 20.0 19 1 20.0 19 1 20.0
201–300 9 0 22.5 9 0 22.5 20 1 19.0 20 1 19.0 31 2 16.8 31 2 16.8
301–400 9 0 22.5 20 1 19.1 20 1 19.1 32 2 16.2 32 2 16.2 43 3 15.2
401–500 9 0 22.5 20 1 19.1 20 1 19.1 32 2 16.3 32 2 16.3 44 3 14.9
501–600 9 0 22.5 20 1 19.2 20 1 19.2 32 2 16.3 45 3 14.6 60 4 12.9
601–800 9 0 22.5 20 1 19.2 33 2 15.9 33 2 15.9 46 3 14.3 60 4 13.0
801–1,000 9 0 22.5 21 1 18.3 33 2 16.0 46 3 14.3 60 4 13.0 75 5 12.2
1,001–2,000 9 0 22.5 21 1 18.4 34 2 15.6 47 3 14.1 75 5 12.2 105 7 11.0
2,001–3,000 9 0 22.5 21 1 18.4 34 2 15.6 60 4 13.2 90 6 11.3 125 8 10.4
3,001–4,000 21 1 18.4 21 1 18.4 48 3 13.8 65 4 12.2 110 7 10.7 155 10 9.8
4,001–5,000 21 1 18.5 34 2 15.7 48 3 13.9 80 5 11.6 110 7 10.8 175 11 9.5
5,000–7,000 21 1 18.5 34 2 15.7 48 3 13.9 80 5 11.6 125 8 10.4 210 13 9.0
7,001–10,000 21 1 18.5 34 2 15.7 65 4 12.3 95 6 11.1 145 9 9.8 245 15 8.6
10,001–20,000 21 1 18.5 34 2 15.7 65 4 12.3 110 7 10.8 195 12 9.0 340 20 7.9
20,001–50,000 21 1 18.5 49 3 13.6 80 5 11.6 145 9 9.8 250 15 8.5 460 26 7.4
50,001–100,000 21 1 18.5 49 3 13.6 95 6 11.1 165 10 9.6 310 18 8.0 540 30 7.1
Source: Reprinted from Dodge, H.F. and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York,
1959, p. 202. With permission.
Notes: n, sample size; c, acceptance number; “All” indicates that each piece in the lot is to be inspected; pt, lot tolerance percent defective with a
­consumer’s risk (Pc) of 0.10.
377
 378

TABLE 14.2
Dodge–Romig Double-Sampling Table for Average Outgoing Quality Limit = 4.0%
Process Average
0%–0.08% 0.09%–0.80% 0.81%–1.20%
Trial 1 Trial 2 Trial 1 Trial 2 Trial 1 Trial 2
Lot Size n1 c1 n2 n1 + n 2 c2 p t% n1 c1 n2 n1 + n 2 c2 p t% n1 c1 n2 n1 + n 2 c2 p t%

1–10 All 0 — — — — All 0 — — — — All 0 — — — —

11–50 8 0 — — — 23.0 8 0 — — — 23.0 8 0 — — — 23.0
51–100 12 0 7 19 1 22.0 12 0 7 19 1 22.0 12 0 7 19 1 22.0
101–200 13 0 8 21 1 21.0 13 0 8 21 1 21.0 15 0 17 32 2 18.0
201–300 13 0 9 22 1 20.5 16 0 18 34 2 17.4 16 0 18 34 2 17.4
301–400 14 0 8 22 1 20.0 16 0 19 35 2 17.0 18 0 28 46 3 15.5
401–500 14 0 8 22 1 20.0 16 0 19 35 2 17.0 19 0 28 47 3 15.3
501–600 16 0 19 35 2 17.0 16 0 19 35 2 17.0 19 0 29 48 3 15.1
601–800 16 0 20 36 2 16.7 16 0 20 36 2 16.7 19 0 30 49 3 14.9
801–1,000 16 0 20 36 2 16.7 16 0 20 36 2 16.7 20 0 45 65 4 13.8
1,001–2,000 17 0 19 36 2 16.6 19 0 31 50 3 14.8 21 0 44 65 4 13.6
2,001–3,000 17 0 19 36 2 16.6 19 0 31 50 3 14.8 21 0 44 65 4 13.6
3,001–4,000 17 0 20 37 2 16.5 19 0 31 50 3 14.8 22 0 58 80 5 13.0
4,001–5,000 17 0 20 37 2 16.5 19 0 31 50 3 14.8 22 0 58 80 5 13.0
5,001–7,000 17 0 20 37 2 16.5 19 0 31 50 3 14.8 22 0 58 80 5 13.0
7,001–10,000 17 0 20 37 2 16.5 19 0 36 55 3 14.6 23 0 57 80 5 12.7
10,001–20,000 17 0 20 37 2 16.5 21 0 44 65 4 13.6 23 0 72 95 6 12.0
20,001–50,000 17 0 20 37 2 16.5 21 0 44 65 4 13.6 43 1 92 135 8 10.6
50,001–100,000 17 0 20 37 2 16.5 23 0 62 85 5 12.5 44 1 106 150 9 10.3
Source: Reprinted from Dodge, H.F. and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons,
New York, 1959, p. 215. With permission.
Trial 1: n1, first sample size; c1, acceptance number for first sample; “All” indicates that each piece in the lot is to be inspected; Trial 2: n2,
second sample size; c2, acceptance number for first and second samples combined; pt, lot tolerance percent defective with a
­consumer’s risk (Pc) of 0.10.
Acceptance Sampling in Quality Control
TABLE 14.3
Dodge–Romig Single-Sampling Table for Lot Tolerance Percent Defective = 4.0%
Series of Lots

Process Average
0%–0.04% 0.05%–0.40% 0.41%–0.80% 0.81%–1.20% 1.21%–1.60% 1.61%–2.00%
AOQL AOQL AOQL
Lot Size n c AOQL (%) n c AOQL (%) n c AOQL (%) n c (%) n c (%) n c (%)

1–35 All 0 0.00 All 0 0.00 All 0 0.00 All 0 0.00 All 0 0.00 All 0 0.00
36–50 34 0 0.35 34 0 0.35 34 0 0.35 34 0 0.35 34 0 0.35 34 0 0.35
51–100 44 0 0.47 44 0 0.47 44 0 0.47 44 0 0.47 44 0 0.47 44 0 0.47
101–200 50 0 0.55 50 0 0.55 50 0 0.55 50 0 0.55 50 0 0.55 50 0 0.55
201–300 55 0 0.57 55 0 0.57 85 1 0.71 85 1 0.71 85 1 0.71 85 1 0.71
301–400 55 0 0.58 55 0 0.58 90 1 0.72 120 2 0.80 120 2 0.80 145 3 0.86
401–500 55 0 0.60 55 0 0.60 90 1 0.77 120 2 0.87 150 3 0.91 150 3 0.91
501–600 55 0 0.61 95 1 0.76 125 2 0.87 125 2 0.87 155 3 0.93 185 4 0.95
601–800 55 0 0.62 95 1 0.78 125 2 0.93 160 3 0.97 190 4 1.00 220 5 1.00
801–1,000 55 0 0.63 95 1 0.80 130 2 0.92 165 3 0.98 220 5 1.10 255 6 1.10
1,001–2,000 55 0 0.65 95 1 0.84 165 3 1.10 195 4 1.20 255 6 1.30 315 8 1.40
2,001–3,000 95 1 0.86 130 2 1.00 165 3 1.10 230 5 1.30 320 8 1.40 405 11 1.60
3,001–4,000 95 1 0.86 130 2 1.00 195 4 1.20 260 6 1.40 350 9 1.50 465 13 1.60
4,001–5,000 95 1 0.87 130 2 1.00 195 4 1.20 290 7 1.40 380 10 1.60 520 15 1.70
5,001–7,000 95 1 0.87 130 2 1.00 200 4 1.20 290 7 1.50 410 11 1.70 575 17 1.90
7,001–10,000 95 1 0.88 130 2 1.10 230 5 1.40 325 8 1.50 440 12 1.70 645 19 1.90
10,001–20,000 95 1 0.88 165 3 1.20 265 6 1.40 355 9 1.60 500 14 1.80 730 22 2.00
20,001–50,000 95 1 0.88 165 3 1.20 295 7 1.50 380 10 1.70 590 17 2.00 870 26 2.10
50,001–100,000 95 1 0.88 200 4 1.30 325 8 1.60 410 11 1.80 620 18 2.00 925 29 2.20
Source: Reprinted from Dodge, H.F. and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 184.
With permission.
Notes: n, sample size; c, acceptance number; “All” indicates that each piece in the lot is to be inspected; pt, lot tolerance percent defective with a consumer’s risk (Pc)
of 0.10.
379
TABLE 14.4
380

Dodge–Romig Double-Sampling Table for Lot Tolerance Percent Defective = 4.0%

Process Average
0.81%–1.20% 1.21%–1.60% 1.61%–2.00%
Trial 1 Trial 2 Trial 1 Trial 2 Trial 1 Trial 2
Lot Size n1 c1 n2 n1 + n 2 c2 p t% n1 c1 n2 n1 + n 2 c2 p t% n1 c1 n2 n1 + n 2 c2 p t%

1–35 All 0 — — — 0.00 All 0 — — — 0.00 All 0 — — — 0.00

36–50 34 0 — — — 0.35 34 0 — — — 0.35 34 0 — — — 0.35
51–75 40 0 — — — 0.43 40 0 — — — 0.43 40 0 — — — 0.43
76–100 50 0 25 75 1 0.46 50 0 25 75 1 0.46 50 0 25 75 1 0.46
101–150 55 0 30 85 1 0.55 55 0 30 85 1 0.55 55 0 30 85 1 0.55
151–200 60 0 55 115 2 0.68 60 0 55 115 2 0.68 60 0 55 115 2 0.68
201–300 60 0 65 125 2 0.75 60 0 90 150 3 0.84 60 0 90 150 3 0.84
301–400 65 0 95 160 3 0.86 65 0 95 160 3 0.86 65 0 120 185 4 0.92
401–500 65 0 100 165 3 0.92 65 0 130 195 4 0.96 105 1 140 245 6 1.00
501–600 65 0 135 200 4 1.00 105 1 145 250 6 1.10 105 1 175 280 7 1.10
601–800 65 0 140 205 4 1.00 105 1 185 290 7 1.20 105 1 210 315 8 1.20
801–1,000 110 1 155 265 6 1.20 110 1 210 320 8 1.20 145 2 230 375 10 1.30
1,001–2,000 110 1 195 305 7 1.30 150 2 240 390 10 1.50 180 3 295 475 13 1.60
2,001–3,000 110 1 260 370 9 1.40 185 3 305 490 13 1.60 220 4 410 630 18 1.70
3,001–4,000 150 2 255 405 10 1.50 185 3 340 525 14 1.60 285 6 465 750 22 1.80
4,001–5,000 150 2 285 435 11 1.60 185 3 395 580 16 1.70 285 6 520 805 24 1.90
5,001–7,000 150 2 320 470 12 1.60 185 3 435 620 17 1.70 320 7 585 905 27 2.00
7,001–10,000 150 2 325 475 12 1.70 220 4 460 680 19 1.90 320 7 645 965 29 2.10
10,001–20,000 150 2 355 505 13 1.70 220 4 495 715 20 1.90 350 8 790 1140 35 2.20
20,001–50,000 150 2 420 570 15 1.70 255 5 575 830 24 2.00 385 9 895 1280 40 2.30
50,001–100,000 150 2 450 600 16 1.80 255 5 665 920 27 2.10 415 10 985 1400 44 2.40
Source: Reprinted from Dodge, H.F. and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York,
1959, p. 192. With permission.
Trial 1: 
n1, first sample size; c1, acceptance number for first sample; “All” indicates that each piece in the lot is to be inspected; Trial 2: n2, second
sample size; c2, acceptance number for first and second samples combined; pt, lot tolerance percent defective with a consumer’s risk
(Pc) of 0.10.
Acceptance Sampling in Quality Control
Series of Lots 381

LTPD double sampling (Table 14.4)

Sample Size Cumulative Sample Size (n1 + n 2) Acceptance Number Rejection Number

n1 = 60 60 c1 = 0 c2 + 1 = 4
n2 = 90 150 c2 = 3 c2 + 1 = 4

AOQL = 0.84%

This dramatically shows the difference in sample size that can result from specifying AOQL
or LTPD. It is vital to select the proper measure for the sampling situation when applying
rectification schemes of this type as with all sampling plans.
Schilling et al. (2002) have developed a set of tables of variables plans (known and
unknown standard deviation) that match the attributes plans of the Dodge–Romig tables.
A representative set is shown in Tables 14.5 and 14.6 for 4% AOQL.
As in the previous example, suppose a 4% AOQL is desired for lot sizes of 200 when the
process is running at about 1.6% defective. The matching variables plans from Tables 14.5
and 14.6 are
Variability known (Table 14.5): n = 8, k = 1.40, LTPD = 17.2%

Variability unknown (Table 14.6): n = 13, k = 1.42, LTPD = 18.9%

These plans indicate the advantage of variables plans when compared to the matched
attributes plan:

Attributes: n = 20, c = 1

Of course, the price is in the more stringent assumptions of the variables plans.

Measures
OC curves are given for all AOQL single- and double-sampling plans. The binomial OC
curves for selected single-sampling plans are also shown for reference. To illustrate
the curves, Figures 14.5 and 14.6 show the OC curves for the AOQL single- and double-­
sampling plans of the previous example. Figure 14.7 gives an example of the OC curves
from the collection of binomial OC curves given by Dodge and Romig.

Further Considerations
Sampling plans meeting the Dodge–Romig criterion for minimum ATI can be derived
using procedures developed by Dodge and Romig (1929) for LTPD plans and by Dodge
and Romig (1941) for AOQL plans. These papers provide the basis and proofs underlying
the technical development of the Dodge–Romig (Dodge and Romig 1959) tables. The tech-
nical background of these plans is interesting and informative. The Dodge–Romig plans
minimize ATI for both the LTPD and AOQL plans.
For the LTPD plans

æ e - np ( n p ) ö
C i

ATI = n + ( N - n ) ç 1 -
ç
è
å i!
÷
÷
ø
i =0
 382

TABLE 14.5
Dodge–Romig Variables Plans for Average Outgoing Quality Limit = 4.0% Variability Known
Process Average
0%–0.08% 0.081%–0.80% 0.81%–1.60% 1.61%–2.40% 2.41%–3.20% 3.21%–4.00%
Lot Size n k p t% n k p t% n k p t% n k p t% n k p t% n k p t%
1–10 1 1.58 38.1 2 1.34 33.2 2 1.34 33.2 2 1.34 33.2 2 1.34 33.2 2 1.34 33.2
11–50 2 1.46 29.1 3 1.39 25.7 4 1.36 23.5 5 1.35 21.9 6 1.34 20.8 7 1.33 20.0
51–100 2 1.47 28.7 4 1.39 22.8 5 1.38 21.1 7 1.37 18.8 9 1.37 17.4 11 1.37 16.4
101–200 2 1.48 28.5 4 1.40 22.4 7 1.39 18.3 9 1.39 16.7 12 1.40 15.2 16 1.41 13.9
201–300 3 1.42 24.7 5 1.40 20.5 8 1.40 17.2 11 1.41 15.4 15 1.42 13.8 20 1.43 12.7
301–400 3 1.42 24.7 5 1.40 20.5 8 1.40 17.1 12 1.42 14.8 17 1.43 13.1 24 1.45 11.8
401–500 3 1.43 24.7 5 1.40 20.4 9 1.41 16.4 13 1.42 14.3 19 1.44 12.6 27 1.46 11.3
501–600 3 1.43 24.7 6 1.40 19.0 9 1.41 16.3 14 1.43 13.9 21 1.45 12.1 30 1.47 10.9
601–800 3 1.43 24.6 6 1.40 19.0 10 1.41 15.6 16 1.44 13.2 24 1.46 11.5 35 1.48 10.3
801–1,000 3 1.43 24.6 6 1.40 19.0 11 1.42 15.1 17 1.44 12.9 26 1.47 11.2 39 1.49 9.9
1,001–2,000 3 1.43 24.6 7 1.41 17.8 13 1.43 14.1 21 1.46 11.9 34 1.49 10.2 57 1.52 8.8
2,001–3,000 4 1.41 22.1 8 1.41 16.9 14 1.44 13.7 24 1.47 11.3 40 1.51 9.6 70 1.54 8.3
3,001–4,000 4 1.41 22.1 8 1.41 16.9 15 1.44 13.3 25 1.48 11.1 44 1.51 9.3 81 1.55 8.0
4,001–5,000 4 1.41 22.1 8 1.41 16.9 16 1.45 13.0 27 1.48 10.8 47 1.52 9.1 91 1.56 7.7
5,001–7,000 4 1.41 22.1 9 1.42 16.1 17 1.45 12.7 29 1.49 10.6 53 1.53 8.8 107 1.57 7.4
7,001–10,000 4 1.41 22.1 9 1.42 16.1 18 1.45 12.4 32 1.49 10.2 59 1.54 8.5 126 1.58 7.1
10,001–20,000 4 1.41 22.1 10 1.42 15.5 20 1.46 12.0 37 1.51 9.8 72 1.55 8.1 173 1.60 6.6
20,001–50,000 5 1.41 20.3 11 1.43 14.9 23 1.47 11.4 45 1.52 9.2 92 1.57 7.6 254 1.62 6.2
50,001–100,000 5 1.41 20.3 12 1.43 14.4 26 1.48 10.9 51 1.53 8.9 108 1.58 7.3 331 1.63 5.9
Source: Reprinted from Schilling, E.G. et al., Qual. Eng., 14(3), 446, 2002. With permission.
Acceptance Sampling in Quality Control
TABLE 14.6
Series of Lots

Dodge–Romig Variables Plans for Average Outgoing Quality Limit = 4.0% Variability Unknown
Process Average
0%–0.08% 0.081%–0.80% 0.81%–1.60% 1.61%–2.40% 2.41%–3.20% 3.21%–4.00%
Lot Size n k p t% n k p t% n k p t% n k p t% n k p t% n k p t%

1–10 3 1.55 40.9 3 1.55 40.9 3 1.55 40.9 3 1.55 40.9 4 1.26 39.1 4 1.26 39.1
11–50 4 1.64 31.1 6 1.44 27.1 7 1.40 25.8 8 1.37 24.8 9 1.34 24.0 10 1.32 23.4
51–100 5 1.55 27.9 7 1.45 24.6 9 1.41 22.6 11 1.38 21.1 13 1.37 20.1 14 1.36 19.7
101–200 5 1.57 27.5 8 1.45 22.9 12 1.41 19.7 15 1.40 18.3 18 1.39 17.2 22 1.39 16.2
201–300 6 1.52 25.4 9 1.44 21.7 13 1.42 18.9 18 1.41 16.9 22 1.41 15.8 27 1.40 14.9
301–400 6 1.52 25.4 10 1.44 20.8 15 1.42 17.8 20 1.42 16.1 26 1.42 14.8 32 1.42 13.9
401–500 6 1.52 25.3 10 1.44 20.7 16 1.42 17.3 22 1.42 15.5 29 1.42 14.2 37 1.43 13.2
501–600 6 1.53 25.3 11 1.44 19.9 17 1.42 16.9 23 1.43 15.2 31 1.43 13.8 41 1.44 12.7
601–800 7 1.49 23.7 11 1.44 19.8 18 1.43 16.5 26 1.43 14.5 36 1.44 13.0 48 1.45 11.9
801–1,000 7 1.49 23.7 12 1.44 19.1 19 1.43 16.1 28 1.44 14.0 40 1.45 12.5 55 1.46 11.4
1,001–2,000 7 1.50 23.7 14 1.44 18.0 23 1.44 14.9 36 1.46 12.7 54 1.47 11.2 81 1.49 10.0
2,001–3,000 8 1.48 22.4 15 1.44 17.5 26 1.45 14.2 41 1.46 12.2 64 1.49 10.6 101 1.51 9.3
3,001–4,000 8 1.48 22.4 16 1.44 17.0 28 1.45 13.8 45 1.47 11.8 72 1.50 10.2 118 1.52 8.9
4,001–5,000 8 1.48 22.4 16 1.44 17.0 29 1.45 13.6 48 1.48 11.5 78 1.50 9.9 133 1.53 8.6
5,001–7,000 9 1.46 21.3 17 1.44 16.6 31 1.45 13.3 52 1.48 11.2 89 1.51 9.5 159 1.54 8.2
7,001–10,000 9 1.46 21.3 18 1.44 16.2 34 1.46 12.9 58 1.49 10.8 101 1.52 9.2 191 1.56 7.8
10,001–20,000 10 1.45 20.4 20 1.44 15.6 39 1.47 12.3 69 1.50 10.2 127 1.54 8.6 269 1.58 7.2
20,001–50,000 11 1.45 19.7 23 1.44 14.8 46 1.48 11.6 85 1.51 9.6 168 1.56 8.0 412 1.60 6.6
50,001–100,000 11 1.45 19.7 25 1.45 14.3 51 1.48 11.2 98 1.52 9.2 202 1.57 7.6 555 1.62 6.2
Source: Reprinted from Schilling, E.G. et al., Qual. Eng., 14(3), 446, 2002. With permission.
383
384 Acceptance Sampling in Quality Control

1.00
N = 201–300
0.90 n c
1. 9 0
0.80 3. 2. 20 1
3. 31 2
1.
Probability of acceptance 0.70

0.60

0.50

0.40

0.30 1.
2.
0.20 3.

0.10

0
0 5 10 15 20 25
Defective (%)

FIGURE 14.5
OC curves, single-sampling plans: AOQL = 4.0%. (Reprinted from Dodge, H.F. and Romig, H.G., Sampling
Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 94. With permission.)

1.00
N = 201–300
5.
0.90
n1 n2 c1 c2
1.
1. 13 9 0 1
0.80 2. 16 18 0 2
3. 17 28 0 3
0.70 4. 18 37 0 4
5. 33 47 1 6
Probability of acceptance

0.60

0.50

0.40

0.30 1.
2.
0.20 3.
4.
0.10 5.

0
0 10 20 30 40
Defective (%)

FIGURE 14.6
OC curves, double-sampling plans: AOQL = 4.0%. (Reprinted from Dodge, H.F. and Romig, H.G., Sampling
Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 151. With permission.)
Series of Lots 385

1.00

0.90 n = 35–50
c=0

0.80

0.70
Probability of acceptance

0.60

0.50

0.40

35
0.30
50

0.20

0.10

0
0 2 4 6 8
Defective (%)

FIGURE 14.7
OC curves, single-sampling plans: acceptance number, c = 0. (Reprinted from Dodge, H.F. and Romig, H.G.,
Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 173. With
permission.)

is minimized subject to

CiNpt Cn -(i
C N 1- pt )

.10 = åi =0
CiN

Thus, the LTPD is calculated using (Type A) hypergeometric probabilities (or approxima-
tions thereto) since LTPD is on individual lots. The AOQL, however, is determined using
the (Type B) Poisson approximation to the binomial distribution since AOQL has meaning
only in terms of a series of lots from a process.
For the AOQL plans, it is necessary to maximize AOQ to find

é N-I
AOQL = pL = max ê p
( ) ùú
êë N úû

which will occur when I , the average number of units inspected in a lot, is at a minimum.
Substituting the formula for ATI in place of I

é é æ C
e - np ( np ) ö ù ùú
i

å
p ên + ( N - n ) ç 1 -
AOQL = max ê p - ÷ú
ê N ê ç i! ÷ú ú
ë ë è i =0 ø û û
386 Acceptance Sampling in Quality Control

which can be shown to be

é N-n æ C
e - np ( np ) ö ù
i
æ
AOQL = max ê p ç
öç
÷
ê è N øç
è
å i!
÷ .ú
÷ú
ø û
ë i =0

and it is obvious that all the probabilities in this calculation involve the Poisson approxi-
mation to the binomial (Type B). Differentiating and setting the results equal to 0 gives

æ N - n öæ ö
C
e-x xi

AOQL = x ç ÷ç
è Nn ø çè åi =0
i!
÷
÷
ø

where x = npM, and pM represents the value of p at which the AOQL occurs.
Then
æ1 1 ö
AOQL = y ç - ÷
è n N ø

where
C
e-x xi

y=x å
i =0
i!

which is shown by Dodge and Romig to equal
e - x x c +2
y=
c!
and finally,
yN
n=
NpL + y

It follows that

æ nö
npL = y ç 1 - ÷
è N ø

Values of x and y are given in Appendix Table T14.1. Use of these values with the accep-
tance number, which gives minimum I (Figure 14.11), forms the basis of the Dodge–Romig
AOQL plans.

Constructing LTPD Plan with Minimum ATI

To find an LTPD plan that will achieve minimum ATI Imin, proceed as follows:

1. Given lot size (N), LTPD, and process average proportion defective (p). Define
pt = LTPD/100 = tolerance fraction defective.
2. Enter Figure 14.8, with the ratio p/pt on the x-axis and the product ptN on the y-axis,
to find the acceptance region that will make ATI a minimum. Use this acceptance
number c.
Series of Lots 387

1000
800
600 31
500 26 TO
400 21 TO 35
TO 30
300 19
25
17 20
15 18
200 13
16
11 14
12
10
[pt N = (tolerance fraction defective)/(lot size)]

r
100

be
8

um
80

n
7
Tolerance number of defectives

ce
tan
60 6
50 p 5
ce
Ac

40 4

30 3

2
20
1

10 0
8

6
5
4

3
Single sampling
consumer’s risk, 0.10
2

1
0 0.1 0.2 0.3 0.4 0.5 0.6
Ratio of process average to tolerance
[p/p1 = (Process average fraction defective)/(tolerance fraction defective)]

FIGURE 14.8
Dodge–Romig curves for finding the acceptance number. (Reprinted from Dodge, H.F. and Romig, H.G., Sampling
Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 14. With permission.)

3. Enter Figure 14.9 with the product ptN on the x-axis and read the product (ptn) on
the y-axis from the appropriate curve for c.
4. Divide (ptn) by pt to obtain n.
5. The plan n, c will give the LTPD protection desired on each lot with minimum ATI.
6. The minimum ATI can be found from Figure 14.10 by finding the point corre-
sponding to p/pt on the x-axis and ptN on the y-axis and interpolating between the
closest curves to obtain pt(ATI) = ptImin on the right axis. Division of pt(ATI) by pt
gives the ATI that will minimize ATI.
388 Acceptance Sampling in Quality Control

24

22
[ptn = (tolerance fraction defectives)/(sample size)]
Single sampling
consumer’s risk, 0.10 19
20
18
17
18
16
Tolerance times sample size

15
16
14

ber
13

um
14
12

en
anc
11
12

ept
10
10 Acc 9
8
8 7
6
5
6
4
3
4
2
2 0 1

0
1 2 3 4 5 7 10 20 30 50 70 100 200 300 500 1000
Tolerance number of defectives
[pt N = (tolerance fraction defectives)/(lot size)]

FIGURE 14.9
Dodge–Romig curves for finding the size of the sample. (Reprinted from Dodge, H.F. and Romig, H.G.,
Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959, p. 15. With
permission.)

For example, suppose N = 250, p = .016, and LTPD = 4%; so pt = .04. Then p/pt = 0.4 and
ptN = 10 so that Figure 14.8 shows c = 2. Entering Figure 14.9 with ptN = 10 and reading
from c = 2 gives ptn = 4.5 so that n = 4.5/.04 = 112.5. The plan is n = 113, c = 2. Entering
Figure 14.10 the point (0.4, 10) is nearest to the curve for pt(ATI) = 5.7 (interpolating) and so
ATImin = 5.7/.04 = 142.5 ~ 143.

Constructing AOQL Plan with Minimum ATI

To find an AOQL plan that minimizes ATI of accepted samples plus 100% inspection of
rejected lots, proceed as follows:

1. Given lot size (N), AOQL (pL), and process average proportion defective (p).
2. Calculate the ratio k = p/pL and the product M = pN .
3. Enter Figure 14.11 with k on the x-axis and M on the y-axis to find the acceptance
number region that will make ATI a minimum. Use the acceptance number c.
4. Use Appendix Table T14.1 to find the values of x and y specified for the c value
obtained in step 3.
5. Compute
yN
n=
pL N + y
Series of Lots 389

1000
800
600

[pt IMIN = (tolerance fraction defective)/(minimum average number inspected per lot )]
500
400
300

200 40
35
[ pt N = (tolerance fraction defective)/(lot size)]

Tolerance times minimum average inspected

100 30
80
Tolerance number of defectives

60 25
50
20
40
30
15
20 13
11
10
9
8
10 7
8 6

6 5
5 4
4 3
3
2
2
1.5
Single sampling
consumer’s risk, 0.10
1
0 0.1 0.2 0.3 0.4 0.5 0.6
Ratio of process average to tolerance
[p/pt = (process average fraction defective)/(tolerance fraction defective)]

FIGURE 14.10
Dodge–Romig curves for finding the minimum amount of inspection per lot. (Reprinted from Dodge, H.F. and
Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York, 1959,
p. 16. With permission.)

6. The plan n, c will give the AOQL specified with minimum ATI. The AOQL will
occur at

npM = x

so that

x
pM =
n
390 Acceptance Sampling in Quality Control

10,000
8,000
6,000
5,000
4,000
3,000

2,000

c
21
1,000
M = pN (process average fraction defective times lot size)

TO
800 19 25
15 20
600 1617
500 14 18
400 12
300 10
9
8
200
7
6
100 5
80
4
60
50
40
3
30

20
2

10
8
6
5
1
4
3

c=0
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
p
k= (ratio of process average fraction defective to AOQL)
pL

FIGURE 14.11
Dodge–Romig curves for determining the acceptance number, c; AOQL protection. (Reprinted from Dodge, H.F.
and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed., John Wiley & Sons, New York,
1959, p. 40. With permission.)

For example, suppose we take AOQL = 4%; so pL = .04, lot size N = 250, and process aver-
age p = .016. Then k = .016/.04 = 0.4 and M = .016(250) = 4. Hence, from Figure 14.11, c = 1.
Appendix Table T14.1 shows y = 0.84 and x = 1.62, and

.84 ( 250 )
n= = 19.4 ~ 20
.04 ( 250 ) + .84
Series of Lots 391

The plan is n = 20, c = 1, and the AOQL will occur at

1.62
pM = = .08
20

at p = .016, the ATI will be

ATI = nPa + N ( 1 - Pa )
= 20 ( .959 ) + 250 ( .041)
= 29.4

Anscombe Rectifying Inspection Procedure

F. J. Anscombe (1949) has presented an adaptable, easy-to-use, inspection procedure that
is appropriate in guaranteeing an LTPD and AOQL in a sequence of inspections. The
method does not rely on 100% inspection of rejected lots. Rather, successive samples are
taken on each lot until a stopping rule is satisfied and the lot accepted or until the lot
is exhausted. The stopping rule is set in such a way that poor lots will, in general, be
extensively sampled while good lots will require minimal sampling. It assumes that any
defective items found will be replaced with effective ones. The LTPD is guaranteed with
minimum ATI.

Operation
The procedure, in the words (notation slightly modified*) of Anscombe (1949, p. 193), is as
follows:

From a batch of N articles, a first sample of f1N articles is inspected, and then further
samples of f2N articles each. Defective articles found are removed or replaced by good
ones. Inspection ceases after the first sample if no defectives have been found, or after
the second sample if altogether one defective has found, or, generally, after the (r + 1)th
sample if altogether r defectives have been found. Inspection is continued until either
this stopping rule operates or the whole batch is inspected.

This ingenious procedure may be summarized as follows:

Sample (i) Sample Size Acceptance Number (ci = i − 1)

1 f1N 0
2 f2N 1
3 f3N 2
⋮ ⋮ ⋮
k fkN k−1

* Anscombe’s original α and β are given here as f1 and f2 after the manner of Duncan (1974).
392 Acceptance Sampling in Quality Control

Selection
The parameters of these plans are f1 and f2. These quantities and associated measures have
been tabulated exactly by Anscombe (1949) and are exemplified in Table 14.7, which is a
part of the original Table IV of the Anscombe paper. Anscombe’s notation compares to that
of this book as follows:

Anscombe Present Notation

Zt ptN
ε β
α f1
β f2
A ASN
N N
Y pN
AOQL (N)AOQL
Y* (last column) (N)pM

In the notation of this book, the heading of Anscombe’s first table would appear as follows:

Scheme ASN/N for Np Equal to

f1 f2 0 1 2 3 4 5 6 8 10 12 (N)AOQL (N)pM
Npt = 5, β = .10

Tables are given for β = .10 and β = .01 by Anscombe; however, only the table for β = .10
(i.e., LTPD protection) is shown here.
Anscombe’s tables are indexed by limiting quality Zt = Npt and associated risk ε = β.
They show the ratio of average sample number to lot size, that is, ASN/N for various
­values of Y = pN. The value of ASN/N that appears in bold type (denoted by asterisk here)
is the minimum ratio for the column. That is, minimum ATI since for these plans

ASN = ATI

Also shown are values of N(AOQL) and NpM for each plan. The original tables are in
terms of number defective in the lot, and not proportion defective. Accordingly, the values
shown must be suitably transformed by multiplying or dividing by the lot size to obtain
the conventional values.
To find a plan having a desired AOQL, compute N(AOQL) and search the second last
column of the Anscombe tables for the desired value. This will not guarantee minimum
ATI for the AOQL given.
To use Table 14.7 to guarantee a specified LTPD = pt for estimated fraction defective p,
proceed as follows:

1. Enter the table with Zt = Npt and ε = β = .10.

2. Find the column corresponding to Y = Np.
3. The value of ASN/N in bold type (denoted by asterisk here) indicates the row for
minimum ATI. Use the values of α = f1 and β = f2 obtained from the row.
TABLE 14.7
Series of Lots

Anscombe Rectifying Inspection Schemes for Lot Tolerance Zt = ptN with Risk e = ß = .10
Scheme Average Sample Size (A/N) for Y Equal to
α β 0 1 2 3 4 5 6 8 10 12 AOQL Ya

Zt = 5, ε = 0.10
0.3690 0.1900 0.369a 0.439a 0.536 0.674 1.4 3
0.4238 0.0982 0.424 0.465 0.515a 0.576a 0.650 0.743 1.8 5
0.4773 0.0639 0.477 0.508 0.542 0.581 0.626 a 0.677 0.737 0.890 1.9 6
0.5241 0.0459 0.524 0.548 0.574 0.603 0.635 0.670a 0.709 0.800 0.916 2.0 7
0.5642 0.0348 0.564 0.584 0.605 0.627 0.652 0.678 0.706a 0.769 0.844 0.935 2.1 8
0.5986 0.0275 0.599 0.615 0.632 0.651 0.670 0.691 0.713 0.760a 0.815 0.877 2.2 9
0.6283 0.0223 0.628 0.642 0.657 0.672 0.688 0.705 0.723 0.761 0.802 0.849 2.2 10
0.6540 0.0185 0.654 0.666 0.679 0.692 0.705 0.719 0.734 0.765 0.799 0.835 2.2 11
0.6767 0.0155 0.677 0.687 0.698 0.709 0.721 0.733 0.745 0.771 0.798a 0.828a 2.2 12
Zt = 10, ε = 0.10
0.2057 0.1694 0.206 a 0.241 a 0.287 0.352 0.444 0.569 0.677 0.820 0.897 0.940 2.8 5
0.2337 0.0967 0.234 0.256 0.283a 0.316a 0.356a 0.406 0.469 3.7 7
0.2669 0.0686 0.267 0.285 0.306 0.330 0.357 0.389a 0.427a 0.526 0.673 4.3 9
0.3002 0.0530 0.300 0.316 0.334 0.353 0.375 0.400 0.427 a 0.495 0.584 0.708 4.6 10
0.3323 0.0429 0.332 0.347 0.362 0.379 0.397 0.418 0.440 0.492a 0.556 0.639 4.9 11
0.3625 0.0357 0.363 0.375 0.389 0.404 0.420 0.438 0.456 0.499 0.549a 0.610 5.1 13
0.3908 0.0304 0.391 0.403 0.415 0.429 0.443 0.458 0.474 0.510 0.552 0.601 5.2 14
0.4170 0.0262 0.417 0.428 0.439 0.452 0.464 0.478 0.492 0.523 0.559 0.599a 5.3 15
0.4414 0.0229 0.441 0.452 0.462 0.473 0.485 0.497 0.510 0.537 0.568 0.602 5.4 16
Source: Reprinted from Anscombe, F.J., J. Roy. Stat. Soc. (Ser. A), 112(Pt. II), 202, 1949. With permission.
a Indicates minimum ratio for column.
393
394 Acceptance Sampling in Quality Control

4. Multiply ASN/N by N to obtain ASN. Divide the value of AOQL shown by N to

put it in terms of fraction defective. Similarly, divide the corresponding value of Y
by N to obtain pM.

For example, if a plan is desired having 4% LTPD with lot size 250 for process average
percent defective 1.6%, we have the following:

1. Zt = 250(.04) =10 and ε = β = 0.10.

2. Y = 250(.016) = 4.
3. Bold (asterisk) ASN/N = .356; so f1 = .2337 and f2 = .0967.
4. Measures are as follows:

ASN = .356 ( 250 ) = 89

AOQL = 3.7 = 250 = .015

pM = 7/250 = .028

Sample sizes will be

f1 ( N ) = .2337 ( 250 ) = 58.4 ~ 59

f 2 ( N ) = .0967 ( 250 ) = 24.2 ~ 25

Application of this plan would involve a first sample of 59 followed by successive

samples of 25. Inspection would be terminated if at any time the accumulated
number of defectives is less than the number of samples minus 1 or when lot is
exhausted. The plan would appear as follows:

Sample Sample Size Cumulative Sample Size Cumulative Acceptance Number

1 59 59 0
2 25 84 1
3 25 109 2
⋮ ⋮ ⋮ ⋮
k 25 59 + 25(k − 1) k−1

Measures
The following approximate measures of scheme performance have been given by
Anscombe. They may be used whenever
1 - f1
p<
f2N

1. Average sample number

é f1 f f 2 pN ù
ASN = N ê - 1 2 ú
êë 1 - pNf 2 ( 1 - pNf 2 ) úû
3

Series of Lots 395

2. AOQ at p

é f1 f1 f 2 ù
AOQ = p ê1 - + ú
êë 1 - pNf 2 ( 1 - pNf 2 ) úû
3

3. AOQL

é
( ) ù
2
1 1 - f1 1
AOQL = ê + - 1ú
Nê f2 f1 ú
êë úû

which is attained at

1 é 1 - f1 3 ù
pM = ê + - 1ú
N ëê f 2 2 f1 úû

For example, for the plan derived earlier where LTPD = 4%, N = 250, p = .16, we
obtained f1 = .2337, f2 = .0967. Hence at p = .016,

é .2337 ( .0967 ) ( .016 )( 250 ) ù

2
.2337
ASN = 250 ê - ú
ê 1 - .016 ( 250 )( .0967 ) ( 1 - ( .0016 )( 250 )( .0967 ) )3 ú
ë û
= 85.8
é .2337 ( .0967 ) ù
.2337
AOQ = .016 ê1 - + ú
ê 1 - .016 ( 250 )( .0967 ) ( 1 - ( .016 )( 250 )( .0967 ) )3 ú
ë û
= .011
é
( ) ù
2
1 ê 1 - .2337 1
AOQL = + - 1ú
250 ê .0967 .2337 ú
êë úû
= .015
1 é 1 - .2337 3 ù
pM = ê + - 1ú
250 ë .0967 2 .2337 û
= .030

Credit-Based Schemes
Although most rectification plans are based on the asymptotic behavior of application
repeated in the marketplace, credit-based plans reflect the current state of the marketplace
while guaranteeing an AOQL. Credit-based schemes depend on the credit principle as put
396 Acceptance Sampling in Quality Control

forth by Klaassen (2001). Credit is defined as the total number of items accepted since the
last rejection. These plans provide a useful alternative in conjunction with accept zero, that
is, c = 0, applications. The basic plan proceeds as follows:

1. Specify AOQL = a and set credit, k, to 0.

2. Choose sample size from

N
n=
( k + N ) a + 1

where
N is the lot size
k is the accumulated credit
a is the AOQL desired

3. Apply plan.
4. Update the credit on the basis of the sampling results:
a. PASS: Increase credit by adding the lot size to k.
b. FAIL: Set the credit to k = 0.
If k is already 0, screen both the lot and the sample.
If k already exceeds 0, exercise option to scrap, screen, or return lot to supplier.
c. Return to (2).

A check sequence for obtaining a credit-based accept zero scheme is given in Figure 14.12.
Occasionally, it may be desirable to put an upper limit or cap, k0, on the credit, thus
freezing the sample size. This is accomplished by using as a modified credit update in the
formula. The sample size formula then becomes

N
n=
{min ( k , k ) + N} a + 1
0

As an example of the application of the credit-based scheme, suppose three lots of size 100
have already passed and an AOQL of 1.0 is desired. Then

N = 100

k = 300

a = 0.01

hence

100 100
n= = = 20
( 300 + 100 ) 0.01 + 1 5
Series of Lots 397

Start

Specify a = AOQL

Set credit k = 0

Determine N = lot size

Determine sample size

N
n=
( k + N)a + 1

Apply accept zero plan

n, c = 0

Pass Fail

k=k+N k=0 k>0

Screen lot Optionally:

and sample Screen lot
Scrap lot
Return lot
to supplier

FIGURE 14.12
Check sequence for obtaining a credit-based accept zero plan.

and the plan is n = 20, c = 0. Now if the lot passes, credit will be increased to 400 and the
procedure starts over. If, however, the lot fails, its disposition is optional between scrap-
ping, screening, or returning to the supplier; however, credit will be reset to 0. Note that if
credit was at 0, the lot, including the sample, would have been screened and the acceptable
items sent to the marketplace.
Credit-based plans reflect the nature of the market in which they are applied. They
provide an imaginative accept zero approach for a series of lots. As such, sample sizes
are automatically adjusted to accommodate for what is in the field, while preserving the
AOQL. They are an excellent alternative for situations in which the product is of very good
quality as confirmed by relatively small sample sizes.
398 Acceptance Sampling in Quality Control

Dodge–Romig Plan Derivatives

The Dodge and Romig sampling plans were first published in 1944 and a second edition
appeared in 1959. These plans have not undergone any major changes since.

ASTM International Standard E1994

E1994 Standard Practice for Use of Process Oriented AOQL and LTPD Sampling Plans is a prac-
tice that is primarily a statement of principles for the guidance of ASTM technical commit-
tees and others in the use of AOQL and LTPD sampling plans for determining acceptable
of lots of product. This standard is an abbreviated compilation of the sampling plans pre-
sented by Dodge and Romig in their classic development of sampling plans for use with
the process associated with a continuing supply of products. The so-called AOQL plans
provide a means for disposition of product whether or not the process is in control as well
as incentives for process improvement in terms of reduced sample size as the process aver-
age percent defective is lowered. In addition, so-called LTPD plans are provided for use
with individual lots of product, not necessarily associated with a process stream.

Software Applications
Excel
Trindade and Meade (1999) proposed a general Excel solution to generate LTPD-type sam-
pling plans based on specific values of the process defective level p. Their solution is an
Excel add-in called “LTPD.xla” that can generate the values of n and c for any desired value
of p. They consider the case of a random sample from a process or a large lot for which
the sample is less than 10% of the lot size N, that is, Type B sampling. As such, they rely
on the binomial probability distribution. According to Trindade and Meade, many product
qualification plans are based on LTPD considerations as a means of assuring the customer of
protection against individual lots of poor quality getting through. A common problem is the
need to adjust n and c based on time, money, or resource considerations while simultane-
ously keeping β constant at a specified quality level. While there are tables available to assist
in selecting proper sample sizes and acceptance numbers, the user is restricted to only the
values specified in the tables. Their Excel add-in handles any specified LTPD at any prob-
ability of acceptance. We can illustrate this add-in with an example.
Suppose we have a case where a given sampling plan calls for a lot to be accepted based
on c ≤ 3 failures out of n = 300 sampled units. Due to the cost of each sampled unit, there is
a need to construct a sampling plan that allows a lot to be rejected on one failure, that is, a
c = 0 plan. One important requirement of the new plan is that the LTPD and the probabil-
ity of acceptance must be the same as the original plan. In this example, the probability of
acceptance is assumed equal to 10% at the LTPD.
We begin by setting up a simple spreadsheet as shown in Figure 14.13. Specify the values
of c, n, and the probability of acceptance (Pa) in cells B4, B5, and B6, respectively. Note that
the Excel inverse beta distribution function (BETAINV) is used to calculate the LTPD. The
general format of the BETAINV function is

BETAINV ( 1 – Pa , c + 1, n – c )
Series of Lots 399

FIGURE 14.13
Simple spreadsheet to demonstrate the Excel template LTPD.xla for an LTPD sampling plan.

In order to determine an alternative plan, we need to create a second table with two
columns as shown in Figure 14.14. The first column contains a list of possible sample sizes
in descending order if a lower value of c is desired (as in this case) or in ascending order if
a higher value of c is needed. The second column contains the Excel cumulative binomial
distribution (BINOM.DIST) function for each sample size in the first column. Note that the
format of the BINOM.DIST function is

BINOM.DIST(c, n, LTPD, 1)

where the last parameter of 1 tells Excel that we want the cumulative distribution func-
tion. Since we want a c = 0 plan, the new sampling plan occurs where the value of Pa most
closely approximates 10%, which is the LTPD. In Figure 14.14, we see that the alternative
plan of n = 103 and c = 0 has a Pa of 10%.
Using the Excel add-in LTPD.xla (found on the CRC book website), we execute the macro
and select Find alternative sampling plan from the dialog box that appears (Figure 14.15) and
click OK. Complete the next dialog box as shown in Figure 14.16. In this dialog box, enter the
c, n, and Pa (in %) for the original plan, and our desired c = 0 for the alternative plan. Then,
tell the macro that you want the output to start in cell $M$1 so it is next to the alternative
plan calculations given in Figure 14.14. When you click OK, the computed alternative plan
is shown in Figure 14.17. This analysis also shows the approximate AQL that would accept
95% of the lots passing this alternative sampling plan. The Excel add-in confirms that the
proper alternative plan is n = 103 and c = 0. As shown in Figure 14.15, the Excel add-in can
also find the LTPD for a given sampling plan or can find a sample size n for a sampling plan
for a given value of c.
Stephens (2016) created a pair of Excel templates to create LTPD and AOQL plans for sit-
uations of a desired process average p and lot size N and for a specified value of the LTPD
or AOQL. Stephens goes much further than Trindade and Meade in creating a means of
deriving nonstandard Dodge-Romig LTPD and AOQL plans. Like Dodge and Romig,
Stephens derives plans that achieve a minimum ATI. His paper discusses the mathemati-
cal basis for the derivation of such plans, which primarily relies on the use of the Poisson
distribution function. In this paper, Stephens compares the plan generated by his Excel
template to an approximate Dodge–Romig tabulated plan that requires availability of
these tables. Neubauer has modified these templates to include the Dodge–Romig tables
 400

FIGURE 14.14
Simple spreadsheet expanded for alternative c = 0 sampling plans.
Acceptance Sampling in Quality Control
Series of Lots 401

FIGURE 14.15
Dialog box for the Excel template LTPD.xla to select an alternative sampling plan.

FIGURE 14.16
Dialog box for the Excel template LTPD.xla to indicate c = 0 for an alternative sampling plan with an original
sampling plan of n = 300 and c = 3.

FIGURE 14.17
Output for the Excel template LTPD.xla to indicate c = 0 for an alternative sampling plan with a probability of
acceptance of 10%.
402 Acceptance Sampling in Quality Control

within each template (they are hidden) so that once the desired values of N, LTPD (or
AOQL), and p are entered, the template plan and the corresponding Dodge–Romig plan
appear to the user. These modified templates are available through the CRC book website
and are named D & R LTPD Plans (Stephens 2016).xlsm and D & R AOQL Plans (Stephens
2016).xlsm. Stephens also includes the ability for the user to enter an alternative plan
either directly through manual entry or via spin buttons to change the values of n and c.
According to Stephens, below each sampling plan in the LTPD template’s first tab is 14
numerical properties in column M from rows 22 to 48. Five probabilities of acceptance, Pa,
values are calculated—two for the LTPD (pt) value using the binomial and hypergeometric
distributions and one each for p, pl as defined, and p1 as shown and defined in row 36. In
addition, four ATI values are calculated—one for each of the four p values. Finally, the
remaining three properties are p1 (as used in cells M38 and M46), AOQL computed using
two formulas, and LTPD, which agrees closely with the input in cell M12.
Stephens also includes a second tab entitled Plan Properties, which contains the OC, AOQ,
and ATI curves for the template-chosen plan. Neubauer has also included the curves for
the alternative plan and the corresponding Dodge–Romig plan on this tab. Stephens notes
that for the LTPD template and associated derived plans, the hypergeometric probability
function is used to evaluate the OC, AOQ, and ATI curves.
As an example, assume that the lot size N = 2000, desired LTPD = 0.03 (or 3%), and
p  =  0.005 (or 0.5%), under the assumption that there is a 10% probability of acceptance
at the LTPD (Dodge–Romig default value). Note that the 0.10 value of this probability
of acceptance can be changed in cell H16 if a nonstandard Dodge–Romig LTPD plan is
of interest. In Figure 14.18, we see that for the entries for N, LTPD, and p, the template-
generated plan is n = 223 and c = 3. If we look in column W (not shown in Figure 14.18), the
corresponding Dodge–Romig plan is n = 175 and c = 2, which has an AOQL of 0.72% that
agrees well with the template plan’s AOQL of 0.77%. The Dodge–Romig plan is found for
lot sizes in the range of 1001–2000. Since our lot size of 2000 is at the extreme of this range,
it is worth noting that the plan for the next consecutive lot size range of 2001–3000 is
n = 220 and c = 3. Figure 14.18 shows that this alternative sampling plan has AOQL and
LTPD values close to the template-generated plan.
But suppose that we really wish to find a plan with the desired design parameters with a
smaller sample size and lower acceptance number. Stephens shows that the sampling plan
n = 170 and c = 2 is an attractive choice and can be found using the spin buttons to tweak
the n and c values of the alternative plan.
The other Excel template file determines optimal AOQL plans in a manner similar to
the LTPD template. Using a new example, assume that N = 6000, desired AOQL = 0.03
(or 3%), p = 0.018 (or 1.8%), under the assumption that there is a 10% probability of accep-
tance at the LTPD (Dodge–Romig default value). The minimum ATI AOQL sampling plan
generated by the template is n = 124 and c = 6. No hypergeometric ATI column is used
for the AOQL sampling plan derivation. Similar to the other template, several properties
for the sampling plan are shown in column L from rows 22 to 48 with two of these proper-
ties (the computed values of the AOQL and LTPD) with duplicated computation using dif-
ferent formulas. Five probabilities of acceptance are calculated—one each for p1, pl, and pt
and two for p using the binomial and hypergeometric. Also, there are three ATI values
calculated—one for each of the three p values p1, pl, and p. Finally, the remaining five prop-
erties are p1, AOQL, np1, d/dp, and LTPD. The user can use the spin buttons to find an alter-
nate sampling plan with similar performance statistics. For example, an alternate sampling
plan of n = 103 and c = 5 is similar to the template-generated plan for a smaller value of c as
shown in Figure 14.19. The corresponding Dodge–Romig AOQL plan is n = 105 and c = 5,
 Series of Lots

FIGURE 14.18
Excel template D & R LTPD Plans (Stephens 2016).xlsm used to generate a nonstandard Dodge–Romig LTPD plan in comparison with the nearest standard plan.
403
 404

FIGURE 14.19
Excel template D & R AOQL Plans (Stephens 2016).xlsm used to generate a nonstandard Dodge–Romig AOQL plan in comparison with the nearest standard plan.
Acceptance Sampling in Quality Control
Series of Lots 405

and these values are seen in cells W20 and W22, respectively, with an LTPD of 8.8%. This
value of the LTPD corresponds roughly with the template plan of 8.5% and the alternate
plan of 9.0%.
The Stephens (2016) article contains more information than covered here and the origi-
nal Excel templates contain two other tabs, which are discussed in that paper. The reader
is encouraged to refer to the paper for more information on the design and use of the
templates. Also, Neubauer has included information on the source of the paper at the
top of the first tab of each modified template that includes a hyperlink for the author’s
name. When clicked, the hyperlink will direct the user to the paper’s website. Quality
Engineering subscribers can download the paper and templates free of charge. Others will
be charged.

Snap Sampling Plans!

Snap Sampling Plans! can generate Dodge–Romig AOQL and LTPD sampling plans.
Click the Dodge/Romig button on the main menu screen. As an example of an LTPD
plan, suppose that the lot size N = 2000, desired LTPD = 0.04 (or 4%), and p = 0.005
(or 0.5%), under the assumption that there is a 10% probability of acceptance at the
LTPD (Dodge–Romig default value). Complete the dialog box as shown in Figure
14.20. Click on the Sample Plan button to generate the plan. The resulting plan is n =
165, c = 3 and is presented in Figure 14.21. This plan agrees with what can be found
in Table 14.3.
As an example of an AOQL plan, suppose that the lot size N = 2000, desired AOQL = 0.04
(or 4%), and p = 0.005 (or 0.5%), under the assumption that there is a 10% probability of
acceptance at the LTPD (Dodge–Romig default value). Complete the dialog box as shown
in Figure 14.22. Click on the Sample Plan button to generate the plan. The resulting plan
is n = 21, c = 1 and is presented in Figure 14.23. This plan agrees with what can be found in
Table 14.1.

FIGURE 14.20
Snap Sampling Plans! used to generate a standard Dodge–Romig LTPD plan.
406 Acceptance Sampling in Quality Control

Dodge–Romig
sampling plan

Lot Size Type Avg % defective AOQL%

2000 LTP D 0.5 3.000
Results Sample sizea Accept Reject
175 2 3
Accept number is the number of allowed defects.
If the inspection sample has defects ≥ the reject number, then
reject the lot.
a If sample size is greater than the lot size, then conduct 100%
inspection.

FIGURE 14.21
Snap Sampling Plans! used to generate a standard Dodge–Romig LTPD plan.

FIGURE 14.22
Snap Sampling Plans! used to generate a standard Dodge–Romig AOQL plan.

Dodge–Romig
sampling plan

Lot Size Type Avg % defective AOQL%

2000 AOQL 0.5 4.000
Results Sample size Accept Reject
21 1 2
Accept number is the number of allowed defects.
If the inspection sample has defects ≥ the reject number, then
reject the lot.
a If sample size is greater than the lot size, then conduct 100%
inspection.

FIGURE 14.23
Snap Sampling Plans! used to generate a standard Dodge–Romig AOQL plan.
Series of Lots 407

Problems
1. Using the Altman diagram, find the AOQLs associated with the following plans:
a. n = 50, c = 1
b. n = 80, c = 1
c. n = 20, c = 0
2. From the Altman diagram, derive a plan for AOQL = 5% when sample size must be
restricted to 30 or less.
3. Find Dodge–Romig single- and double-sampling plans for AOQL = 4.0% for the lot
sizes and process average percent defective shown.
a. N = 125, p = 1%
b. N = 1250, p = 1%
c. N = 5500, p unknown
4. Find Dodge–Romig LTPD plans for LTPD = 4% meeting the specifications given in
Problem 3.
5. Construct an LTPD plan for N = 1250, pt = 4% when the process average is at
1% defective. What is the minimum ATI for this plan?
6. Construct an AOQL plan for N = 1250, pL = 4% when the process average is at
1% defective. At what fraction defective will the AOQL occur?
7. A lot of size 125 is to be screened using the Anscombe procedure. LTPD protection
of pt = .04 is desired, while the process average p = .008. Construct the appropriate
Anscombe scheme. What are its average sample number, the AOQL, and the point at
which the AOQL occurs?
8. What Dodge–Romig single-sampling plan corresponds to the Anscombe plan devel-
oped in Problem 8? Find the minimum amount of inspection per lot from Figure 14.10
and compare it to the ASN of the Anscombe plan.
9. Using f1 = .300 and f2 = .053, verify the values given in the table for Zt = 10, ε = .10
when Np = 0.
10. Suppose four lots of size 50 have already passed and an AOQL of 2.0 is desired.
Determine the credit-based scheme that should be used.

References
Altman, I. B., 1954, Relationship between sample size and AOQL for attributes single sampling plans,
Industrial Quality Control, 10(4): 29–30.
American Society for Testing and Materials, Standard Practice for Use of Process Oriented AOQL and
LTPD Sampling Plans, ASTM Standards E1994, Vol. 14.02, West Conshohocken, PA.
Anscombe, F. J., 1949, Tables of sequential inspection schemes to control fraction defective, Journal of
the Royal Statistical Society (Series A), 112(Part II): 180–206.
Dodge, H. F. and H. G. Romig, 1929, A method of sampling inspection, The Bell System Technical
Journal, 8(10): 613–631.
Dodge, H. F. and H. G. Romig, 1941, Single sampling and double sampling inspection tables, The Bell
System Technical Journal, 20(1): 1–61.
408 Acceptance Sampling in Quality Control

Dodge, H. F. and H. G. Romig, 1959, Sampling Inspection Tables, Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York.
Duncan, A. J., 1974, Quality Control and Industrial Statistics, 4th ed., Richard D. Irwin, Homewood, IL.
Klaassen, C. A. J., 2001, Credit in acceptance sampling on attributes, Technometrics, 43(2): 212–222.
Schilling, E. G., J. H. Sheesley, and K. J. Anselmo, 2002, Minimum average total inspection plans
indexed by average outgoing quality limit, Quality Engineering, 14(3): 435–451.
Stephens, K. S., 2016, Practitioner advice: Dodge and Romig sampling tables: Revisited, refined, and
extended-practitioner advice, Quality Engineering, 28(2): 238–244.
Trindade, D. C. and D. Meade, 1999, A General Excel Solution for LTPD Type Sampling Plans, Joint
Statistical Meetings, Baltimore, MD.
15
Continuous Sampling Plans

In the sampling of some processes, lots are not clearly defined. In a sense, lot size is N = 1
since units are produced item by item. Examples might be cars coming off an assembly
line, soft drink bottles from a continuous glass ribbon machine, or welded leads emanating
from a welding operation. Yet average outgoing quality limit (AOQL) and perhaps some
form of lot tolerance percent defective (LTPD) protection may be desired. Sometimes in
such situations, it is possible to artificially define a lot, such as the production of an hour,
a shift, a day, or a week. This is often quite arbitrary, ­however, and other alternatives exist.
Emanating from the original CSP-1, published by Dodge (1943), several d ­ ifferent
­continuous sampling plans have been developed to deal with this situation, ­usually with
AOQL protection. They apply to a steady stream of individual items from the ­process and
require sampling of a specified fraction, f, of the items in order of ­production, with 100%
inspection of the flow at specified times. Several such plans have been described in detail
by Stephens (1980) in a manual prepared for the American Society for Quality Control.
Special measures of performance apply to continuous plans, they include

• AOQ as the average outgoing quality

• AOQL as the average outgoing quality limit
• Pa as the average fraction of production accepted under sampling
• AFI as the average fraction of production inspected

The AOQ and AOQL are previously defined. The symbol Pa is used to denote the average
fraction of production accepted under sampling since in concept Pa implies the probability
of an item being accepted on a sampling basis (whether included in the sample or not). In
this sense, Pa will be seen to be analogous to the lot-by-lot probability of acceptance under
rectification. AFI indicates the average fraction of product actually inspected including
items inspected during sampling or in screening. Then
1 - AFI
Pa =
1- f

Dodge Continuous Plans

Dodge CSP-1
The most celebrated continuous sampling plan and the plan that undoubtedly has received
the most application is also the original—the Dodge CSP-1 plan. It is carried out on a stream
of product, with items inspected in order of production. The procedure is as follows:

1. Specify sampling fraction (f) and clearing interval (i).

2. Begin 100% inspection.

409
410 Acceptance Sampling in Quality Control

Start

Inspect i
Replacet successive
defective units
with
nondefective

Defective No defective
found found

Randomly sample
fraction f units

Defective No defective
found found

FIGURE 15.1 
Dodge CSP-1 procedure.

3. After i units in succession have been found without a defective, start sampling
inspection.
4. Randomly inspect a fraction f of the units.
5. When a defective is found, revert to 100% inspection (step 2).

A diagrammatic representation of CSP-1 will be found in Figure 15.1.

A detailed discussion of the Dodge CSP-1 plan and its relation to other approaches for
the sampling of a steady stream of product is given in Dodge (1947).
The Dodge CSP-1 plan and its later modifications CSP-2 and CSP-3 (see the fol-
lowing text) use a special measure to evaluate protection against spottiness, that is,
surges of highly defective product. Based on a finite production run of 1000 units pt%
is the ­percent defective in a consecutive run of 1000 units for which the probability of
­remaining under sampling is 10% for a sample of f. It shows the percent defective that
will result in a 90% chance of reverting to 100% inspection within a run of 1000 consecu-
tive units.
The notation here is somewhat deceptive, in that pt% is not the LTPD of the continuous
plan. Burr (1976) has shown that, if f is the fraction sampled in a CSP-1 plan,

1000 f
æ p %ö
.10 = ç 1 - t ÷
è 100 ø

which gives

é 1
ù
pt % = 100 êë1 - (.10) 1000 f úû

a function of f only.
Continuous Sampling Plans 411

If the LTPD is to be determined in terms of a 10% probability of acceptance Pa as defined

earlier for continuous plans, we find, for CSP-1, when Pa = .10

é æ f ö1/i ù
LTPD = 100 ê1 - ç ÷ ú
ê è 9+ f ø ú
ë û

which is a function of both f and i.

Both of these equations can be solved using logarithms. For example, for the CSP-1 plan
f = .1, i = 38, and pt% = 2.3% while LTPD = 11.2%.
A particular CSP-1 plan is determined by the values of f and i selected. The choice of f
and i, of course, depends on the value of AOQL and possibly the protection against spotty
quality desired for the plan. Dodge presented a simple diagram, shown in Figure 15.2, to
be used in setting these quantities. Values of f are given on the left axis while values of pt%
are given on the right. Values of i are shown on the abscissa. The curves represent various
levels of AOQL in percent so that any curve defines alternative f and i to obtain the AOQL.
The corresponding pt% can be read from the right axis.
For example, if an AOQL of 2.9% is desired with protection against spotty quality of 2.3%
we have, from the diagram, i = 38 and f = .10. This would result in 100% inspection until
38 units are found good in succession with random sampling of 10% of the units thereafter
until a defective is found.
The Dodge CSP-1 plan provides AOQL protection for most practical conditions; how-
ever, when the process is not in a state of control, instances may be found where the AOQL
may be exceeded. This was pointed out by Wald and Wolfowitz (1945). Subsequently,

Pt in % = the value of percent

f defective. In a consecutive run pf
in % of N = 1000 product units for in %
50
which the probability of acceptance. 0.3
Pa is 0.10 for a sample size of f %. 0.4
40 0.5
30 0.6
0.8
20 1.0

10 2.0
8
3.0
6
0.01 = AOQL 4.0
5 10 8 6 5 4 3 2 1 0.5 0.3 0.2 0.1 0.16 0.08 0.02
in %
4 5.0
6.0
3
8.0
2 10.0

1
10 20 30 40 60 80 100 200 300 400 600 1,000 2,000 4,000 6,000 10,000 20,000
Number of units (i)

FIGURE 15.2 
Dodge CSP-1 curves for f and i. (Reprinted from Dodge, H.F., Ind. Qual. Control, 4(3), 6, 1947. With permission.)
412 Acceptance Sampling in Quality Control

Lieberman (1953) showed that the Dodge CSP-1 procedure absolutely guarantees an
unlimited AOQL (UAOQL) of

UAOQL =
(1/f ) - 1
(1/f ) + i

when defective items are replaced with good items.

For the plan f = .1, i = 38, this becomes

UAOQL =
(1/.1) - 1 = .188
(1/.1) + 38

which is considerably worse than the nominal AOQL of 2.9%. Under this highly con-
servative approach, i would have to be 310 to guarantee a 2.9% UAOQL. Fortunately,
the nominal AOQL of 2.9% will hold in most practical situations and so can safely be
used to characterize the plan. When defective items are not replaced, the UAOQL guar-
anteed becomes

1- f
UAOQL =
f ( i - 1) + 1

as given by Banzhaf and Brugger (1970). This would give UAOQL = .191 for the plan f = .1,
i = 38 when defectives are not replaced.

Dodge–Torrey CSP-2 and CSP-3

H.F. Dodge and M.N. Torrey later improved upon CSP-1 somewhat, particularly with
regard to the occurrence of an occasional stray random defective (CSP-2) and, in addition,
with regard to “spotty” quality (CSP-3). That is, short bursts of bad quality. In the Dodge
and Torrey (1951) paper, they proposed to extend the CSP-1 procedure by changing the
steps given earlier for CSP-1 as follows:
For CSP-2,

Step 5. When a defective is found, continue sampling for k successive sample units. If
no defective is found in the k samples, continue sampling on a normal basis (step 4). If a
defective is found in the k samples, revert to 100% inspection immediately (step 2).

For CSP-3,

Step 5. Same as CSP-2, except, in addition, begin step 5 as follows: When a defective is
found, inspect the next 4 units, if an additional defective is found revert to 100% inspec-
tion (step 2); otherwise, continue sampling for k...

The other steps remain the same in each procedure. A flowchart for the CSP-1, 2, and 3
procedures is given in Figure 15.3, which highlights the differences between them.
Continuous Sampling Plans 413

Start
CSP 1, 2, 3
Inspect i
Replace
successive
defective
units
with
nondefective
Defective No defective
found found

Randomly sample
fraction f units

Defective No defective
found found

(CSP–1)
CSP–3 (CSP–3) (CSP–2)
(only)
Inspect
next 4
units

Defective No defective
found found

CSP–2, 3
(only) Continue sampling
for k samples

Defective No defective
found found

FIGURE 15.3 
Dodge CSP-1, 2, and 3 procedures.

Since CSP-3 is a very slight modification of CSP-2, the curves for finding f and i for
CSP-2 are used also for CSP-3 as a very good approximation. A set of curves is presented
in Dodge and Torrey (1951) for CSP-2 for the case when k = i. These curves are given in
Figure 15.4. They are employed in a manner identical to those for CSP-1. In fact, the CSP-1
curves are superimposed on the diagram as dotted lines. We find the CSP-2 plan f = .1,
i = 50 gives an AOQL of 2.9% as did the CSP-1 plan previously discussed. The value of pt%
for this plan is about 4.2% higher than the value of 2.3% from the CSP-1 plan because of the
increased difficulty of reverting to 100% inspection under CSP-2.
CSP-2 can also be shown to guarantee an AOQL even when the process is not in a state of
control. The upper limit on AOQL, as given by Banzhaf and Brugger (1970), when ­defective
items are not replaced, is

2 (1 - f )
UAOQL =
if + 2 ( 1 - f )

This formula may also be used as an upper limit on the AOQL of CSP-3.
 414

Continuous sampling plans pt in % = the value of percent defective in a consecutive run of N = 1000
f CSP-2 (k = i) product units for which the probability of acceptance, pa, is 0.10
in % for CSP-1, pt in a unique value, constant for all values of 1.
CSP-1 pt in %
i = 10 20 30 40 50 60 80 100 200 300 400 500 800 1000 2000
CSP-2 CSP-1
50
40
0.5
1
30

20 1
2

3
2
10 CSP-2 AOQL,
4
% = 10 8 6 5 4 3 2 1 0.5 0.3 0.2 0.1
8 5 3
6 6
4
5
8
5
4
10 6
3 7
8
15
10
2
CSP - 1 20
AOQL, % = 10 8 6 5 4 3 2 0.5 0.3 0.2 0.1

1
10 20 30 40 50 60 80 100 200 300 400 600 800 1000 2000 3000
Number of units (i)

FIGURE 15.4 
Dodge–Torrey CSP-2 curves for f and i. (Reprinted from Dodge, H.F. and Torrey, M.N., Ind. Qual. Control, 7(5), 8, 1951. With permission.)
Acceptance Sampling in Quality Control
Continuous Sampling Plans 415

100 CSP-1 plan

accepted on a sampling basis

Percent of total production
CSP-2 plan
80
(k = i)
60

40 AOQL = 3% AOQL = 6%
f = 0.05 f = 0.10
20 AOQL = 1%
f = 0.10
0
0 2 4 6 8 10 12 14 16 18 20
Incoming percent defective

FIGURE 15.5 
OC curves of three CSP-1 and CSP-2 plans. (Reprinted from Dodge, H.F. and Torrey, M.N., Ind. Qual. Control, 7(5),
9, 1951. With permission.)

The operating characteristic (OC) curves of continuous plans are expressed in terms of
the percent of total production accepted on a sampling basis (100Pa) plotted against incom-
ing values of percent defective. A set of such curves from the Dodge and Torrey (1951)
paper is given in Figure 15.5.

Measures of CSP-1, 2, and 3

Formulas are available to more precisely determine the various measures of CSP plans
for incoming proportion defective p, where q = 1 − p, u is the average number of units
inspected on a 100% inspection basis

1 - qi
u=
pqi
( CSP-1, CSP- 2 )

and v is the average number of units passed during sampling inspection

1
v= ( CSP-1)
fp
2 - qk
v= ( CSP- 2 )
(
fp 1 - qk )

The formulas are as follows:

Average fraction inspected

F, average fraction of total product inspected in the long run,

u + fv
F=
u+v
( CSP-1, CSP- 2 )

416 Acceptance Sampling in Quality Control

Specifically,

f
F= ( CSP-1)
f + q (1 - f )
i

F=
( )( ) ( )
f 1 - qi 1 - q k + qi f 2 - q k
( CSP-2 )
f (1 - q ) (1 - q ) + q ( 2 - q )
k i i k

f
F= ( CSP-2, k = i)
( ) + q (2 - q )
2
f 1- q i i i

f ( 1 - q ) ( 1 - q ) + fq + 4 fpq + fq ( 1 - q )
i k+4 i i i+ 4 k

F= ( CSP- 3 )
f ( 1 - q ) ( 1 - q ) + q + 4 fpq + q ( 1 - q )
i k+4 i i i+ 4 k

AOQ (defectives replaced by good)

pA = AOQ , average outgoing quality

AOQ = p [1 - F ] ( defectives replaced with good )

Specifically,

é ( 1 - f ) qi ù
AOQ = p ê iú ( CSP-1)
êë f + ( 1 - f ) q úû

AOQ = p
é
ê
( 1 - f ) qi 2 - q k ( ) ù
ú ( CSP- 2 )
( )(
ê f 1 - qi 1 - q k + qi 2 - q k
ë ) ( ) ú
û

é ( 1 - f ) qi 2 - qi
AOQ = p ê
( ) ù
ú ( CSP-2, k = i)
ê f + ( 1 - f ) qi 2 - qi
ë ( ) ú
û

é
AOQ = p ê
( 1 - f ) qi 1 + q 4 - q k + 4 ù
ú
( ) ( CSP- 3 )
ë
i
(
ê f 1- q 1- q k+4 i
)(4
+ q 1+ q - q k+4 i
)
+ 4 fpq ú
û ( )

AOQ (defectives removed but not replaced by good)

p¢A = AOQ¢, average outgoing quality

p (1 - F ) AOQ
AOQ¢ = =
(1 - pF ) q + AOQ
Continuous Sampling Plans 417

Specifically,

é ( 1 - f ) qi ù
AOQ¢ = p ê iú ( CSP-1)
êë fq + ( 1 - f ) q úû

AOQ¢ = p ê
(
é ( 1 - f ) qi 2 - qi ) ù
ú ( CSP-2, k = i)
ë (
ê fq + ( 1 - f ) q i 2 - q i ) ú
û

The AOQL may be found by the differentiation–iteration technique used by Dodge (1943)
or by trial and error. The relationship between pL, the AOQL, and the point at which it
occurs pM is

AOQL = pL =
( i + 1) p M - 1
i

so that

1 + ipL
pM =
i+1

These formulas can be used to calculate specific measures for a given CSP plan. As an
example, consider the CSP-1 plan f = .1, i = 38 evaluated at p = .054. We have

q = .946
1 - .946 38
u= = 134.148
(
(.054 ) .94638 )
1
v= = 185.185
( )(.054 )
. 1

Average fraction inspected

134.148 + .1 ( 185.185 )
F= = .478
134.148 + 185.185

or alternatively,

.1
F= = .478
.1 + .946 38 ( .9 )

AOQ (with replacement)

AOQ = .054 ( 1 - .478 ) = .028

418 Acceptance Sampling in Quality Control

or alternatively,

é ( .9 ) .946 38
AOQ = .054 ê
( ) ù
ú = .028

ê .1 + .9 .946 38
ë ( ) ú
û

AOQ (without replacement)

.054 ( 1 - .478 ) .028

AOQ¢ = = = .029
( 1 - . 054 ( . 478 ) ) . 946 + .028

or alternatively,

é
AOQ¢ = .054 ê
.9 .946 38 ( ) ù
ú = .029

ê .1 ( .946 ) + .9 .946 38
ë ( ) ú
û

The AOQL will occur at

1 + 38 ( .029 )
pM = = .054
38 + 1

giving, of course,

AOQL =
( 39 )(.054 ) - 1 = .029
38

Now consider the CSP-2 plan such that f = .1, i = 50, k = 50 evaluated at p = .054. We have

q = .946
1 - .946 50
u= = 278.682
(
.054 .946 50 )
2 - .946 50
v= = 382.676
(
.1 ( .054 ) 1 - .946 50 )

Average fraction inspected

278.682 + .1 ( 382.676 )
F= = .479
278.682 + 382.676

or alternatively,

F=
( )( )
.1 1 - .946 50 1 - .946 50 + .946 50 ( .1) 2 - .946 50 ( ) = .479
(
.1 1 - .946 50
) (1 - .946 ) + .946 ( 2 - .946 )
50 50 50

Continuous Sampling Plans 419

or when k = i,

.1
F= = .479
( ) ( )
2
50
.1 1 - .946 + .946 50 2 - .946 50

AOQ (with replacement)

AOQ = .054 ( 1 - .479 ) = .028

or alternatively,

é
AOQ = .054 ê
(
.9 .946 50 2 - .946 50 )( ) ù
ú = .028
( )(
ê .1 1 - .946 50 1 - .946 50 + .946 50 2 - .946 50
ë ) ( ) ú
û

or when k = i,

(
é .9 .946 50 2 - .946 50
AOQ = .054 ê
)( ) ù
ú = .028
ë (
ê .1 + .9 .946 50 2 - .946 50 )( ) ú
û

AOQ (without replacement)

.054 ( 1 - .479 ) .028

AOQ¢ = = = .029
1 - .054 ( .479 ) .946 + .028

or when k = i,

é
AOQ¢ = 054 ê
(
.9 .946 50 2 - .946 50 )( ) ù
ú = .029
ë (
ê .1 ( .946 ) + .9 .946 50 2 - .946 50 )( ) ú
û

The AOQL for this plan should also be .029. Hence, it should occur at pM = .054 as before
for the CSP-1 plan.
Finally, evaluating the corresponding CSP-3 plan f = .1, i = 50, k = 50, we have

F=
(.1(1 - .946 ) (1 - .946 ) + .1(.946 ) + 4 (.1)(.054 ) (.946 ) + .1(.946 ) (1 - .946 ))
50 54 50 50 54 50

(.1(1 - .946 ) (1 - .946 ) + .946 + 4 (.1) (.054 ) (.946 ) + .946 (1 - .946 ))

50 54 50 50 54 50

= .508

and with replacement of defectives

AOQ = .054 ( 1 - .508 ) = .027

420 Acceptance Sampling in Quality Control

or alternatively,

é
AOQ = .054 ê
( )(
.9 .946 50 1 + .946 4 - .946 54 ) ù
ú
ë ( )( ) ( ) (
ê .1 1 - .946 50 1 - .946 54 + .946 50 1 + .946 4 - .946 54 + 4 ( .1)( .054 ) .946 50 ) ú
û
= .027

The AOQL for this plan should also be .029. This too should occur at pM = .054.

Stopping Rules and Selection of CSP-1 Plans

Occasionally, it may become obvious that the process level of fraction defective has
moved upward from nominal levels. This is exhibited by excessively long sequences of
100% inspection. It is for such an eventuality that stopping rules, rules that indicate when
the process should be stopped for corrective action, were devised. Such rules have been
extensively investigated by Murphy (1959a). The rules studied by Murphy are summa-
rized in Table 15.1.
Typical of the stopping rules is the so-called rule r that involves stopping as soon as r
defective units are found in any screening sequence. We shall consider only this rule here.
To uniquely determine r for a given CSP-1 plan having
f as the fraction inspected
i as the clearing interval
A as AOQL = pL
pM as the proportion defective at AOQL

it is necessary to specify
E as the average number of units produced between successive stops wheen p = pM
recalling

1 + i ( AOQL )
pM =
i+1

Note that for large values of i (say greater than 100)

1
pM  + AOQL
i+1

TABLE 15.1
Stopping Rules for CSP-1 Plans
Rule (n* − i) Stop as soon as a defective unit is found in any one screening sequences after the sequence has
exceeded n* − i units.
Rule (r) Stop as soon as a specified number r of defective units are found in any one screening
sequences.
Rule (N, R) Stop as soon as a specified number R of defective units are found in any block of a specified
number N of inspected units. (Blocks do not overlap.)
Rule (n*) Stop as soon as a specified number n* of units have been inspected in any one screening
sequences without ending it.

Source: Reprinted from Murphy, R.B., Ind. Qual. Control, 16(5), 10, 1959a. With permission.
Continuous Sampling Plans 421

Algebraically manipulating the formulas given by Murphy, we find that to assure an inter-
val between stops of E at proportion defective pM, it is necessary to set

- log ( 1 - f + EFA )
r=
(
log ( F - f ) /F )

using any base for the logarithms employed.

Thus, for the plan f = .1, i = 38 with AOQL = .029, we have

f = .1

i = 38

A = .029

pM = .054

F = .478

so for an interval between stops of, say, 2i = 76 so that E = 76 and

- log ( 1 - .1 + 76 ( .478 )( .029 ) )

r=
log ( ( .478 - .1) /.478 )

- log 1.9535 -.2908

r= = = .285 ~ 3
log .7908 -.1019

As we have seen, the Dodge charts give a wide variety of choice of f and i for a given
AOQL. Murphy (1959b) has presented a way to uniquely define a CSP-1 plan given
A as the AOQL
P¢ as the producer’s nominal quality level
F¢ as the fraction inspected at producer’s nominal quality level

For a given AOQL this allows the producer to specify a quality level, which is to have
minimal inspection. P′ is chosen to be a fraction defective, which is to require a reasonably
small fraction inspected F′ when quality is at the specified level. The procedure given by
Murphy is as follows for plans where (P′ < A)

1. Calculate
A - P¢
B=
2.3 (1 - A ) (1 - P¢ )

and

H=
(1 - F¢ ) (1 - P¢ ) B
AF¢
422 Acceptance Sampling in Quality Control

Factor for determining i

2.5

2.0

1.5
C

1.0

0.5

0
1 2 3 5 7 10 20 30 50 70 100 200 300 500 700 1000
H

FIGURE 15.6 
C factor for determining i. (Reprinted from Murphy, R.B., Ind. Qual. Control, 16(6), 20, 1959b. With permission.)

2. Using the graph given by Murphy (Figure 15.6), find the value of C corresponding
to H.
3. Then i = C/B.
4. Find f from the Dodge chart for CSP-1 (Figure 15.2).

Murphy gives, as an example, the selection of a plan having A = .10, P′ = .05, and F′ = .10.
We have

.10 - .05
B= = .0235
2.3 ( 0.90 )( 0.95 )

and

H=
(.90 )(.95 )(.0235 ) = 2.01
(.10 )(.10 )
C = 0.34 ( from Figure 15.6 )
0.34
i= = 14.5 ~ 15
.0235
f = .043 ( from Figure 15.2 )

So the plan having the desired characteristics is

f = .043 i = 15
Continuous Sampling Plans 423

Note that for a plan selected in this way, it is possible to compute a value of r for the stop-
ping rule that will give an interval between stops of E′ when quality is at the producer’s
nominal quality level P′ by using the formula

r=
(
- log 1 + E¢P¢F¢ ( 1 - P¢ )
i
)
(
log 1 - ( 1 - P¢ )
i
)

Any base logarithms may be used. For the plan just selected in the example, if E′ = 10,000,
we have

r=
(
- log 1 + 10,000 ( .05 )( .10 ) ( 1 - .05 )
15
)
(
log 1 - ( 1 - .05 )
15
)
- log ( 24.1646 )
=
log ( .5367 )
-1.3832
= = 5.12 ~ 5
-.2703

Multilevel Plans
In an ingenuous extension of the Dodge CSP-1 concept, Lieberman and Solomon (1955)
conceived the idea of sampling fewer items as quality gives increasing evidence of being
acceptable. This notion resulted in the so-called multilevel plan (MLP), which reduces the
sampling frequency as successively more product is passed without finding a defective.
This involves less inspection than CSP-1 under certain conditions to achieve the same
AOQL. The AOQL given for the multilevel plans assumes that the production process is in
control as in the Dodge plans.
Based on a Markov chain approach, the plans thus produced may be characterized theo-
retically as a random walk with reflecting barriers.
The procedure, allowing the possibility of infinite levels, is as follows:

1. Specify
i as the clearing interval
f as the initial sampling frequency
k0 as the maximum number of levelsto beused
2. Set k = 1 and begin 100% inspection.
3. After i units in succession have been found without a defective, sample at a rate
of f k.
4. If i sampled units are found free of defects, increase k by one and go to step 3.
However, k must not exceed k0, that is k ≤ k0.
5. If a defective is found, decrease k by one and go to step 3. If k = 0, go to step 2.
424 Acceptance Sampling in Quality Control

Start

.
Inspect i
Replace successive
defective units
with
nondefective

Defective No defective
found found

k=1

Randomly sample
k
fraction f units

Defective No defective
found in 1 found in i

k=1 k>1 k < k0 k = k0

k=k–1 k=k+1

FIGURE 15.7 
Multilevel procedure.

While the number of levels in a multilevel plan may be unrestricted, that is, k0 = ∞, it is
often desirable to stop the progression of levels at a certain number of stages. For this rea-
son, a value of k0 may be specified at the outset, which k is not allowed to exceed. Thus, we
have two-level plans (k0 = 2), three-level plans (k0 = 3), and so forth. It should be noted that
when k0 = 1, the multilevel plan reduces to the Dodge CSP-1 plan.
A schematic representation of the multilevel plan is presented in Figure 15.7.
Lieberman and Solomon have provided charts, similar to those of CSP-1, to determine
the values of f and i for specified AOQL. Figure 15.8 shows curves for the infinite-level
plans (k0 = ∞) as solid lines, as contrasted with the Dodge CSP-1 equivalent (k0 = 1) as dot-
ted lines. Figure 15.9 gives the AOQL curves for a two-level plan (k0 = 2).
For example, it can be seen from Figure 15.9 that the two-level multilevel plan i = 38,
f  =  .10 has an AOQL of 4% compared to a 2.9% AOQL for a CSP-1 plan with the same
f and i. From Figure 15.8, an infinite-level plan having 4% AOQL and a clearing interval of
i = 38 would require f = .27.
For an infinite-level plan with defective units replaced by good items, measures can be
determined as follows at fraction defective p for a plan having AOQL = A.
Initial sampling frequency

(1 - A )
i

f =
1 - (1 - A )
i

Continuous Sampling Plans 425

1.00
0.80 k=∞
0.60 k=1
0.50
0.40
0.30
0.20
8 6 54 3 2 0.5 0.3 0.2
0.2 0.1 0.05 0.03 0.02 0.01 AOQL in %
0.10
f

0.08
0.06
0.05 10
0.04
0.03
0.02
10 8 6 5 4 3 2 0.5 0.3 0.2 0.1 0.05 0.03 0.02 0.01
0.01
10 20 3040 60 80100 200 400 600 1,000 2,000 4,000 10,000 20,000 40,000
Number of units (i)

FIGURE 15.8 
Multilevel AOQL curves for k0 = 1, ∞. (Reprinted from Lieberman, G.J. and Solomon, H., Ann. Math., Stat., 26,
696, 1955. With permission.)

1.00
0.80
0.60
0.50
0.40
0.30
0.20

AOQL in %
0.10
f

0.08
0.06
0.05
0.04
0.03
0.02 10 8 6 5 4 3 2 1 0.5 0.3 0.2 0.1 0.05 0.03 0.02 0.01

0.01
10 20 30 40 60 80100 200 400 600 1,000 2,000 4,000 10,000 20,000 40,000
Number of units (i)

FIGURE 15.9 
Multilevel AOQL curves for two-level plan, k0 = 2. (Reprinted from Lieberman, G.J. and Solomon, H., Ann. Math.
Stat., 26, 698, 1955. With permission.)

Average outgoing quality limit

1/i
æ f ö
AOQL = 1 - ç ÷
è 1+ f ø

Fraction defective pM at which AOQL occurs

pM = AOQL
426 Acceptance Sampling in Quality Control

Average fraction inspected

( (1 - A ) / (1 - p ) ) - 1
i

F¥ = , p>A
( (1 - A ) / (1 - p ) ) - 2 (1 - A )
i i

F¥ = 0, p £ A

Average outgoing quality

AOQ ¥ = p ( 1 - F¥ )

Thus, for the infinite level plan i = 38, f = .27 having AOQL = .04, we have

(1 - .04 )
38

f = = .27
1 - ( 1 - .04 )
38

with
1/38
æ .27 ö
AOQL = 1 - ç ÷ = .04
è 1 + .27 ø

which occurs at

pM = .04

The AFI at fraction defective p = .041 is

( (1 - .04 ) / (1 - .041) ) - 1
38

F¥ =
( (1 - .04 ) / (1 - .041) ) - 2 (1 - .04 )
38 38

1.0404 - 1
= = .066
1.0404 - .4240

and

AOQ ¥ = .041 ( 1 - .066 ) = .038

Note that the AFI at p = AOQL = .04 is

AFI = F = 0

hence

AOQ ¥ = .04 ( 1 - 0 ) = .04

Continuous Sampling Plans 427

The mathematical development of multilevel plans is described in the Lieberman and

Solomon (1955) paper. In discussing the advantages of the multilevel procedure, the
authors state (p. 686) that their purpose was “…to consider an extension of Dodge’s first
plan which (1) allows for smoother transition between sampling inspection and 100%
inspection, (2) requires 100% inspection only when the quality submitted is quite inferior,
and (3) allows for a minimum amount of inspection when quality is definitely good.” Burr
(1976) has pointed out, however, that f must be fairly large “…in order to avoid extremely
low fractions on higher powers of f. This makes the saving small at f or even f  2, and the
multilevels make scheduling of workloads difficult.”
In any event, the multilevel procedure provides a useful alternative in the application of
continuous sampling plans and has found application, “to a variety of products, ranging
from EAM cards to very complicated equipment…with substantial savings,” as reported
by Ireson and Biedenbender (1958).

Tightened Multilevel Plans

A set of tightened multilevel plans has been developed by Derman et al. (1957). They
offered “three generalizations of MLP, accomplished by altering the manner in which the
transition can occur….” One of these, the simplest, will be discussed here. It is the tightest
and has been labeled MLP-T.
The MLP-T plan is simply a multilevel plan that requires a switch all the way back to
100% inspection, at any level, whenever a defective unit is found. This provides quick rec-
tification in the event of a shift in quality.
Measures for the MLP-T infinite level plan when defectives are replaced with effective
units are the following:

Average fraction inspected

F=
(
1 - ( 1 - p ) /f
i
) f > (1 - p )
i

1 - (1 - p )
i

f £ (1 - p )
i
f =¥

Average outgoing quality

p (1 - p ) æ 1 - f ö
i

f > (1 - p )
i
AOQ = i ç ÷
1 - (1 - p ) è f ø
f £ (1 - p )
i
AOQ = p

Average outgoing quality limit

AOQL = 1 - f 1/i

with

pM = AOQL
428 Acceptance Sampling in Quality Control

For the infinite level plan i = 38, f = .27 at p = .04, we have

f = .27 > ( 1 - p ) = .21

i

so

F=
(
1 - ( 1 - .04 ) /.27
38
) = .273
1 - ( 1 - .04 )
38

.04 ( 1 - .04 ) æ 1 - .27 ö

38

AOQ = 38 ç ÷ = .029
1 - ( 1 - .04 ) è .27 ø
AOQL = 1 - .27 1/38 = .034

and

pM = .034

Here, we see AFI is increased over the corresponding multilevel plan while the AOQ at
p = .04 and the AOQL are decreased by the quick return at 100% inspection.

Block Continuous Plans

Both the multilevel and the Dodge continuous plans basically assume a steady flow of
production with no attempt to segregate the product into lots or segments. While either of
these procedures may be carried by sampling at random fraction f units from successive
segments of a given size, special plans have been designed for this purpose and may be
characterized as block continuous plans. These plans divide the sequence of production
into successive blocks, taking a prescribed sample from each block. A specified proportion
of the block is screened once a signal for 100% inspection is given. Block continuous plans
are easily adapted to the inspection of successive lots and are useful when the process itself
generates natural segments. It should be noted that while the Dodge plans essentially fix
both i and f, the multilevel plans fix i but allow the fraction inspected to vary with the level
of the plans. The block continuous plans typically sample one unit from a group of items of
specified size (that is f fixed) and allow the clearing interval to vary. Action is usually taken
on the cumulative number of defectives found.

Wald–Wolfowitz Plan
The first block continuous plans were proposed by Wald and Wolfowitz (1945). Of the
three plans they proposed, only one will be discussed here, their statistical process control
(SPC) plan. They divide the production flow into segments of size N0 that are sampled by
Continuous Sampling Plans 429

taking one item from groups of size 1/f to achieve a sampling frequency of f. The plan is
applied as follows for specified N0, M*, and f:

1. Demark the flow of production into segments of fixed size N0.

2. Break each segment into fN0 groups of size k = 1/f.
3. Start with partial inspection of one item from each group.
4. Continue accumulating the sum of defectives found, Σd, until the following are
performed:
a. M* defectives are found and then begin 100% inspection of the remainder of
the segment starting with the next group.
b. The segment has been completely partially inspected.
5. Repeat the procedure anew on the next segment.

A schematic representation of the procedure is given in Figure 15.10.

This procedure can be adapted to the inspection of a sequence of lots by substituting the
word lot for the word segment in the previous steps. In such applications, it is often convenient
to draw successive samples from the entire lot rather than from individual groups. In this case,
a sample of fN0 items is taken successively at random from the lot. As soon as the number
of defectives found equals M* sampling is discontinued. At this point, if the M*th defective
occurred on the N′th sample, an additional (N0 − (1/f)N′) units are 100% inspected and the lot is
released. If fN0 items are sampled without reaching M* defectives, the lot is also released. This
procedure guarantees the same UAOQL as the procedure for sampling from groups.
The Wald–Wolfowitz procedure guarantees the AOQL regardless of the state of control
of the process. For this plan, when defectives are replaced with good items,
M * (1 - f )
UAOQL =
fN 0

Start

Inspect next production

segment of N0 items

Randomly sample 1 unit

from next group of 1/f
items within segment Accept
group

Σd ≥ M* Σd < M*

Inspect 100%
f N0 < f N0
of remainder
sampled sampled
of segment

FIGURE 15.10 
Wald–Wolfowitz SPC procedure.
430 Acceptance Sampling in Quality Control

so that, given that UAOQL = A, N0, and f, a plan can be set up using

AfN 0
M* =
(1 - f )
Other measures of these plans at fraction defective p are as follows:

Average fraction inspected

A (1 - f )
M * -1

å ( M * - 1) ( C ) p ( 1 - p )
fN0 fN0 - i
AFI = F = 1 - + i
i
p pfN 0
i =0

Average outgoing quality

é M * -1
fN0 - i ù
å ( )
1
AOQ = A ê1 - ( M * - i ) CifN0 pi ( 1 - p ) ú
êë M * i = 0 úû

As an example of application, consider setting up a plan that is to have UAOQL = A = .029

and f = .1 for production segments of N0 = 310. Then

.029 ( .1) 310

M* = = .999 ~ 1
( 1 - .1 )

The plan would be applied to segments of 310 units. From each segment, one unit would
be sampled from each of .1(310) = 31 groups of size 1/.1 = 10. As soon as a defective was
found in a segment, the remaining groups in the segment would be 100% inspected before
starting afresh with the next segment. For this plan at fraction defective, p = .054,

.029 (1 - .1) 1 C 31 .0540 1 - .054 31

AFI = 1 - +
.054 .054 ( .1) 310
( ) 0 ( ( ))
= 1 - .537 + .538 ( .179 ) = .559
é 1 31 ù

ë 1
( )
AOQ = .029 ê1 - ( 1) C031 .0540 ( 1 - .054 ) ú
û
= .029 ( 1 - .179 ) = .024

Girshick Plan
M.A. Girshick (1954) has provided a modification of the Wald–Wolfowitz approach that
avoids the necessity for segmenting production, but that achieves essentially the same
result. The procedure is as follows for specified f, m, and N:

1. Divide the flow of production up into groups of size 1/f.

2. Start with partial inspection of one item from each group.
3. Cumulate the number of defective Σd and the number of samples taken, n.
Continuous Sampling Plans 431

4. When the cumulative number of defectives equals m, compare the number of sam-
ples inspected to the integer N.
5. a. If n ≥ N, the product previously inspected is confirmed as good.
b. If n < N, 100% inspect the next N − n groups (i.e., (N − n) (1/f) units) replacing
defectives with good.
6. Start anew.

The Girshick procedure guarantees

UAOQL £
(1 - f ) m
N
regardless of the state of control of the process. It will be seen that this is essentially a modi-
fication of the Wald–Wolfowitz SPC plan with N = fN0 and m = M*. To set up such a plan,
N should be small enough that 100% inspection can reasonably be performed if necessary.
Then for a given UAOQL = A, f, and N,
NA
m=
(1 - f )
If UAOQL = .029, f = .1, and N = 310, the plan is essentially the same as the previously
mentioned Wald–Wolfowitz example if
310 ( .029 )
m=  10
( 1 - .1 )
without the necessity for setting up arbitrary divisions on the flow of production. The
Girshick (1954) monograph presents the mathematical characterization and measures of
the procedure.
Considering block continuous plans in general, it would seem that the Wald–Wolfowitz plans
are particularly well suited where lot inspection is involved or when the production stream is
naturally divided into segments of a given size. The Girshick plan would appear to be quite
good for a continuous flow of product might be set aside as required for screening without the
need for immediate 100% inspection as would be required in the Dodge or multilevel plans.

MIL-STD-1235B
Military standard 1235B entitled Single and Multi-level Continuous Sampling Procedures and
Tables for Inspection by Attributes is a collection of continuous sampling plans indexed by
acceptable quality level (AQL). Like MIL-STD-105E, it was discontinued in 1995; however,
it is a superb collection of continuous sampling plans. An excellent source of the relevant
theory and tables will be found in Stephens (2001). The standard takes care to point out in
its definition of AQL that “for continuous sampling plans, the AQL is an index to the plans,
and has no other meaning.” The AQL index is used to tie the standard to contractual levels
of protection incorporated in contracts involving MIL-STD-105E, MIL-STD-414, and other
such sampling plans. It is not an AQL plan and has no switching rules.
Since continuous sampling plans are usually used, specified, and indexed by AOQL, the
AOQL of the plans given in MIL-STD-1235B is always shown together with the AQL index.
In fact, the plans included were chosen to match, as well as possible, representative values
432 Acceptance Sampling in Quality Control

of the scheme AOQLs of the MIL-STD-105E plans having the AQL index shown. Reference
to Appendix Table T11.20 shows the scheme AOQLs range from 2.9 to 3.2 over the 2.5 AQL
column for nonzero acceptance numbers.*
Thus, the continuous plans indexed under 2.5 AQL in MIL-STD-1235B show AOQL val-
ues of 2.9%. In this way, the results of the MIL-STD-1235B plans correspond to the results
of the MIL-STD-105E system when MIL-STD-105E is used to guarantee AOQLs per para-
graphs 11.3 and 11.4 of that standard.
Five different types of continuous plans are given in MIL-STD-1235B. The user has the
option of selecting the plan that is the most suitable for the inspection situation involved.
The plans included are CSP-1, CSP-2, CSP-F, CSP-T, and CSP-V. These plans are character-
ized in Table 15.2.
Of course, CSP-1 and CSP-2 are the standard plans used in the standard way. The CSP-F
plan is intended for use with short production runs, short periods of production within a
production interval (defined by the standard to be a period of homogeneous quality such
as a shift, but at most a day). The criteria are adjusted to account for a finite period of pro-
duction, N items in length. The CSP-T plans are tightened multilevel plans incorporating
three levels. They are modified from those of Lieberman and Solomon (1955) and Derman
et al. (1957) after the manner of Guthrie and Johns (1958) in that the sampling frequency is
cut in half from level to level, rather than by powers of f as in the conventional multilevel
plans. As a multilevel plan, CSP-T allows a reduction in sampling frequency as quality
improves, reducing the amount of sampling necessary. It is sometimes desirable to cut
the clearing interval rather than sampling frequency with improved quality, particularly
when the sampling inspector cannot be switched to other work. The screening crew will
then have less to do. The CSP-V plans are designed to do just this by reducing the clearing
interval if evidence of superior quality exists.
The structure of MIL-STD-1235B is shown in Figure 15.11. Tables of f and i are provided
for each type of plan. In addition, except for CSP-F, tables of S values are also given as

TABLE 15.2
Type and Purpose of MIL-STD-1235B Plans
Section Plan Type Purpose

2 CSP-1 Standard CSP-1 plan Simple, popular, easy to use and

administer
3 CSP-F CSP-1 procedure with parameters modified Smaller clearing intervals for short
for application to sequence of specified production runs or when long clearing
length intervals are impractical
4 CSP-2 Standard CSP-2 plan Provides against stray defectives and gives
warning that screening crew may be
needed
5 CSP-T Tightened three-stage multilevel plan with Permits reduction of sampling frequency
modified sampling frequencies (f) when superior quality is demonstrated
6 CSP-V Modified CSP-1 procedure with shortened Permits reduction of clearing interval (i)
clearing interval if previous i units free of when superior quality is demonstrated
defectives

* The MIL-STD-105E system AOQLs were calculated using tightened–normal switching only and so correspond
only roughly to the Schilling and Sheesley (1978) values given in Appendix Table T11.20, which incorporates
switching to reduced inspection also.
Continuous Sampling Plans 433

Section 1
general Section 2 Section 3 Section 4 Section 5 Section 6 Section 7
description CSP-1 CSP-F CSP-2 CSP-T CSP-V definitions
Sampling frequency
Description

Description

Description

Description

Description

Description
code letters

Procedure

Procedure
Procedure

Procedure

Procedure
Tables

Tables

Tables

Tables

Tables
Values of S

Values of S

Values of S

Values of S
Values of i

Values of i

Values of i

Values of i
Values of i

FIGURE 15.11 
Structure of MIL-STD-1235B.

criteria to allow termination of excessively long periods of screening. The stopping rule
employed is the “rule n* − i” of Murphy (1959a). That is, clearing is stopped as soon as
a defective is found in any one screening sequence exceeding S units. Clearing is started
anew at the beginning of the clearing interval after corrective action has been taken.
MIL-STD-1235B also provides for the possibility of a check inspection of screened lots.
If the check inspector finds one defective, the customer is to be notified and corrective
action taken on the screening crew. If two defectives are found, product acceptance may
be suspended.
A diagrammatic representation of the application of MIL-STD-1235B is presented in
Figure 15.12, which gives a check sequence for the operation of the standard. This can be
used to ensure that the standard is properly employed.
To illustrate the application of MIL-STD-1235B, consider its use to obtain a CSP-T plan
to be employed on a contract specifying an AQL of 0.65% with the production interval
expected to be about 5000 units. The operation of the CSP-T plan is illustrated in Figure 5-A
of the standard and given here as Figure 15.13. Table 1 of the standard, shown here as
Table 15.3, indicated that Code H may be used. Values of f, i, and S can be obtained from
Tables 5-A and 5-B of MIL-STD-1235B, which correspond to Tables 15.4 and 15.5. They
show that the plan to be employed should be
f = 1/25
i = 217
S = 1396
It should be emphasized that AQL is used here as an index only. In fact, the plan given has
an AOQL equal to 0.79%. Note that, if AOQL protection is desired, an appropriate plan
434 Acceptance Sampling in Quality Control

Start

Determine AQL index

to be used

Determine number units

in production interval

Obtain sampling frequency

code letter

Choose type continuous

plan desired

CSP-1 CSP-F CSP-2 CSP-T CSP-V

Determine period N
for which plan
is to apply

Obtain f, i, and S from

tables in applicable section

Apply plan using given f, i

If number screened
exceeds S, inspection Clearing i
may be suspended for If check inspector
corrective action finds defect in
Screening crew
Sampling f screened product,
starts new count
consumer notified and
of defect-free units
corrective action taken
Product shipped on screening crew

If check inspector
finds second defect,
product acceptance
may be suspended

FIGURE 15.12 
Check sequence for applying MIL-STD-1235B.

can be selected from the tables simply by using the AOQL listed with any desired value
of f to find the corresponding value of i. The value of S for such a plan can also readily be
located if a stopping rule is to be employed. For example, the MIL-STD-105E, Code F, 2.5%
AQL system has an AOQL of 2.9% from Appendix Table T11.20. If a CSP-T plan is to be
employed having equal AOQL protection with a sampling frequency of, say, f = .2, we find
the plan f = .2, i = 29 to be appropriate. For this plan, S = 93.
The theoretical development of the standard and particularly the CSP-F, CSP-T, and
CSP-V plans was largely accomplished by Banzhaf and Brugger of the U.S. Army,
Armament Procurement and Supply Agency, Product Quality Evaluation Division,
Continuous Sampling Plans 435

Start

The screening crew inspects 100% of the units

When i consecutive units are found free of the

defects concerned

The screening crew is released from 100% inspection

and the sampling inspector inspects a fraction, f, of the
units, where the units are selected in a random manner

If the sampling inspector If the sampling inspector

finds one of the finds i consecutive sample
defects concerned units are free of the defects
concerned

The sampling inspector begins inspecting a fraction,

f/2, of the units, which are selected in a random manner

If the sampling inspector If the sampling inspector

finds one of the finds i consecutive sample
defects concerned units are free of the defects
concerned

The sampling inspector begins inspecting a fraction,

f/4, of the units, which are selected in a random manner

When the sampling inspector finds one of the defects concerned

FIGURE 15.13 
Procedure for CSP-T plans. (From United States Department of Defense, Military Standard, Single and Multi-Level
Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-1235B, U.S. Government Printing
Office, Washington, DC, 1981, 41.)

TABLE 15.3
Sampling Frequency Code Letters
Number of Units in Production Interval Permissible Code Letters

2–8 A, B
9–25 A through C
26–90 A through D
91–500 A through E
501–1,200 A through F
1,201–3,200 A through G
3,201–10,000 A through H
10,001–35,000 A through I
35,001–150,000 A through J
150,001 and up A through K

Source: United States Department of Defense, Military Standard, Single

and Multi-Level Continuous Sampling Procedures and Tables for
Inspection by Attributes, MIL-STD-1235B, U.S. Government
Printing Office, Washington, DC, 1981, p. 7.
436 Acceptance Sampling in Quality Control

TABLE 15.4
Values of i for CSP-T Plans
Sampling AQLa (%)
Frequency
Code Letter f 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10.0
A 1/2 87 58 38 25 16 10 7 5
B 1/3 116 78 51 33 22 13 9 6
C 1/4 139 93 61 39 26 15 11 7
D 1/5 158 106 69 44 29 17 12 8
E 1/7 189 127 82 53 35 21 14 9
F 1/10 224 150 97 63 41 24 17 11
G 1/15 266 179 116 74 49 29 20 13
H 1/25 324 217 141 90 59 35 24 15
I 1/50 409 274 177 114 75 44 30 19
J, K 1/100 499 335 217 139 91 53 37 23

0.53 0.79 1.22 1.90 2.90 4.94 7.12 11.46

AOQL (%)

Source: United States Department of Defense, Military Standard, Single and Multi-Level
Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-
STD-1235B, U.S. Government Printing Office, Washington, DC, 1981, p. 42.
a AQLs are provided as indices to simplify the use to this table, but have no other

meaning relative to the plans.

TABLE 15.5
Values of S for CSP-T Plans
Sampling AQLa (%)
Frequency
Code Letter f 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10.0

A 1/2 159 117 77 52 34 22 13 12

B 1/3 256 197 128 80 59 35 25 18
C 1/4 379 253 167 103 78 43 38 24
D 1/5 444 320 210 130 93 54 43 30
E 1/7 725 460 289 188 137 81 59 34
F 1/10 857 619 398 261 189 104 88 58
G 1/15 1254 900 584 368 376 152 126 84
H 1/25 1885 1396 923 545 421 235 198 122
I 1/50 3283 2477 1604 1013 764 408 374 223
J, K 1/100 5753 4541 2948 1754 1341 708 653 391

0.53 0.79 1.22 1.90 2.90 4.94 7.12 11.46

AOQL (%)

Source: United States Department of Defense, Military Standard, Single and Multi-Level
Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-
STD-1235B, U.S. Government Printing Office, Washington, DC, 1981, p. 43.
a AQLs are provided as indices to simplify the use to this table, but have no other

meaning relative to the plans.

Continuous Sampling Plans 437

using a Markov chain approach; their work on the original MIL-STD-1235 standard
has already been cited (Banzhaf and Brugger 1970). The theoretical background of the
Dodge CSP-1 and CSP-2 plans has already been given. CSP-F developed out of the
Markov chain approach to CSP-1 formula derivation presented by Roberts (1965) and
the study by Lasater (1970) of the theory and performance of CSP-1 when applied to a
finite number of units.
Measures of the plans incorporated in MIL-STD-1235B are presented in a compan-
ion document, MIL-STD-1235A-1, Appendices A–D. All plans included in MIL-STD-
1235B are represented except for the CSP-F plans. For each plan, curves are given for
AOQ, AFI, and OC. These are defined by MIL-STD-1235A-1, Appendices A through D,
as follows:

The AOQ for a particular process average is the long run expected percentage of
­defective material in the accepted material, if the associated sampling plan is fol-
lowed faithfully (Figure 15.14).
The AFI is the fraction of product that will be inspected over the long run if the pro-
cess average is a particular value (Figure 15.15).
The OC of a continuous sampling plan describes the percent of product accepted
­during the sampling phases of the plan over the long run if the process average is
a particular value (Figure 15.16).

5 CSP-T
Functional curves
Bottom left = AOQ
Top right = AFI
Bottom right= OC
RQL = 2.50 Target AOQL = 2.80%
4 Sampling Frequency
code letter i f f/2 f/4

A 18 1/2 1/4 1/8

B 22 1/3 1/6 1/12
Average outgoing quality

C 28 1/4 1/8 1/16

D 29 1/5 1/10 1/20
3

1 (A)
(B)
(C)
(D)

0 4 8 12 16 20 24 28
Percent defective of submitted product

FIGURE 15.14 
MIL-STD-1235A-1, Appendix C AOQ curves—CSP-T. (From United States Department of Defense, Military
Standard, Single and Multi-Level Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-
1235B, U.S. Government Printing Office, Washington, DC, 1981, C2.)
438 Acceptance Sampling in Quality Control

100

C
A
Percent of product accepted under sampling

80

60

40

20

0 4 8 12 16 20 24 28
Percent defective of submitted product

FIGURE 15.15 
MIL-STD-1235A-1, Appendix C AFI curves—CSP-T. (From United States Department of Defense, Military
Standard, Single and Multi-Level Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-
1235B, U.S. Government Printing Office, Washington, DC, 1981, p. C3.)

100
Percent of product accepted under sampling

80

60

40

B A
20 DC

0 4 8 12 16 20 24 28
Percent defective of submitted product

FIGURE 15.16 
MIL-STD-1235A-1, Appendix C OC curves—CSP-T. (From United States Department of Defense, Military
Standard, Single and Multi-Level Continuous Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-
1235B, U.S. Government Printing Office, Washington, DC, 1981, p. C3.)
Continuous Sampling Plans 439

The standard states that “curves for CSP-F are not provided, since exact methods for
their determination have not been developed.” MIL-STD-1235C superseded MIL-STD-
1235A-1 in 1988.
Illustrations of the MIL-STD-1235A-1, Appendix C, curves are given here for the multi-
level plan CSP-T: Code D, 2.5% AQL (f = 1/5, i = 29).

MIL-STD-1235B Derivatives
ASTM International Standard E2819
E2819 Standard Practice for Single- and Multi-Level Continuous Sampling of a Stream of Product
by Attributes Indexed by AQL is a practice that establishes tables and procedures for apply-
ing five different types of continuous sampling plans for inspection by attributes using
MIL-STD-1235B as a basis for sampling a steady stream of lots indexed by AQL. This prac-
tice provides the sampling plans of MIL-STD-1235B in ASTM format for use by ASTM
committees and others. It recognizes the continuing usage of MIL-STD-1235B in industries
supported by ASTM. Most of the original text in MIL-STD-1235B is preserved in Sections 6
through 10 of this practice.

Software Applications
Excel
The Excel template Continuous Sampling Plan.xlsm can be used to generate and evalu-
ate a CSP-1 continuous sampling plan. To illustrate its use, suppose we desire a CSP-1
plan for a process with a production interval of 1000 units and a desired AOQL of 2.2%.
In the template, we enter 1000 in cell F6 and the desired AOQL% in cell F8 as shown in
Figure 15.17. Click on the button called “Find a range of acceptable frequency values.”
The macro shows the range of code letters that can be used for this particular length of
production interval. In this example, for an interval of length 1000, the choices of fre-
quencies range from code letter A (1/2) to F (1/10). These choices are also displayed in a
table in columns E through O.
Suppose we choose a frequency of 1/10, f = .1. To determine the value of the clearing
interval, click on the button called “Find value of i.” The macro indicates the value is
i = 57. The i and f values are carried over to the Calculations tab where all of the com-
putations are made to construct the OC, AOQ, and AFI curves (all available on separate
tabs). The Calculations tab shows that for this plan the value of spotty quality pt% is
2.3% while the AOQL is 1.9% (if defectives are replaced, and 1.93% if not replaced) and
the LTPD = 7.61%. If an AOQL of 1.9% is desired with protection against spotty qual-
ity of 2.3%, we have, from Figure 15.2, i = 465 when f = .10. This would result in 100%
inspection until 465 units are found good in succession with random sampling of 10%
of the units thereafter until a defective is found (see Figure 15.18). Also, note that if the
 440

FIGURE 15.17 
Excel template Continuous Sampling Plan.xlsm used to generate a MIL-STD-1235B CSP-1 continuous sampling  plan.
Acceptance Sampling in Quality Control
 Continuous Sampling Plans

FIGURE 15.18 
Probability calculations supporting the Excel template Continuous Sampling Plan.xlsm used to generate a MIL-STD-1235B CSP-1 continuous sampling plan.
441
442 Acceptance Sampling in Quality Control

user desires a nonstandard value of β (other than 10%), this value can be changed in
cell I2 on the Calculations tab.

Snap Sampling Plans!

Snap Sampling Plans! can generate CSP-1, CSP-2, CSP-F, and CSP-T (but not CSP-V)
MIL-STD-1235 continuous sampling plans. Click on the Mil Std 1235 button from
the main menu. To illustrate its use, suppose that we want to obtain a CSP-T plan to
be employed on a contract specifying an AQL of 0.65% with the production interval
expected to be about 5000 units. Note that the AQL is used here as an index only. The
operation of the CSP-T plan is illustrated in Figure 5-A of the standard and given as
Figure 15.13. Table 1 of the standard, shown as Table 15.3, indicated that Code H may
be used. The completed dialog box is presented in Figure 15.19. The collection of CSP-T
plans generated by Snap Sampling Plans! For Levels A–H is given in Figure 15.20. Here,
we see that for Code H the proper CSP-T sampling plan is f = 1/25, i = 217, S = 1396.
These values of i and S agree with those found in Tables 15.4 and 15.5, ­respectively, for
this example.

FIGURE 15.19 
Snap Sampling Plans! dialog box used to generate an MIL-STD-1235B CSP-T continuous sampling plan.
Continuous Sampling Plans 443

MIL-STD-1235
Continuous sampling plan

Production Inspection
size: type: AQL%:
5000 CSP-T 0.65
Sample Corrective
Level Clearance every action Accept Reject
A 58 2 Parts 117 0 1
B 78 3 Parts 197 0 1
C 93 4 Parts 253 0 1
D 106 5 Parts 320 0 1
E 127 7 Parts 460 0 1
F 150 10 Parts 619 0 1
G 179 15 Parts 900 0 1
H 217 25 Parts 1396 0 1

FIGURE 15.20 
Snap Sampling Plans! MIL-STD-1235B CSP-T continuous sampling plan.

Problems
1. Construct a CSP-1 plan for AOQL = 4% that will have a sampling frequency of about
10%. What is the UAOQL for this plan when defectives are replaced by good units?
2. Find a CSP-2 plan with k = i that will afford about the same protection as the plan in
Problem 1. What is its UAOQL when defectives are not replaced with good items?
3. What is the AFI when the proportion defective submitted is .08 for
a. Problem 1?
b. Problem 2?
4. Stopping rule r is to be instituted on the plan of Problem 1. Find r if 50 units may be
produced between successive stops when the process proportion defective is such
that the AOQL is realized (i.e., p = pM).
5. Use Murphy’s procedure to find a CSP-1 plan that will have a 5% AOQL with 25% of
the product inspected when the process average is expected to be 1% defective.
6. Find an infinite multilevel plan that will have an AOQL of 4% with an initial sampling
frequency of about 10%. What is the AOQL if it is used as a tightened multilevel plan?
7. Production is boxed in crates of 24 units with 24 crates to a skid. A UAOQL of 6% is
desired. Construct the appropriate Wald–Wolfowitz plan.
8. What would be the parameters of a Girshick plan corresponding to Problem 7?
9. An MIL-STD-105E scheme is being used with screening of rejected lots to provide
AOQL protection. For the scheme, AQL = 6.5 with lots of size 550. A process change
mandates a change to continuous sampling. At present, a tightened/normal sample
size of 80 is being used, which implies a sample size–lot size ratio of .145 ~ 1/7.
What MIL-STD-1235B, CSP-T, plan should be used? What should be the maximum
number screened before clearing is stopped?
444 Acceptance Sampling in Quality Control

10. Compare the code letters of MIL-STD-1235B to those of MIL-STD-105E. Are they
comparable? That is, if Code J is being used on MIL-STD-105E should J be used on
MIL-STD-1235B?

References
American Society for Testing and Materials, Standard Practice for Single- and Multi-Level Continuous
Sampling of a Stream of Product by Attributes Indexed by AQL, ASTM Standards E2819, Vol. 14.02,
West Conshohocken, PA.
Banzhaf, R. A. and R. M. Brugger, 1970, MIL-STD-1235 (ORD), single and multi-level continuous sam-
pling procedures and tables for inspection by attributes, Journal of Quality Technology, 2(1): 41–53.
Burr, I. W., 1976, Statistical Quality Control Methods, Marcel Dekker, New York.
Derman, C., S. Littauer, and H. Solomon, 1957, Tightened-multi-level continuous sampling plans,
Annals of Mathematical Statistics, 28: 395–404.
Dodge, H. F., 1943, A sampling plan for continuous production, Annals of Mathematical Statistics,
14(3): 264–279.
Dodge, H. F., 1947, Sampling plans for continuous production, Industrial Quality Control, 4(3): 5–9.
Dodge, H. F. and M. N. Torrey, 1951, Additional continuous sampling inspection plans, Industrial
Quality Control, 7(5): 7–12.
Girshick, M. A., 1954, A sequential inspection plan for quality control, Technical report no. 16,
Applied Mathematics and Statistics Laboratory, Stanford University, Stanford, CA.
Guthrie, D. and M. Johns, 1958, Alternative sequences of sampling rates for tightened multi-level
continuous sampling plans, Technical report no. 36, Applied Mathematics and Statistics
Laboratory, Stanford University, Stanford, CA.
Ireson, W. G. and R. Biedenbender, 1958, Multi-level continuous sampling procedures and tables for
inspection by attributes, Industrial Quality Control, 15(4): 10–15.
Lasater, H. A., 1970, On the robustness of a class of continuous sampling plans under certain types
of process models, PhD dissertation, Rutgers—The State University, New Brunswick, NJ.
Lieberman, G. J., 1953, A note on Dodge’s continuous sampling plan, Annals of Mathematical Statistics,
24(3): 480–484.
Lieberman, G. J. and H. Solomon, 1955, Multi-level continuous sampling plans, Annals of Mathematical
Statistics, 26(4): 686–704.
Murphy, R. B., 1959a, Stopping rules with CSP-1 sampling inspection plans, Industrial Quality Control,
16(5): 10–16.
Murphy, R. B., 1959b, A graphical method of determining a CSP-1 sampling inspection plan, Industrial
Quality Control, 16(6): 20–21.
Roberts, S. W., 1965, States of Markov chains for evaluating continuous sampling plans, in
Transactions of the 17th Annual All-Day Conference on Quality Control, Metropolitan Section ASQC
and Rutgers University, New Brunswick, NJ, pp. 106–111.
Stephens, K. S., 1980, The ASQC basic references in quality control: Statistical techniques, in How to
Perform Continuous Sampling (E.J. Dudewicz, Ed.), vol. 2. American Society for Quality Control,
Milwaukee, WI.
Stephens, K. S., 2001, The Handbook of Applied Acceptance Sampling, ASQ Quality Press, American
Society for Quality, Milwaukee, WI.
United States Department of Defense, 1981, Military Standard, Single and Multi-level Continuous
Sampling Procedures and Tables for Inspection by Attributes (MIL-STD-1235B), U.S. Government
Printing Office, Washington, DC.
Wald, A. and J. Wolfowitz, 1945, Sampling inspection plans for continuous production which insure
a prescribed limit on the outgoing quality, Annals of Mathematical Statistics, 16(1): 30–49.
16
Cumulative Results Plans

Except for continuous sampling plans, the acceptance sampling plans discussed so far
have been applied on an individual lot-by-lot basis. The acceptable quality level (AQL)
schemes incorporated in MIL-STD-105E and MIL-STD-414 do, in fact, utilize the results
from the most recent lots as part of the switching rules, but the acceptance criteria applied
to any one lot do not specifically incorporate the results of the inspection of the immedi-
ately preceding lots.
The continuous sampling plans discussed in Chapter 15 require knowledge of the
results from the immediately preceding samples as part of the action rule for any sample
inspected. This is particularly evident in CSP-2 and CSP-3, but applies to all the plans
discussed. Thus, continuous sampling plans are a member of the class of, so-called, cumu-
lative results plans. Other members include skip-lot plans, chain sampling plans, and the
Cone and Dodge (1963) cumulative results plan. Cumulative results plans, however, usu-
ally involve lot-by-lot inspection of a stream of product. It is the purpose of this chapter
to examine these plans as a means of dealing with the frequent problem of minimizing
sample size because of economic constraints while still affording a reasonable amount of
protection.
In general, cumulative results plans require certain assumptions about the nature of the
inspection to be met. As described by Dodge (1955a) in introducing chain sampling plans,
these are as follows:

1. The lot should be one of a continuing series of supply.

2. Lots should normally be expected to be of the same quality.
3. The consumer should have no reason to believe that the lot to be inspected is
poorer than any of the immediately preceding lots.
4. The consumer must have confidence in the supplier, in that advantage would not
be taken of a good record to slip in a substandard lot.

Under these conditions, it is reasonable to use the record of previous inspections as a means
of reducing the amount of inspection required on any given lots.

Skip-Lot Sampling Plans

SkSP-1
Continuous sampling plans are intended to be applied on individual units produced in
sequence from a continuing source of supply. The principles of continuous sampling can,
however, be applied to individual lots received in a steady stream from a trusted supplier.

445
446 Acceptance Sampling in Quality Control

Just as units are “skipped” during the sampling phase of a continuous sampling plan,
so lots may be skipped (and passed) under an analogous skip-lot plan. It is surprising, but
fortuitous, that skip-lotting can actually increase protection per unit sampled.
The first skip-lot plan, SkSP-1, was introduced by Dodge (1955b) as an adaptation of
CSP-1 to the inspection of raw materials purchased regularly from a common source.
Materials are often inspected using bulk sampling procedures with an output of one labo-
ratory determination per lot. Disposition of the lot is in accord with whether or not the
laboratory determination conforms to specification requirements. Thus, by regarding each
lot of raw material as a single “unit”—either conforming or not conforming to specifica-
tions—continuous sampling plans are readily applied. In this case, however, the average
outgoing quality limit (AOQL) provides an upper bound on the average percentage of
accepted lots that will be nonconforming in the long run. Similarly, other measures of con-
tinuous sampling plans may be interpreted as referring to “lots” rather than units. Dodge
(1955b) contrasts CSP-1 and SkSP-1 for 2.0% AOQL in tabular form, shown in Table 16.1.
The skip-lot procedure may be represented schematically as in Figure 16.1, which
­presents the skip-lot concept.
In such applications, rejected lots are not usually 100% inspected or replaced by known
good material. They are simply rejected and disposed of. When this is the case, Dodge
(1955b) points out that i should be increased by one in CSP-1 plans to maintain the protec-
tion guaranteed.
While any continuous sampling plan may be used in this application, Dodge (1955b)
proposes a CSP-1 plan with 2% AOQL for general use:

Procedure A-1 (Each nonconforming lot corrected or replaced by a conforming lot.)

i = 14, f = 1/2
Procedure A-2 (Each nonconforming lot rejected and not replaced by a conforming
lot.) i = 15, f = 1/2

Other choices of f and i can be made from Figure 15.2 for CSP-1 or by using appropriate
means for other continuous sampling plans.

TABLE 16.1
Comparison of CSP-1 and SkSP-1
CSP-1 (Product Units) SkSP-1 (Lots of a Raw Material)

Series of units. Series of lots (or batches).

Inspect a unit. Make laboratory analysis of a sample of
material.
Defective unit (a unit that fails to meet the applicable Nonconforming lot (a lot whose sample fails to
specification requirement). meet the applicable specification requirement).
Units in succession found clear of defects. Lots in succession found conforming.
Incoming % defective: % of incoming units that are Incoming % defective: % of incoming lots that
defective. are nonconforming.
Meaning of 2% AOQL: an average of not more than 2% of Meaning of 2% AOQL: an average of not more
accepted units will be defective for the characteristic under than 2% of accepted lots will be
consideration. nonconforming for the characteristic under
consideration.
Source: Reprinted from Dodge, H.F., Ind. Qual. Control, 11(5), 4, 1955b. With permission.
Cumulative Results Plans 447

Start

Inspect i
successive
lots

Lot No lot
rejected rejected

Randomly inspect
a fraction f of
the lots

Lot No lot
rejected rejected

FIGURE 16.1
Skip-lot procedure.

SkSP-1 plans are implemented as follows, illustrated using procedure A-2 on lots that are
rejected and not replaced:

1. Apply plan separately to each characteristic under consideration. When several

characteristics are involved, try to test at least one characteristic in each lot. That
is, lots skipped for one characteristic should be tested for another so that each lot
receives tests on some of the characteristics.
2. Start by testing each lot consecutively as received until 15 lots in succession are
found conforming.
3. When 15 lots in succession are found conforming, test only half the lots at random.
Accept the lots not tested.
4. When a lot is rejected, revert to step 2.

This will guarantee that on an average a maximum of 2% of the accepted lots will be
nonconforming. Procedure A-1, for use when nonconforming lots are replaced with
­
­conforming lots, is the same with i = 14 rather than i = 15 as given in the earlier text.
Other skip-lot plans of this sort can easily be devised using the procedures for the
­continuous sampling plans of Chapter 15.

SkSP-2
While SkSP-1 was intended to be used in circumstances leading to a simple and absolute
go/no-go decision on each lot, the continuous sampling approach to skipping lots may
be utilized when a standard sampling plan is applied to each lot. When sampling plans
448 Acceptance Sampling in Quality Control

are used, a lot is accepted or rejected with an associated producer’s or consumer’s risk.
These risks have been factored into the skip-lot procedure by Dodge and Perry (1971) in
their development of SkSP-2. These plans are intended to be applied to a series of lots or
batches of discrete items that are sampled using a standard “reference” sampling plan.
Two stages of sampling are distinguished:

Normal inspection: Use of the reference plan on every lot as received.

Skipping inspection: Use of the reference plan on a randomly selected fraction f of the
lots. Skipped lots are accepted.

CSP-1 is applied to the inspection results to determine whether normal or skipping inspec-
tion is to be used. The procedure is as follows:

1. Start with normal inspection of each lot using the reference plan.
2. When i consecutive lots have been accepted, switch to skipping inspection, inspect-
ing a fraction f of the lots at random as received.
3. When a lot is rejected, switch to normal inspection, step 1.
4. Screen each rejected lot replacing nonconforming units with conforming units.

Thus, an SkSP-2 plan is specified by

1. The reference sampling plan applied to each lot

2. i, the clearing interval
3. f, the sampling frequency

Application of the plan n = 20, c = 1, i = 4, and f = .20 would proceed as follows:

1. Inspect consecutive lots using the reference plan n = 20, c = 1 on each under normal
inspection.
2. When i = 4 lots in succession have passed the reference plan, go to skipping
­inspection. Inspect at random only a fraction f = .20 of the lots using the reference
plan on each. Pass the lots not inspected.
3. When a lot is rejected, revert to normal inspection, step 1.
4. Screen all rejected lots.

Measures of SkSP-2 are analogous to SkSP-1. Let

P be the probability of acceptance of reference plan

U = (1 − Pi)/(Pi(1 − P)), which is the expected number of lots during normal inspection
V = 1/f(1 − P), which is the expected number of lots during skipping inspection

Then, measures of the SkSP-2 plan are as follows as derived by Perry (1970) using both a
power series approach, as in the derivation of CSP-1, and Markov chains:

Pa is the probability of acceptance (long-run proportion of lots accepted):

Pa =
(1 - f ) P i + fP
(1 - f ) Pi + f
Cumulative Results Plans 449

where F is the average fraction lots inspected (long-run average fraction of lots
inspected)

U + fV f
F= =
U +V (1 - f ) Pi + f
ASNSK is the average sample number (long-run average sample size over lots
inspected):

ASN SK = F ( ASN R )

where ASNR is the average sample number of reference plan.

For a single-sampling plan (n, c), this becomes

ASN SK = Fn

AOQL1 = unit AOQL, which is the upper bound on the long-run average proportion of
outgoing product units that are defective:

Y
AOQL1 =
n

where selected values of Y from Perry (1970) are given in Appendix Table T16.1 for single-
sampling reference plans for various values of c, f, and i (assume type B sampling).
AOQL2 = lot AOQL, which is the upper bound on the long-run average proportion of
outgoing lots that are nonconforming, that is, lots that would have failed the reference plan:

AOQL 2 = AOQL of CSP - 1 plan having same f and i

Values of the lot AOQL2 have been tabulated by Perry (1973a) and are shown in Table 16.2.
Consider the SkSP-2 plan n = 20, c = 1, i = 4, and f = .25. Assume an incoming propor-
tion defective of p = .05. For the reference plan, at this fraction defective, P = .736. Also,
AOQL = .042. The measures of the skip-lot plan are

TABLE 16.2
Values of Lot AOQL2 for Given f and i
f
i 2/3 3/5 1/2 2/5 1/3 1/4 1/5

4 0.034 0.044 0.060 0.081 0.098 0.126 0.148

6 0.024 0.030 0.042 0.057 0.069 0.089 0.105
8 0.018 0.023 0.032 0.044 0.053 0.069 0.081
10 0.015 0.019 0.026 0.035 0.043 0.056 0.066
12 0.013 0.016 0.022 0.030 0.037 0.047 0.056
14 0.011 0.014 0.019 0.026 0.031 0.041 0.048
Source: Reprinted from Perry, R.L., J. Qual. Technol., 5(3), 130, 1973. With permission.
450 Acceptance Sampling in Quality Control

Probability of acceptance

(1 - .25 )(.736 ) + .25 (.736 ) = .860

4

Pa =
(1 - .25 )(.736 ) + .25
4

Average fraction lots inspected

.25
F= = .532
(1 - .25 )(.736 )
4
+ .25

Average sample number

ASN SK = .532 ( 20 ) = 10.6

Unit AOQL

.9861
AOQL1 = = .049
20

Lot AOQL

AOQL 2 = .122

Thus, at this fraction defective, the probability of acceptance is higher than the reference
plan under skip-lotting, from .736 to .860, while the AOQL is increased only slightly from
.042 to .049.
The operating characteristic (OC) curve for the plan n = 20, c = 1, i = 4, and f = .25 is given
in Figure 16.2. Also included is the OC curve for a similar plan with i = 10 and for the ref-
erence plan. Note that skip-lotting swells the shoulder of the OC curve, improving the

1.00
f = 1/4
0.90
1. i = 4
0.80 3
2. i = 10
0.70 3. Reference plan
2
0.60 n = 20, c = 1

0.50
Pa

0.40 1

0.30
0.20
0.10
0
0 0.020 0.040 0.060 0.080 0.100 0.120 0.140
p

FIGURE 16.2
OC curves for some skip-lot plans. (Reprinted from Perry, R.L., J. Qual. Technol., 5(3), 125, 1973. With permission.)
Cumulative Results Plans 451

1
20.00
18.00

16.00
2
14.00

12.00
3
ASN

10.00

8.00

6.00
1. Reference plan: n = 20, c = 1
4.00 2. f = 2/3, i = 4
3. f = 1/5, i = 4
2.00
0
0 0.020 0.040 0.060 0.080 0.100
p

FIGURE 16.3
ASN curves for some skip-lot plans. (Reprinted from Perry, R.L., J. Qual. Technol., 5(3), 127, 1973. With permission.)

producer’s risk, but leaves the lot tolerance percent defective (LTPD) essentially unchanged
from the reference plan. A typical set of ASN curves is illustrated in Figure 16.3. Note the
substantial reductions in ASN in regions of good quality (Pa > .5).
Using unity values from the Poisson distribution, Dodge and Perry (1971) developed
a table, which can be used to easily derive skip-lot plans to match single-sampling plans
having acceptance numbers from 2 to 10. It is presented here as Appendix Table T16.2. It
shows alternate skip-lot plans having a given operating ratio (OR). Of course, the ORs of
skip-lot plans cover a wide range of possible values and are not restricted to those of the
single-sampling plans. Additional unity values have been given by Perry (1970).
To use the Dodge–Perry table, given the OR desired, perform the following:

1. Find the OR listed closest to the OR desired, where OR = p.10/p.95.

2. Find the corresponding single-sampling plan for that OR by using the associated
acceptance number c and sample size n found by dividing the value np.95 given for
the single-sampling plan by the value of p.95 used to compute the OR.
3. To find the matched SkSP-2 plan, consider the following:
a. Pick convenient values of f and i listed for the OR given.
b. The reference plan will have the value of c listed and sample size found by
dividing the value of np.95 for the SkSp-2 plan by the value of p.95 used to obtain
the OR.

For example, suppose a plan is desired such that p.95 = .01 and p.10 = .04. The OR is
.04
OR = =4
.01
452 Acceptance Sampling in Quality Control

1. The closest OR for a single-sampling plan is 4.057.

2. The single-sampling plan closest to OR = 4 has c = 4 and

1.97
n= = 197
.01

3. A corresponding SkSP-2 plan would have f = .5, i = 4, with a reference plan having
c = 3 and a sample size

1.645
n= = 164.5 ~ 165
.01

The table shows the ratio of these two sample sizes to be .830.

Of course, Appendix Table T16.2 can be used in a number of ways to derive and evalu-
ate SkSP-2 plans. For example, an SkSP-2 plan matching the single-sampling plan n = 200,
c = 3 has f = .2, i = 14, c = 2 and a reference plan sample size of 73.1% of the matched single-
sampling plan or

n = .731 ( 200 ) = 146.2 ~ 147

This utilizes the column of ratios of reference sample sizes of SkSP-2 plans to match single-
sampling plans. From this column, it is apparent that, in matching single-sampling plans,
not only are some lots skipped, but also the sample size and acceptance number of the
reference plan applied to the lots inspected will be less than that of the matched single-
sampling plan. This is because, in SkSp-2 inspection, a tight reference plan is used a frac-
tion of the time to achieve the same result as consistent application of a looser-matched
single-sampling plan. Sizable savings in average sample size can be achieved by using
SkSp-2 plans. Of course, the gain is achieved by not inspecting all the lots, which may, at
times, be a serious disadvantage.
The skip-lot concept has been extended by Perry (1973b) to achieve greater flexibility in
application by using two stages. Since the skip-lot plans may be derived from any continu-
ous sampling procedure, not just CSP-1, three procedures are proposed by Perry based on
other continuous plans:

Plan 2L.1. Two-stage plan based on the multilevel plan of Lieberman and Solomon
(1955)
Plan 2L.2. Two-stage plan based on the tightened multilevel plan of Derman et al.
(1957) as extended by Guthrie and Johns (1958)
Plan 2L.3. A unique two-stage plan developed by Perry that determines the sampling
rate on the basis of the number of consecutive lots accepted

The plans presented allow for any combination of sampling rates to be used and thus are
more general than those of the conventional multilevel plans that prescribe a geometric
relationship between sampling rates. The details of these plans together with an exposition
of their properties are presented in Perry (1973b).
Cumulative Results Plans 453

Chain Sampling Plans

A prime example of the use of cumulative results to achieve a reduction of sample size
while maintaining or even extending protection can be found in the chain sampling plans
introduced by Dodge (1955a). These plans were originally conceived to overcome the
problem of lack of discrimination in c = 0 sampling plans. The procedure was devel-
oped to “chain” together the most recent inspections in a way that would build up the
shoulder of the OC curve of c = 0 plans. This is especially desirable in situations in which
small samples are demanded because of the economic or physical difficulty of obtaining
a sample.

ChSP-1
The original chain sampling inspection procedure as developed by Dodge (1955a) is as
follows:

1. From each lot, select a sample of n units.

2. Accept if
a. No defectives are found in the sample.
b. One defective is found in the sample, but no defectives were found in the
­previous i samples of n.
3. Reject otherwise.

Specification of n and i completely determines a ChSP-1. The ChSP-1 procedure is illus-

trated schematically in Figure 16.4.

Start

Inspect sample
of n
from current lot

No One Two or more

defectives defective defectives
found found found

No defectives One defectives

in previous in previous
i lots i lots

Accept Reject

FIGURE 16.4
ChSP-1 procedure.
454 Acceptance Sampling in Quality Control

The OC curve of the ChSP-1 procedure is shown by Dodge (1955a) to be determined by

i
Pa = p ( 0 ) + p ( 1) éë p ( 0 ) ùû

where p(x) is the probability of x defectives in a sample of n.

Clearly, when i = ∞, ChSP-1 reverts to the single-sampling plan having c = 0. Thus, for the
plan n = 10, i = 2 when p = .10, we have

p ( 0 ) = .3487 p ( 1) = .3874

so

Pa = .3487 + .3874 ( .3487 ) = .3958

2

Of course, average outgoing quality can be found by

AOQ = pPa

which gives

AOQ = .10 ( .3958 ) = .04

A comparison of ChSP-1 plans for n = 10 and several values of i is shown in Figure 16.5
from Dodge (1955a). The solid line represents the single-sampling plan n = 10, c = 0. It
illustrates the shoulder built up on the c = 0 OC curve when the chain sampling criterion is
imposed. The curve for i = 1 is shown dotted since its use is not recommended by Dodge.
Note that in the region of low probability of acceptance, the OC curves for various values
of i seem to coincide with the exception of the curve for i = 1.
Chain sampling plans are easily evaluated using Poisson unity values developed by
Soundararajan (1978a). Table 16.3 shows values of np corresponding to various probabilities
of acceptance for values of i from 1 to 6 and for i = ∞, which is simply the single-­sampling
plan with c = 0. Notice that with values of low probability of acceptance, the unity values
are those of the corresponding single-sampling plan having c = 0. When divided by sample
size, the unity values give the proportion defective corresponding to the probability of
acceptance shown. Thus, for the plan n = 10, i = 3, we have the following

Pa p

.99 .0057
.95 .0139
.50 .0768
.10 .2303
.05 .2996
.01 .4605

A table for constructing ChSP-1 plans has also been given by Soundararajan (1978a) and
is presented here as Appendix Table T16.3. Based on Poisson unity values, it allows deter-
mination of a ChSP-1 plan from the desired operating ratio p2/p1. Values of np are given
Cumulative Results Plans 455

100

Sample size: 10
ChSP-1 plan
Acceptance criterion:
90 1=i c = 0, except c = 1
2 if no defects in
immediately preceding
3 i samples of 10.
80
4
5
70

60

50

SS plan
40 n = 10, c = 0

30

20

10

0
0 5 10 15 20

FIGURE 16.5
OC curve for ChSP-1 (n = 10; i = 1, …, ∞). (From Dodge, H.F., Ind. Qual. Control, 11(4), 11, 1955. With permission.)

TABLE 16.3
Unity Values for Evaluation of ChSP-1 OC Curves
f
i 0.99 0.95 0.50 0.10 0.05 0.01

1 0.086 0.207 1.007 2.490 2.996 4.605

2 0.067 0.162 0.840 2.325 2.996 4.605
3 0.057 0.139 0.768 2.303 2.996 4.605
4 0.051 0.124 0.733 2.303 2.996 4.605
5 0.046 0.114 0.714 2.303 2.996 4.605
6 0.042 0.106 0.703 2.303 2.996 4.605
∞ 0.010 0.051 0.693 2.303 2.996 4.605
Source: Reprinted from Soundararajan, V., J. Qual. Technol., 10(2), 56, 1978a. With permission.
456 Acceptance Sampling in Quality Control

at .95 and .10 probability of acceptance for i = 1 to 10 and i = ∞. Also, the AOQL of the
ChSP-1 procedure can be found from sample size using nAOQL values or from the pro-
ducer’ quality level p1 using values of AOQL/p1. Further, the proportion defective pM at
which the AOQL occurs can be determined from values of npM.
For example, we find for the plan n = 10, i = 3, the following properties are given by
Appendix Table T16.3.

Pa = .95 at p = .0139

Pa = .10 at p = .2303

p2/p1 = 16.568

AOQL = 2.798(.0139) = .0389

pM = .0902

It is sometimes desirable to construct a ChSP-1 plan having a specified AOQL. For this pur-
pose, Table 16.4 has been developed by Soundararajan (1978a), showing values of sample
size n and cumulative results criterion (CRC) i. We see that an AOQL of .04 is achieved by
the plan n = 10, i = 3.
While Poisson unity values provide an excellent device for constructing plans as an
approximation to the binomial distribution and are exact when dealing with defects, it is
sometimes desirable to have exact tables for the selection of plans based on the binomial
distribution itself. Appendix Table T16.4 from Soundararajan (1978b) gives ChSP-1 plans
indexed by AQL (p.95 value) and LTPD (p.10 value). It shows that, using the binomial
distribution, for p.95 = .015 and p.10 = .220 the plan n = 10, i = 3 would be appropriate.
In addition, plans may be constructed for a given AQL/AOQL combination (AQL = p.95)

TABLE 16.4
ChSP-1 Plans Having Given AOQL
AOQL (%)
i 0.10 0.25 0.50 0.75 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 6.0 7.0 8.0 9.0 10.0

1 504 202 101 68 51 34 26 21 17 15 13 12 11 9 8 7 6 5

2 420 168 89 56 42 28 22 17 14 12 11 10 9 7 6 6 5 5
3 389 156 78 52 39 26 20 16 13 12 10 9 8 7 6 5 5 4
4 377 151 76 51 38 26 19 16 13 11 10 9 8 7 6 5 5 4
5 372 149 74 50 38 25 19 15 13 11 10 9 8 7 6 5 5 4
6 369 148 74 50 37 25 19 15 13 11 10 9 8 7 6 5 4 4
7 369 148 74 50 37 25 19 15 13 11 10 9 8 7 6 5 4 4
8 368 148 74 49 37 25 19 15 13 11 10 9 8 7 6 5 4 4
9 368 148 74 49 37 25 19 15 13 11 10 9 8 7 6 5 4 4
10 368 148 74 49 37 25 19 15 13 11 10 9 8 7 6 5 4 4
∞ 368 148 74 49 37 25 19 15 13 11 10 9 8 7 6 5 4 4
Source: Reprinted from Soundararajan, V., J. Qual. Technol., 10(2), 58, 1978a. With permission.
Cumulative Results Plans 457

using Appendix Table T16.5 given by Soundararajan (1978b). For an AQL/AOQL combi-
nation of .015/.035, Appendix Table T16.5 gives the plan n = 15, i = 1, which should give
protection roughly equivalent to the plan n = 20, c = 1 that has p.95 = .018 and AOQL = .035.
Note the obvious saving in sample size.

Two-Stage Plans
Two-stage chain sampling plans generalizing ChSP-1 have been the subject of extensive
work by H. F. Dodge and K. S. Stephens. These plans provide a generalization of ChSP-1
plans in that two stages for the implementation of the plan are defined.

1.
Restart procedure: The period during which the chain sampling procedure is started
or immediately following a rejection. During this phase, samples of n1 are chained
with a CRC of c1 allowable defectives in the cumulative results. When k1 lots have
been accepted, the normal procedure is instituted.
2.
Normal procedure: After k1 lots have been accepted, additional lots are chained
until a running total of k2 lots is reached and maintained. During this period,
samples of n2 are taken from each lot using a CRC of c2 allowable defectives
in the cumulative results. The restart procedure is initiated as soon as a lot is
rejected.

This approach as introduced by Dodge and Stephens (1966) can be represented schemati-
cally as in Figure 16.6.
The solution of the operating characteristic problem of the general family of chain sam-
pling plans is described by Stephens and Dodge (1974). It involves imbedding a Markov
chain in the chain sampling process by an appropriate definition of states. The two-stage
plans have been designated by Stephens and Dodge (1976b) as ChSP (n1, n2)-c1, c2 with
k1, k2 separately specified. The original ChSP-1 plan is equivalent to ChSP (n, n)-0, 1 with
i = k1 = k2 − 1. The first two-stage plans by Dodge and Stephens (1966) maintained a con-
stant sample size in both the restart and normal procedures. These plans will be found
designated ChSP-c1, c2 in the literature with n, k1, k2 separately specified. The advantages
of greater generality in the selection of chain sampling parameters are greater flexibility
in matching and use, and, of course, improved the discrimination through the use of the
generalized two-stage procedure.
Stephens and Dodge (1976a) have provided a comparison of ChSP-1 and two-stage
chain plans against single- and double-sampling plans. For example, they have found the
­following to be matched using k1 = 1, k2 = 2, n = n1 = n2 = 50. Double-sampling rejection
numbers are c2 + 1 on both samples.

Chain Sampling Single Sampling Double Sampling

ChSP-1 i = 1, n = 50 n = 85, c = 1 n1 = 40, n2 = 80, ASN = 53.2

c1 = 0, c2 = 1
ChSP-0, 2, n = 50 n = 105, c = 2 n1 = 72, n2 = 144, ASN = 79.2
c1 = 1, c2 = 3
ChSP-0, 3, n = 50 n = 120, c = 3 n1 = 54, n2 = 108, ASN = 79.6
c1 = 0, c2 = 4
ChSP-0, 4, n = 50 n = 135, c = 5 n1 = 69, n2 = 138, ASN = 75.5
c1 = 1, c2 = 6
458 Acceptance Sampling in Quality Control

Start

k=1
k = Number lots
cumulated
Sample n1
units from
each lot and
cumulate results
D = Cumulative number
defective
D ≤ C1 D > C1

k < k1 k = k1

Sample n2 from
each lot and
continue to
cumulate results

D ≤ C2 D > C2

k < k2 k = k2

Delete k2th
Preceding lot
from cumulative
results

FIGURE 16.6
Two-stage chain sampling procedure.

The values of ASN shown are for a proportion defective of .005, which had greater than
.92 probability of acceptance under the double-sampling plan. Since for fractions defective
greater than .005, the ASN of the double-sampling plans were much higher, the compari-
son seems favorable to the chain sampling plans shown.

Deferred Sentencing Schemes

Deferred sentencing schemes were among the earliest of cumulative results plans. They
trace their origins to the British Ministry of Supply when Spalding, Halliday, and Sealey
applied the method during World War II under the name of “rational sentencing.” The
term sentencing was regularly used by the government inspectorates in connection with
acceptance of ammunition. Deferred sentencing involves delay of disposition of question-
able lots until subsequent lots have been inspected.
Cumulative Results Plans 459

The scheme has been aptly described and evaluated by Anscombe et al. (1947), who pre-
sented several approaches to this type of sentencing. Their simplest scheme has been the
most fully investigated and is described by them (p. 199) in the following:

Sentencing rule. The product, as it leaves the line, is divided into small lots, and one item
is selected from each for test. D and n being given integers, whenever n defective items
are encountered out of D or fewer consecutive lots tested, all the lots consecutively from
that giving the first to that giving the nth defective in the cluster are rejected. Lots not
rejected by this rule are accepted.

This simple deferred sentencing scheme is represented diagrammatically in Figure 16.7.

The basic idea is to defer sentencing lots after a defect is found until it can be shown
that the subsequent D − 1 lots are of acceptable quality, in which they give less than n − 1
defectives. When this is shown, lots are released up to the next defective and the process
is repeated. This approach may be used as a continuous sampling plan on individual items,

Start

Sample one
item per lot

Nondefective Defective

Accept Hold
lot lot

Continue to hold subsequent

lots, sampling one item from
each, until a sequence of D
lot is accumulated or until
n defectives have been found

Determine cumulative number

n defectives defective, Σd, in the
found sequence of D lots

Σd = n Σd = 1 2 ≤ Σd ≤ n – 1

Reject all Accept all Accept first lot

lots being lots being in sequence of D
held held and all subsequent
lots having no
sample defectives
up to the next
lot having a
sample defective

FIGURE 16.7
Simple deferred sentencing scheme.
460 Acceptance Sampling in Quality Control

TABLE 16.5
Values of Dp Having Specified Probability of Acceptance
Percent Output Accepted
n 99% 90% 50% 10% 1%

3 0.35 0.89 2.20 4.5 7.3

4 0.63 1.34 2.84 5.3 8.1
5 0.97 1.84 3.54 6.1 9.0
6 1.36 2.39 4.28 7.0 10.0
7 1.78 2.96 5.04 7.9 11.0
10 3.25 4.85 7.43 10.8 14.2
Source: Reprinted from Anscombe, F.J. et al., J. R. Statist. Soc., 9, 200, 1947. With permission.

which does not require screening and so may be employed with destructive tests. Deferred
sentencing is primarily intended, however, for use with lots of product sufficiently small
that one test per lot is reasonable. Note that it is particularly well suited to bulk sampling
applications.
The selection of n and D is facilitated by a table of the percentage points of the product
Dp presented by Anscombe et al. (1947) shown here as Table 16.5.
Using Lagrangian interpolation to determine the p.95 values, the ORs R = p.10/p.95 for these
plans are approximately the following:

n R

3 7.5
4 5.7
5 4.4
6 3.8
7 3.4
10 2.7

A graphical representation of Table 16.5 has also been given by Anscombe et al. (1947),
which is shown in Figure 16.8. It gives curves for n plotted by Dp and the percentage of
product accepted. Note that the latter corresponds roughly to Pa and can be used to con-
struct an OC curve for these schemes.
To use the table, or the figure, n may be chosen to correspond with one of the ORs given.
If it is desired to hold a value of LTPD, say 100pt, the value of Dp corresponding to a pro-
portion of output accepted of 10% is divided by pt. This gives the value of D be used in
the plan. For example, if a deferred sentencing scheme is to be determined having about
the same protection as the plan n = 50, c = 5, which has an operating ratio R = 3.6 and an
LTPD = 18.6%, we have
n=7
Dp 7.9
D= = = 42.5 ~ 43
pt .186
The plan is implemented by taking one unit from each lot. As long as no defectives are
found, the lots are passed. As soon as a defective is obtained, lots are held. If six or more
Cumulative Results Plans 461

20.0

10.0

5.0
4.0
n = 10
3.0

2.0
Dp

n=7
n=6
1.0 n=5

n=4
0.5
0.4
n=3
0.3

0.2
99.9 99.8 99.5 99 98 95 90 80 70 60 50 40 30 20 10 5 2 1 0.5 0.2 0.1
Percentage of product accepted

FIGURE 16.8
Operating characteristics of simple deferred sentencing schemes. (Reprinted from Anscombe, F.J. et al., J. R.
Statist. Soc., 9, 201, 1947. With permission.)

defectives are found in the next 42 lots, all lots are rejected up to and including that provid-
ing the seventh defective. If five or less defectives are forthcoming, the first lot is passed
together with all subsequent lots up to the next lot showing a defective. From that point,
the procedure is applied again.
The OC curve for this scheme may be obtained by dividing the values of Dp given in
Table 16.5 for n = 7 by D = 43.

Deferred Sentencing Single Sampling

Pa p n = 50, c = 5

.99 .041 .036

.90 .069 .063
.50 .117 .113
.10 .184 .186
.01 .256 .262

The chart shown in Figure 16.8 is also very useful for this purpose.
There are many variations possible for the deferred sentencing scheme. Some are given
by Anscombe et al. (1947). In one procedure, the lots held for deferred sentencing may
extend from the lot in which a defect is found, forward, and back for a number of lots.
A double-sampling approach has been given by Hill et al. (1959), which incorporates into
the scheme samples of size greater than one from a lot. Deferred sentencing is attractive in
certain applications where time is not at a premium and lots can be put aside. However, it
suffers from the delay inherent in holding lots for disposition for any period.
462 Acceptance Sampling in Quality Control

Demerit Rating Plan

The check inspection and demerit rating plan used extensively by Western Electric and
described by Dodge and Torrey (1956) is an audit plan intended to characterize quality
levels and to provide a check inspection with relatively small samples. It supplies manage-
ment with a demerit rating of defects on specific products and a demerit index of quality
across defect and product types. At the same time, the inspection results necessary for sur-
veillance of quality are used in product acceptance. The plan provides continuing surveil-
lance through control charts. Dodge (1962) has recommended its use in conjunction with
the CRC plan discussed in the following text. Its successful application in this regard has
been described by Cone and Dodge (1963).
The demerit rating plan is initiated by taking small samples at regular intervals (by shift,
day, or week) across the product and product types to be included. Sample sizes are chosen
with regard to the homogeneity of the universe of product to be sampled and its complex-
ity. For a given quantity of output N to be represented by the sample, the sample size is
chosen as in Table 16.6. These samples sizes were arrived at empirically and represent the
result of more than 15 years of experience.
Recognizing differences in the variety and nature of defect types, a classification of
defects was developed as follows:

Class A (very serious) Will surely cause an operating failure

Class B (serious) Will probably cause an operating failure
Class C (moderately serious) May possibly cause an operating failure
Class D (not serious) Minor defect that will not affect operation, maintenance, or life

These classes are assigned demerits for use in constructing a demerit rating and demerit
index in the operation of the plan. These demerits are

Class A: 100 demerits

Class B: 50 demerits
Class C: 10 demerits
Class D: 1 demerit

Table 16.7 as given by Dodge and Torrey (1956) describes the classification further.

TABLE 16.6
Sample Size for Check Inspection and Demerit Rating
Universe
Heterogeneous Homogeneous
Product Comprising a Variety of Product of a Specific Type or a Group
Product Types of Different Construction of Types of Similar Construction

Complex: Construction subject n = 2.5 2N n = 1.5 2N

to variations in adjustment
Simple: Construction n = 2 2N n = 2N
nonadjustable or stable
Source: Reprinted from Dodge, H.F. and Torey, M.N., Ind. Qual. Control, 13(1), 7, 1956. With permission.
 Cumulative Results Plans

TABLE 16.7
Important Aspects of Classification of Defects
Appearance
Cause Cause Cause Intermittent Cause Involve Increased Cause Increase in Finish or
Defect Demerit Personal Operating Operating Trouble Substandard Maintenance or Installation Effort Workmanship
Class Weight Injury Failure Difficult to Locate in Field Performance Decreased Life by Customer Defects

A 100 Liable to Will surelya Will surely — — — —

B 50 — Will surelyb — Will surely Will surely Major increase —
Will probably
C 10 — May possibly — Likely to Likely to Minor increase Major
D 1 — Will not — Will not Will not — Minor
Source: Reprinted from Dodge, H.F. and Torrey, M.N., Ind. Qual. Control, 13(1), 8, 1956. With permission.
a Not readily corrected in the field.

b Readily corrected in the field.

463
464 Acceptance Sampling in Quality Control

TABLE 16.8
Nonconformance Criteria
Maximum No. of Defects in Sample
No. of Units in Sample, n Class A Class B Class C Class D

1–2 0 0 0 0
3–4 0 0 0 1
5–8 0 0 1 1
9–16 0 0 1 2
17–18 0 0 2 3
19–25 0 1 2 3
26–31 0 1 2 4
32–36 0 1 3 4
37–48 1 1 3 5
49–50 1 1 3 6
51–65 1 1 4 6
66–75 1 2 5 7
76–90 1 2 5 8
91–100 1 2 6 9
Over 100 a a a a

Source: Reprinted from Dodge, H.F. and Torrey, M.N., Ind. Qual. Control, 13(1), 9, 1956. With permission.
a
Class A = 0.0025n + 0.150 n .
Class B = 0.0050n + 0.212 n .
Class C = 0.0200n + 0.424 n .
Class D = 0.0400n + 0.600 n .

The sample size having been determined, samples are taken periodically over the time
for which the demerit rating is to be constructed. If the number of defects observed in any
sample exceeds the nonconformance criteria given in Table 16.8, a second sample twice
as large is taken. If the combined number of defectives in the first and second samples
exceeds the nonconformance criteria, the lot or batch represented by the samples is rejected
subject to action by the proper authority.
The nonconformance criteria are set at three standard deviations distant from noncon-
formance levels (NL) that roughly correspond to AQLs for the class of defects involved
in the sense that, according to Dodge and Torrey (1956), “…if products are maintained
at acceptable quality levels, the chances of the criteria being exceeded are very remote.”
These are the following:

Defect Class NL Proportion Defective

A .0025
B .005
C .02
D .04

Thus, the limits become

n ( NL ) ± 3 n ( NL )

using Poisson limits for the number of defects.

Cumulative Results Plans 465

After the samples for the period have been collected, a demerit rating is calculated as

D = w A dA + wBdB + wC dC + wD dD

where the weights wK are simply the demerits assigned to defects of class K and the num-
ber of defects found in that class is dK. Demerits per unit may be calculated as

D
U=
n

where n is the sample size collected for the period. These values are plotted on control
charts with limits set to reflect a “standard quality level” (SQL) that represents engineer-
ing estimates of what quality at delivery should be, taking into account considerations of
quality and cost. For defect classes A, B, C, and D, these are represented by μA, μB, μC, and
μD, each in terms of defects per unit. A given unit of product would then have a SQL US, in
defects per unit, determined as

US = w Am A + wBmB + wCmC + wDmD

This linear combination of Poisson variates has a standard variance factor

CS = w A2 m A + wB2mB + wC2 mC + wD2 mD

Hence, limits for a control chart showing the sample value of demerits per unit, D, plotted
for samples of n taken each period are

CS
US ± 3
n

When products or dissimilar product types are to be combined to give a quality index for
a line, a department, or plant, the types included should each be weighted to represent
the number produced or other salient considerations. In this case, the overall demerit
index I0 is

å wiU i /USi
I0 =
å wi

with

1 æ wi2 CSi ö
s I0 = S ç ÷
ni ç ( å wi ) (USi )2 ÷
2
è ø

where ni represents the sample size for the period for the ith type. The values I0 are plotted
on a control chart with limits

1 ± 3sI 0
466 Acceptance Sampling in Quality Control

X, Y, and Z type relays—unmounted monthly demerit rate

Type of Demerit No. Defects No. Demerits
Defect Weight June J-A Sept. June J-A Sept.
A 100 0 1 0 0 100 0
B 50 0 2 0 0 100 0
C 10 5 10 6 50 100 60
D 1 1 1 3 1 4 3
Total demerits 51 304 63
Sample size, n 232 240 165
Demerits
Demerits per units = = 0.22 1.27 0.38
No delays in sample

0
Dem, per unit

0.5 Standard
quality level = .52
1.0
1.5 Lower control
limit 0.52 + 3 25
2.0 n
J F M A M J J-A S O N D

X, Y, and Z type relays—unmounted quality standards

Contributions to Standards
Defects Demerits Defects Demerits (Demerits)2
Class per Deffect w per Unit µ per Unit per Unit
A 100 0.0014 0.140 14.00
B 50 0.0034 0.170 8.50
C 10 0.0205 0.205 2.05
D 1 0.0097 0.010 0.01
Total 0.525 24.56
Standards Us = 0.52 Cs = 25

Us = wAµA + wBµB + wCµC + wDµD, and

Cs = wA2µA + wB2µB + wC2µC + wD2µD
Rounded to two significant figures.

FIGURE 16.9
Example of summary of demerits per unit and control chart. (Reprinted from Dodge, H.F. and Torrey, M.N., Ind.
Qual. Control, 13(1), 10, 1956. With permission.)

since the expected demerits per unit of the index calculated in this way is 1. When all types
included in the index have the same standard value, wi = 1, and the index degenerates into
a simple “demerit index,” I.
An example of the calculation and display of demerits per unit as shown by Dodge and
Torrey (1956) is given in Figure 16.9.

CRC Plan
The necessity for small samples when tests are costly or difficult to administer often
reduces protection to the consumer. In cases of audit inspection and demerit rating, check
samples in production, small lots, and destructive tests; sample sizes of 5 or 10 or even less
Cumulative Results Plans 467

are common. Under such circumstances, a CRC can be used to increase the effectiveness
of the inspection in protecting the consumer. Cone and Dodge (1963) outlined such a plan
that has had successful application at the Sandia Corporation. The procedure as described
by Dodge (1962) is as follows:

1. For a given quality characteristic, choose a SQL that estimates what quality should
be at delivery considering costs and needs of service.
2. Choose a standard acceptance sampling plan for lot acceptance to be used regard-
less of whether the CRC plan is also applied.
3. When a lot fails the standard acceptance sampling plan, reject the lot and advise
the supplier that the CRC will apply to subsequent lots.
4. Cumulate the results of the standard acceptance sampling plan over subsequent
lots and compare the results for each lot to the CRC1. A stated moving cumulative
sample size m shall be maintained once attained.
5. If at any lot, the cumulative results fail to meet the CRC, the immediate lot is
rejected, and the process is also declared nonconforming.
6. Declaration of the process as nonconforming entails the following:
a. Ceasing inspection until the supplier submits written evidence that corrective
action has been taken.
b. Starting a new sequence of cumulative results when inspection is resumed.
c. If inspection is stopped a second time during the period in which the CRC is
in force, it is not resumed until evidence, satisfactory to higher authority, has
been furnished.
7. The CRC is continued until a succession of m units has been found to have results
equal to or better than the criterion for discontinuance CRC2. At that time, the
­supplier is notified that the CRC has been removed.

To determine the CRC, let

Y be the statistic generated by standard acceptance sampling plan (p, c, X, etc.)

Cum-Y be cumulative results moving average
CRC1 be cumulative results action criterion
CRC2 be cumulative results discontinuance criterion
YS be the standard quality level for Y
σYn be the standard error of Y for given YS and cumulative sample size n

A specific CRC plan is determined by three constants:

m is the maximum moving sample size

Z1 is the multiple of standard deviation for action on the CRC
Z2 is the multiple of standard deviation for discontinuance of the CRC

then

CRC1 = YS + Z1sYn
468 Acceptance Sampling in Quality Control

and

CRC 2 = YS + Z2sYn

Usually, Z1 is taken to be 1.65, 2, or 3, while Z2 may be 0 or 1. In application at Sandia, the

constants

Z1 = 3 Z2 = 0 m = 100

have been found very effective. Cone and Dodge (1963) describe the favorable experience
generated at Sandia over more than 2 years.
It should be emphasized that Y, the statistic cumulated, can take many forms. For
example,

Y = p (fraction defective)
Y = u (defects per unit)
Y = U (demerits per unit)
Y = X (sample mean)

In all cases, the sampling distributions involved will be those of a known universe with
parameters specified by the SQL employed. Thus, the procedure can be adapted to a wide
variety of sampling situations. Troxell (1972) has made an extensive investigation of types
of suspension systems for small sample inspections exemplified by the CRC plan, together
with applications of the procedure to MIL-STD-105E.

Software Applications
Excel
The Excel template SkSP-1 Sampling Plan.xlsx can be used to generate and evaluate a SkSP-1 skip-
lot sampling plan. To illustrate its use, suppose we desire an SkSP-1 plan based on Procedure
A-2 (each nonconforming lot is rejected and not replaced by a conforming lot) with i = 15 and
f = 1/2. Enter 15 into cell N3 for I and 2 into cell D3 so that the fraction inspected is 1/2 so 50
is entered into cell E3 to represent 50% (the actual fraction, f, appears in cell F3) as shown in
Figure 16.10. The AOQL for the plan is 1.78% (cell J3) and the LTPD or LQL is 17.83% at the 10%
probability of acceptance (cell J2). The OC, AOQ, and AFI curves are shown on separate tabs.
Another Excel template SkSP-2 Sampling Plan.xlsx can be used to generate and evaluate
an SkSP-2 skip-lot sampling plan. Using the earlier example, suppose the reference single-
sampling plan is n = 20 and c = 1 using i = 4 and f = 0.25. In the template, the reference plan
is entered into cells B1 (n) and B2 (c). The value of i goes into cell G3 and 20 is entered into
cell I3 to represent 20% (the actual fraction, f, appears in cell J3) as shown in Figure 16.11.
Assume an incoming proportion defective of p = .05. For the reference plan, at this fraction
defective, the probability of acceptance is Pa = .7358 (cell B507).
The measures of the skip-lot plan as shown previously are Pa ~ .860 (cell I507), F = .532
(cell J507), average sample number ASNSK = 10.6 (cell K507), unit AOQL (AOQL1) = 4.87%
(cell M507), and lot AOQL (AOQL2) = 12.17% (cell O507) as shown in Figure 16.11. The OC,
AOQ, and AFI curves are shown on separate tabs. The OC curve for this particular skip-lot
 Cumulative Results Plans

FIGURE 16.10
Excel template SkSP-1 Sampling Plan.xlsx sampling plan entry form.
469
 470

FIGURE 16.11
Excel template SkSP-2 Sampling Plan.xlsx sampling plan entry form.
Acceptance Sampling in Quality Control
 Cumulative Results Plans 471

Operating characteristic (OC) curve

comparison of SkSP-2 plan to reference single sampling plan
100%
Pa(SkSP-1) for i = 4, f = 1/4
90% Pa(Refplan: n = 20, c = 1)
Percent of total lots accepted on sampling basis, Pa

80%

70%

60%

50%

40%

30%

20%

10%

0%
0.00 0.05 0.10 0.15 0.20 0.25
Product quality in percent nonconforming, p

FIGURE 16.12
OC curve for SkSP-2 sampling plan and reference single-sampling plan.

plan is shown in Figure 16.12, which agrees with the reference plan and i = 4 curves shown
in Figure 16.2.
The Excel template Chain Sampling Plan.xlsx is a table for constructing ChSP-1 plans
based on calculations given by Soundararajan (1978a). Based on Poisson unity values,
it allows determination of a ChSP-1 plan from the desired operating ratio p2/p1. Values of
np are given at .95 and .10 probability of acceptance for i = 1 to 10.
For example, we find for the plan n = 10 (cell B11) with p1 = AQL = .0139 (cell B5) and
p2 = LTPD = .2302 (cell B6) that i = 3 (B13) as shown in Figure 16.13. As discussed earlier, the
following properties are given by Appendix Table T16.3 and confirmed by the template.

Pa = .95 at p = .0139 ( cell G144 )

Pa = .10 at p = .2303 ( cell B6 )

p2/p1 = .2303/.0139 = 16.568 ( cell B7 )

AOQL = 2.798 ( .0139 ) = .0389 ( template : .0372, cell B8 )

pM = .0902 ( template : .0824, cell B9 )

It is sometimes desirable to construct a ChSP-1 plan having a specified AOQL. Using

Table 16.4, developed by Soundararajan (1978a), values of sample size n and CRC i indicate
that an AOQL of .04 is achieved by the plan n = 10, i = 3 (as also determined by the template).
 472

FIGURE 16.13
Excel template Chain Sampling Plan.xlsx sampling plan entry form.
Acceptance Sampling in Quality Control
 Cumulative Results Plans 473

Operating characteristic curve

Lot release single sampling plans n, verying c
100%

90%
Percent of lots expected to be accepted

80%
P0, n = Prob(no defects in lot
70%
of quality p) or n, c = 0 plan
60% P0 or 1, n = Prob(one or less
defects in lot of quality p) or
50% n, c = 1 plan
P0, 1 or 2, n = Prob(two or
40%
less defects in lot of quality p)
or n, c = 2 plan
30%
P0, 1, 2 or 3, n = Prob(three
20% or less defects in lot of quality p)
or n, c = 3 plan
10%

0%
0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
Product quality in percent nonconforming

FIGURE 16.14
OC curves for c = 0, 1, 2, and 3 single-sampling plans for a sample of size n.

Operating characteristic curve

Lot release
Single vs. chain sampling plans n, varying c, varying i
100%
P0, n = Prob(no defects in lot
90%
of quality p) or n, c = 0 plan
Percent of lots expected to be accepted

80% Pacc=P0, n + P1,n (P0,n)^1 or n,

c = 0 ChSP (i = 1) plan
70%
Pacc=P0, n + P1,n (P0,n)^2 or n,
c = 0 ChSP (i = 2) plan
60%
Pacc=P0, n + P1,n (P0,n)^3 or n,
50% c = 0 ChSP (i = 3) plan

40% Pacc=P0, n + P1,n (P0,n)^4 or n,

c = 0 ChSP (i = 4) plan
30% Pacc=P0, n + P1,n (P0,n)^5 or n,
c = 0 ChSP (i = 5) plan
20% P0 or 1, n = Prob(one or less
defects in of quality p) or n,
10%
c = 1 plan
0%
0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
Product quality in percent nonconforming

FIGURE 16.15
OC curves for c = 0 and 1 single-sampling plans for a sample of size n compared to the chain sampling plans with
i = 1 to 5.
474 Acceptance Sampling in Quality Control

Average outgoing quality curve

Chain sampling plans
n, c, optimal i
4.00%
AOQ AOQL (ChSP-1) = 0.0372
3.50%
Percent of lots expected to be accepted

3.00%

2.50%

2.00%

1.50%

1.00%

0.50%

0.00%
0.0% 5.0% 10.0% 15.0% 20.0% 25.0%
Product quality in percent nonconforming

FIGURE 16.16
AOQ curve for the chain sampling plan with n, c = 0 and optimal value of i.

The template also provides two OC curves (on separate tabs). The first OC curve compares
the n, c = 0 plan with the n, c = 1; n, c = 2; and the n, c = 3 sampling plans (Figure 16.14).
The second OC curve compares the n, c = 0 and n, c = 1 plans with the n, c = 0 (i = 1 to 5)
ChSP-1 plans. This OC curve illustrates the usefulness of an n, c = 0, i = 3 ChSP-1 plan,
which is a fair compromise among other i values for ChSP-1 plans for a given value of n
(Figure 16.15). Finally, an AOQ curve for the ChSP-1 plan for n and c with optimal i is pro-
vided on a separate tab (Figure 16.16).

Problems
1. The plan n = 50 and c = 3 is being used in lot-by-lot inspection. Derive an SkSP-2 plan
that will afford the same protection.
2. The SkSP-2 plan n = 165, c = 3, i = 4, and f = .5 is being used in sampling inspection
of a continuing series of lots. The reference plan has p.95 = .008. For this plan, evaluate
the following when the process level is p = .008:
a. Probability of acceptance
b. Average fraction lots inspected
c. Average sample number
d. Unit AOQL
e. Lot AOQL
Cumulative Results Plans 475

3. Find a single-sampling plan that roughly matches the SkSP-2 plan n = 143, c = 6,
i = 12, and f = .5. Also find matching (a) double- and (b) multiple-sampling plans.
Compare the ASN of the single, double, and multiple plans to that of the skip-lot plan
for p.95. (Hint: Use the Dodge–Perry and the Schilling–Johnson tables.)
4. Draw the OC curves for the ChSP-1 plan i = 3 and n = 20. What is its AOQL? Evaluate
the formula for Pa when p = .10.
5. Find a ChSP-1 plan having an AOQL of 6% where, for administrative purposes,
i should be no greater than 2.
6. The MIL-STD-105E system for Code C, 2.5 AQL has an overall operating ratio
R = 20.14 with an LTPD = 28.8%. Find a ChSP-1 plan that will give this protection.
What is p.95 for this plan? What is its AOQL and at what process average does it
occur?
7. Find a deferred sentencing plan matching the single-sampling plan n = 50, c = 3. The
OC curves should match as closely as possible at the LTPD. What is the indifference
quality for the plan?
8. A certain simple component used in one specific product is made at the rate of 10,000
units per month. What should be the sample size per month to be used in a demerit
rating plan?
9. Using the demerit weight given by Dodge and Torrey together with the NL of the
defect classes, in terms of defects per hundred units, compute the SQL US and stan-
dard variance factor CS. What would be the control limits on a chart for n = 1000?
Would a signal result if, for classes A, B, C, and D, there were found 0, 2, 1, and 4
defectives, respectively, in a sample of 1000?
10. A CRC for p is set up on the attributes inspection plan n = 10, c = 1, where Z1 = 3,
Z2 = 0, and m = 100. The SQL is YS = .02. After 10 lots have been inspected under the
criterion, one defect has been found. What action should be taken?

References
Anscombe, F. J., H. J. Godwin, and R. L. Plackett, 1947, Methods of deferred sentencing in testing the
fraction defective of a continuous output, Supplement to the Journal of the Royal Statistical Society,
9(2): 198–217.
Cone, A. F. and H. F. Dodge, 1963, A cumulative results plan for small sample inspection, in American
Society for Quality Control Technical Conference Transactions, Chicago, IL, pp. 21–30. [Also
published in Industrial Quality Control, 21(1): 4–9.]
Derman, C., S. Littauer, and H. Solomon, 1957, Tightened multi-level continuous sampling inspec-
tion plans, Industrial Quality Control, 7(5), 7–12.
Dodge, H. F., 1955a, Chain sampling inspection plan, Industrial Quality Control, 11(4): 10–13.
Dodge, H. F., 1955b, Skip-lot sampling plan, Industrial Quality Control, 11(5): 3–5.
Dodge, H. F., 1962, A cumulative-results sampling plan for small sample inspection, Technical report
no. 11, Rutgers—The State University, New Brunswick, NJ.
Dodge, H. F. and R. L. Perry, 1971, A system of skip-lot plans for lot by lot inspection, in American
Society for Quality Control Technical Conference Transactions, Chicago, IL, pp. 469–477.
Dodge, H. F. and K. S. Stephens, 1966, Some new chain sampling inspection plans, Industrial Quality
Control, 23(2): 61–67.
476 Acceptance Sampling in Quality Control

Dodge, H. F. and M. N. Torrey, 1956, A check inspection and demerit rating plan, Industrial Quality
Control, 13(1): 5–12.
Guthrie, D. and M. Johns, 1958, Alternative sequences of sampling rates for tightened multi-level
continuous sampling plans, Technical report no. 37, Applied Mathematics and Statistics
Laboratory, Stanford University, Stanford, CA.
Hill, I. D., G. Horsnell, and B. T. Warner, 1959, Deferred sentencing schemes, Applied Statistics, 8(2):
76–91.
Lieberman, G. J. and H. Solomon, 1955, Multi-level continuous sampling plans, Annals of Mathematical
Statistics, 26(4): 686–704.
Perry, R. L., 1970, A system of skip-lot sampling plans for lot inspection, PhD dissertation, Rutgers—
The State University, New Brunswick, NJ.
Perry, R. L., 1973a, Skip-lot sampling plans, Journal of Quality Technology, 5(3): 123–130.
Perry, R. L., 1973b, Two-level skip-lot sampling plans—Operating characteristic properties, Journal of
Quality Technology, 5(40): 160–166.
Soundararajan, V., 1978a, Procedures and tables for construction and selection of chain sampling
plans (ChSP-1)—Part 1, Journal of Quality Technology, 10(2): 56–60.
Soundararajan, V., 1978b, Procedures and tables for construction and selection of chain sampling
plans (ChSP-1)—Part 2, Journal of Quality Technology, 10(3): 99–103.
Stephens, K. S. and H.F. Dodge, 1974, An application of Markov chains for the evaluation of the oper-
ating characteristics of chain sampling inspection plans, IAQR Journal, 1(3): 131–138.
Stephens, K. S. and H. F. Dodge, 1976a, Comparison of chain sampling plans with single and double
sampling plans, Journal of Quality Technology, 8(1): 24–33.
Stephens, K. S. and H. F. Dodge, 1976b, Two-stage chain sampling inspection plans with different
sample sizes in the two states, Journal of Quality Technology, 8(4): 207–224.
Troxell, J. R., 1972, An investigation of suspension systems for small sample inspections, PhD
­dissertation, Rutgers—The State University, New Brunswick, NJ.
17
Compliance Sampling

Consumer protection has always been a prime factor in the construction of industrial
acceptance sampling plans. The methods and procedures presented in this chapter attest
to that fact. A typical example is the set of Dodge–Romig lot tolerance percent defective
(LTPD) plans developed as early as 1929. Increased use of acceptance sampling plans in
connection with compliance testing to government standards, validation testing of sup-
plier’s inspection, and verification of extremely tight standards set by regulatory agencies,
original equipment manufacturers, and consumers of all kinds suggests the need for sam-
pling plans especially designed and adapted for this area of application.
The popularity of c = 0 attributes sampling plans is due to the importance of con-
sumer protection to automotive, pharmaceutical, and other companies who are sensitive
to the threat of litigation from customers who are harmed by nonconforming products.
Furthermore, customer satisfaction is paramount to the retention of market share in the
global economy, so compliance sampling is a vital part of acceptance control for companies
who want to remain competitive.
It has been pointed out by M.G. Natrella in Muehlhause et al. (1975) that “there is a
need to demonstrate the effectiveness of sampling schemes for compliance testing. Such
experience, and related mathematical investigations, is needed for the formulation in gen-
eral terms of the overall objectives of sampling schemes, so that the statistician and the
regulator—given the standard—can make and explain an appropriate selection.” In the
area of compliance testing, and especially for safety-related items, the following features
seem desirable in a sampling plan:

1. Rejection of the lot if any defective items are found in the sample
2. A well-defined relationship between the sampling plan and the size of the lot
being inspected
3. A clear indication of the economic impact of the quality levels utilized in the plan
4. Simplicity and clarity in use

In safety and compliance testing, an acceptance number of zero is particularly desirable,

since, to the uninformed, it would appear that the use of any greater acceptance num-
ber implies passing lots that have been shown to have defectives in them. Duncan (1979)
points out that in 1972, the National Highway Traffic Safety Administration proposed to
change the rule regulating the performance of hazard warning flashers to eliminate an
acceptance number of c = 3, quoting Duncan (1979, p. 21):

…It took the point of view that “permissible failure rates raise difficult problems of
interpretation and enforcement.” It was thus indicated that any sample the agency took
would have to be 100% conformance for the lot to pass. In other words, the acceptance
number would be zero.

477
478 Acceptance Sampling in Quality Control

The lot sensitive sampling plan (LSP) and tightened-normal-tightened (TNT) plans pre-
sented here are illustrative of plans that are particularly appropriate for use in compliance
sampling as well as in other areas of acceptance control. Plans for verification of quality
levels should be capable of demonstrating compliance to stated levels in as economic a
manner as possible. The simplified grand lot scheme is particularly useful in this regard
since it can be used to provide consumer protection at very-low-quality levels while main-
taining reasonable protection for the producer.
Further development in acceptance sampling plans for compliance testing has occurred
in the area of quick switching system (QSS) plans that had its genesis in the original work
by Dodge (1967). Romboski (1969) extended Dodge’s work on the original QSS (n; cN; cT)
plans to extend them to QSS (n; kn; 0), QSS (n; 2n; 0), QSS-2, and QSS-3 plans. This work on
zero acceptance number quick switching systems for compliance sampling (QSS(n; kn; 0))
was further developed by statisticians in India and New Zealand. In 2011, work on QSS
plans was coupled with chain sampling culminating in a zero acceptance number chained
quick switching system (ChQSS).

LSP
The LSP developed by Schilling (1978) is applicable in general acceptance sampling and
is particularly useful in compliance and safety-related testing. A consumer-oriented LTPD
plan is intended to meet the objectives outlined for compliance testing. Based on the hyper-
geometric probability distribution, it gives the proportion of the lot that must be sampled
to guarantee that the fraction defective in the lot is less than a prescribed limit with LTPD
protection.
The LSP plan is easy to use and is based on the concept of acceptance with zero defectives
in the sample. It relates the sample size to lot size in a straightforward way and provides,
as a baseline, a minimum sample size for sampling applications, since single-sampling
plans allowing acceptance with one or more defectives in the sample usually require larger
sample sizes. The economic impact of the plan vis-à-vis 100% inspection is shown by the
fraction of the lot to be inspected.
The disadvantages of plans allowing no defectives in the sample is, in an economic
sense, in terms of good product rejected because of the severity of the acceptance criteria.
Where possible, various acceptance sampling schemes and strategies should be considered
as an alternative to plans of this type. However, if it is required that no defectives are to be
allowed in the sample, the LSP plan has real advantages, particularly if the inspection is to
be carried out on a unique lot.

Procedure
An LSP plan may be derived in the following manner:

1. Specify lot size N.

2. Specify the limiting quality level pt that is to be protected against by the plan.
3. Compute the product D = Npt.
Compliance Sampling 479

4. Enter the body of Table 17.1 at the nearest value of D and read the corresponding
value of f as the sum of the associated row and column headings.

f = fraction of lot inspected

5. The sampling plan is

Sample size = n = f N

Acceptance number = c = 0

Sample size is always rounded up.

The plan is applied as follows:

1. Randomly sample n items for a lot of N items (i.e., sample a fraction f of the lot).
2. Reject if any defective units are found in the sample.

Protection
The use of the LSP plan as outlined provides LTPD protection to the consumer at the limit-
ing fraction defective pt specified. Specification of LTPD protection is equivalent to a reli-
ability confidence coefficient of 90%. In other words, we can be 90% confident that a lot that
has passed the plan has a fraction defective less than the value of pt specified (or, equiva-
lently, that it has a reliability of at least (1 − pt)). This statement is made in the sense that
in repeated applications of the plan, lots that are composed of exactly pt fraction defective
would be rejected 90% of the time.
To portray the probability of acceptance of the plan, it is possible to approximate the
Type B operating characteristic (OC) curve of the plan showing probability of acceptance
Pa plotted against possible fractions defective p that could occur in the manufacturing pro-
cess from which the lot was taken. This may be done using the factors given in Table 17.2,
which, when multiplied by the selected value of pt, will give the approximate fractions
defective associated with various probabilities of acceptance.
Table 17.2 may also be used to approximate the well-known quantities descriptive of the
protection afforded by the plan such as indifference quality, limiting quality, acceptance
quality level (AQL) (defined as having 95% probability of acceptance), and so on, since
these quantities are determined by probability of acceptance. Furthermore, it provides the
factors necessary to allow the derivation of plans having probability of acceptance other
than 10% at the specified fraction defective.
Suppose a plan was desired having approximately 5% probability of acceptance for
a specified fraction defective p*, that is, a plan that would assure passing lots had at least
1 − p* reliability with 95% confidence. Table 17.2 can be used to obtain such a plan as follows:

1. Since Table 17.2 shows p = 1.301pt at Pa = .05, set pt = 1.301pt and solve for pt

p*
pt =
1.301
 480

TABLE 17.1
Values of D = Npt Corresponding to f
f .00 .01 .02 .03 .04 .05 .06 .07 .08 .09

.9 1.0000 0.9562 0.9117 0.8659 0.8184 0.7686 0.7153 0.6567 0.5886 0.5000
.8 1.4307 1.3865 1.3428 1.2995 1.2565 1.2137 1.1711 1.1286 1.0860 1.0432
.7 1.9125 1.8601 1.8088 1.7586 1.7093 1.6610 1.6135 1.5667 1.5207 1.4754
.6 2.5129 2.4454 2.3797 2.3159 2.2538 2.1933 2.1344 2.0769 2.0208 1.9660
.5 3.3219 3.2278 3.1372 3.0497 2.9652 2.8836 2.8047 2.7283 2.6543 2.5825
.4 4.5076 4.3640 4.2270 4.0963 3.9712 3.8515 3.7368 3.6268 3.5212 3.4196
.3 6.4557 6.2054 5.9705 5.7496 5.5415 5.3451 5.1594 4.9836 4.8168 4.6583
.2 10.3189 9.7682 9.2674 8.8099 8.3902 8.0039 7.6471 7.3165 7.0093 6.7231
.1 21.8543 19.7589 18.0124 16.5342 15.2668 14.1681 13.2064 12.3576 11.6028 10.9272
.0 a 229.1053 113.9741 75.5957 56.4055 44.8906 37.2133 31.7289 27.6150 24.4149
Source: Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 119, 1979. With permission.
a For values of f < .01 use f = 2.303/D; for infinite lot size use sample size n = 2.33/p .
t
Acceptance Sampling in Quality Control
Compliance Sampling 481

TABLE 17.2
Factors for Constructing the OC Curve
Pa p Pa p Pa p Pa p

0.999 0.00043pt
0.995 0.00218pt 0.900 0.046pt 0.100 1.000pt 0.005 2.300pt
0.990 0.0044pt 0.750 0.125pt 0.050 1.301pt 0.001 2.996pt
0.975 0.0110pt 0.500 0.301pt 0.025 1.602pt
0.950 0.0223pt 0.250 0.602pt 0.010 2.000pt

Source: Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 119, 1979. With permission.

2. Use the value of pt obtained to set up a sampling plan using the standard LSP
procedure.
3. Resulting plan will have approximately Pa = .05 for fraction defective p*.

It should be noted that, for a stream of successive lots, the average outgoing quality limit
(AOQL) can be approximated for LSP plans as

.3679 æ 1 ö
AOQL = ç - 1÷
N è f ø

Producer’s Risk
Since acceptance is allowed only when no defectives are found in the sample (c = 0), the
producer must produce at a fraction defective that is less than about 5% of the level, pt,
protected against by the plan in order to assure a reasonably small probability (about 1 in
10 odds) of good lots being rejected.
Clearly, a perfect lot has 100% probability of acceptance under the LSP plan since no
defectives can be found in the sample. For such lots, the producer’s risk of rejection is 0.
Duncan (1977) has shown that for lots containing only a single defective unit (i.e., lots of
fraction defective 1/N) the probability of acceptance is just

Pa = 1 - f

The corresponding producer’s risk of such a lot being rejected is

1 - Pa = f

Thus, with a fraction of the lot inspected of f = .21 and lot size of 100, as in Example 1, there
is a probability of acceptance of

Pa = 1 - .21 = .79

for lots containing a fraction defective

1
p= = .01
100
482 Acceptance Sampling in Quality Control

and a corresponding producer’s risk at that level of fraction defective of

1 - Pa = .21

This gives a minimum estimate of the producer’s risk since a lot containing more than
one defective unit would have a higher probability of rejection. Duncan (1977) has indi-
cated that “computations of producer’s risk…reveal that…plans with zero acceptance
numbers…can be hard on the producer unless most of his lots are perfect.” It is important
to remember that single-sampling plans that require no defectives in the sample for lot
acceptance (such as LSP) should be used only when the state of the art permits near perfect
quality levels to be economically produced.

Examples of LSP Applications

The following are examples of applications of the LSP plan.

Example 17.1
A part is received at incoming inspection in lots of 100 items. Protection against a fraction
defective of 10% is desired. The LSP plan is derived as follows:

1. N = 100
2. pt = .10
3. D = Npt = 100(.10) = 10
4. Table 17.1 gives f = .21 closest to D = 10
5. The sampling plan is

n = .21(100 ) = 21, c = 0

The plan is implemented in the following way:

1. Randomly sample 21 items from each lot of 100.

2. Reject the lot if any defectives are found; accept otherwise.

If rejected material is 100% inspected with rejected items replaced by good ones, the
AOQL is estimated as

.3679 æ 1 ö
AOQL = - 1 ÷ = .014
100 èç .21 ø

Also, from Table 17.2, it is possible to approximate other characteristics of the sampling
plan, such as

1. Indifference quality (fraction defective having 50/50 chance of lot acceptance)

IQ = .301pt = .301(.10 ) = .0301

2. Limiting quality having probability of acceptance of 5% (fraction defective hav-

ing 5% probability of acceptance).

LQ = 1.301pt = 1.301(.10 ) = .1301

Compliance Sampling 483

3. AQL (defined as fraction defective having 95% probability of acceptance).

AQL = .0223 pt = .0223 ( .10 ) = .00223

This example illustrates the simplicity of calculation in deriving an LSP plan.

Example 17.2
In bidding on a new contract, it is necessary to evaluate the consequences of quality
requirements of 1% probability of acceptance at a fraction defective of 2% for products
produced in lots of 100. Thus, p* = .02, and

p* = 2.0pt (from Table 17.2)

p * .02
pt = = = .01
2.0 2.0

The resulting sampling plan is derived as follows:

1. N = 100.
2. pt = .01.
3. D = Npt = 100(.01) = 1.0.
4. Table 17.1 gives f = .90 closest to D = 1.0.
5. The sampling plan is

n = .90 (100 ) = 90, c = 0

This plan requires inspection of 90% of every lot, which may or may not be economically
feasible. If this is the case, 100% inspection may be the only practical alternative. The LSP
plan thus makes explicit the economic consequences of sampling in terms of the fraction
of each lot to be inspected.

Example 17.3
A lot of 10,000 items has been set aside for 100% inspection. It is uneconomical to inspect
the lot if the fraction defective is 7% or more. Derive an LSP plan to test if 100% inspec-
tion is practical.

1. N = 10,000.
2. pt = .07.
3. D = Npt = 10,000(.07) = 700.
4. Since f < .01, use

2.303 2.303
f = = = .0033
D 700

5. The sampling plan is

n = f N = .0033 (10,000 ) = 33, c = 0

484 Acceptance Sampling in Quality Control

Further Considerations
As shown by Schilling (1978), derivation of the LSP plans is based on the notion that, for
the hypergeometric distribution, when c = 0 with D = Np defective pieces in the lot, the
probability of acceptance becomes
Np
æ nö
Pa £ ç 1 - ÷
è N ø

Note that this is equivalent to the f-binomial approximation for the hypergeometric when
c = 0. Values of D for Table 17.1 are obtained as

log Pa
D = NpT =
log ( 1 - f )

Since

np
D=
f

when the Poisson approximation applies

2.303
f =
D

for a probability of acceptance of .10.

A useful set of curves for quick and easy assessment of LSP sample sizes has been given
by Hawkes (1979), which includes levels of consumer protection of .01, .05, .10, .20, and .50.
When a sample is required that is less than 10% of the lot size (D > 22) and, of course, for
conceptually infinite populations, an excellent nomograph for use when c = 0, based on the
Poisson distribution, has been prepared by Nelson (1978). This may be employed to obtain
confidence limits on reliability from the results of sampling inspection and to determine
the sample size, which will give the required protection.

TNT Scheme
While LSP plans are intended for application to unique lots, when product is forthcoming in
a stream of lots and a zero acceptance number is to be maintained, the TNT scheme devised
by Calvin (1977) is particularly appropriate. This scheme utilizes two c = 0 ­sampling plans
of different sample sizes together with switching rules to build up the shoulder of the OC
curve after the manner of the switching rules of MIL-STD-105E (United States Department
of Defense 1989). This is done by a change in sample size rather than acceptance number.
Calvin (1977) points out that, while increasing producer protection, the switching rules
have no real effect on LTPD, which remains essentially that of the tightened plan. Similar
results were shown by Schilling and Sheesley (1978) for MIL-STD-105D even when switch-
ing to reduced inspection was added. The procedure is as follows.
Compliance Sampling 485

Procedure
A TNT scheme is specified by

n1 = tightened (larger) sample size

n2 = normal (smaller) sample size
t = criterion for switching to normal inspection
s = criterion for switching to tightened inspection

It is carried out as follows starting with tightened inspection:

1. Inspect using tightened inspection with the larger sample size n1, c = 0.
2. Switch to normal inspection when t lots in a row are accepted under tightened
inspection.
3. Inspect using normal inspection with the smaller sample size n2, c = 0.
4. Switch to tightened inspection after a rejection if an additional lot is rejected in the
next s lots.

A diagrammatic representation of the switching rules for the TNT scheme is shown in
Figure 17.1.

Protection
The TNT plans correspond to the MIL-STD-105E normal-tightened plans when the ­switching
criteria are set at t = 5, s = 4. In fact, TNT plans correspond directly to the MIL-STD-105E
scheme (using normal-tightened switching only) when the normal plan has a zero accep-
tance number. For example, Code F, 0.65% AQL gives

Normal: n = 20, c = 0

Tightened: n = 32, c = 0

2 out of
s + 1 consecutive
lots rejected

Normal Tightened Start

t consecutive
lots accepted

FIGURE 17.1
Switching rules for TNT procedure.
486 Acceptance Sampling in Quality Control

which correspond to the TNT plan with t = 5, s = 4, n1 = 32, n2 = 20. Calvin (1977) shows the
scheme probability of acceptance of the TNT plan to be

Pa =
( )( )
P1 1 - P2 s 1 - P1t ( 1 - P2 ) + P2P1t ( 1 - P1 ) 2 - P2 s ( )
( 1 - P )(
2 1s
- P ) (1 - P ) + P
1
t
2 1
t
(
(1 - P1 ) 2 - P2s )

where
P1 = (1 − p)n1 is the probability of acceptance of tightened plan at fraction defective p
P2 = (1 − p)n2 is the probability of acceptance of normal plan at fraction defective p

The average sample number is

ASN = n =
( )( )
n1 1 - P2 s 1 - P1t ( 1 - P2 ) + n2P1t ( 1 - P1 ) 2 - P2 s ( )
(1 - P ) (1 - P ) (1 - P ) + P
2
s
1
t
2 1
t
(1 - P1 ) ( 2 - P2 s
)

with average outgoing quality (AOQ) at fraction defective p:

æ N-n ö
AOQ = pPa ç ÷ defective replaced
è N ø

and

pPa ( N - n )
AOQ = defective not replaced
N - np - p ( 1 - Pa ) ( N - n )

The improvement in the operating ratio of a TNT plan over that of its tightened component
(n1 = n2 = 20) is shown in Figure 17.2.

100 n1 = 20, n2 = 5

80 n1 = 20, n2 = 10

60
Pa

40

n1 n2 = 20
20

0
0 2 4 6 8 10 12 14
p

FIGURE 17.2
OC curves of TNT plans (t = 5, s = 4). (Reproduced from Calvin, T.W., TNT zero acceptance number sampling,
in American Society for Quality Control Thirty-First Annual Technical Conference Transactions, American Society for
Quality Control, Philadelphia, PA, 1977, p. 37. With permission.)
Compliance Sampling 487

Selection
Schilling and Sheesley (1978) have pointed out that “…the scheme OC curve might be
approximated by the normal OC curve for Pa ≥ 90% and the tightened OC curve for
Pa ≤ 50% with the intermediate region appropriately interpolated.” The suggestion is from
their work on MIL-STD-105D that includes switching to reduced inspection and so this
approximation should apply even better to the TNT plans. As an illustration, for the plan
n1 = 20, n2 = 5, t = 5, s = 4, shown in Figure 17.2, for p = .01

Pa =
(.8179 (1 - .9510 ) (1 - .8179 ) (1 - .9510 ) + .9510 (.8179 ) (1 - .8179) ( 2 - .9510 ))
4 5 5 4

((1 - .9510 ) (1 - .8179 ) (1 - .9510 ) + (.8179 ) (1 - .8179) ( 2 - .9510 ))

4 5 5 4

= .943

and for p = .11

Pa =
(.0972 (1 - .5584 ) (1 - .0972 ) (1 - .5584 ) + .5584 (.0972 ) (1 - .0972) ( 2 - .5584 ))
4 5 5 4

((1 - .5584 ) (1 - .0972 ) (1 - .5584 ) + (.0972 ) (1 - .0972) ( 2 - .5584 ))

4 5 5 4

= .097

For n = 5, c = 0 we have p.95 = .01 while for n = 20, c = 0, p.10 = .11. Thus, it would seem that,
at least when using the analogous MIL-STD-105E switching criterion, a TNT plan can be
derived using n1 from the c = 0 plan having the desired LTPD and n2 from a c = 0 plan hav-
ing a producer’s quality level at Pa = .95 equal to a specified value.
An easy way to find this value is to divide the c = 0 unity values for 10% and 95% prob-
ability of acceptance by the desired p.10 and p.95, respectively. Thus, the factors to obtain n1
and n2 are

n1 p.10 = 2.303

n2 p.95 = .0513

Further, the ratio of sample sizes will be

n1 44.89
=
n2 R

where R is the desired operating ratio. For example, to obtain a TNT plan having p.10 = .11
and p.95 = .005, we have
2.303
n1 = = 20.9
.11
.0513
n2 = = 10.3
.005

and, for the desired R = 22,

n1 44.89
= = 2.04
n2 22
488 Acceptance Sampling in Quality Control

Thus, the plans n1 = 20, c = 0 having p = .109 at Pa = .10 and n2 = 10, c = 0 having p = .005 at
Pa = .95 would appear to suffice to give TNT: t = 5, s = 4, n1 = 20, n2 = 10. And they do as part
of the Calvin (1977) tabulation. Thus, the approach, as suggested by Schilling and Sheesley,
can be used to quickly set up two-point TNT plans.
To find a TNT plan to match the plan n = 20, c = 1, which has p.95 = .018 and p.10 = .18, it is
necessary to find c = 0 plans that have these probability points. Using the binomial distri-
bution, they have sample sizes 3 and 12, respectively. Hence, the plan is TNT: t = 5, s = 4,
n1 = 12, n2 = 3. For this plan applied to lots of size N = 100 at p = .018:

P =
(.8042 (1 - .9470 ) (1 - .8042 ) (1 - .9470 ) + .9470 (.8042 ) (1 - .8042) ( 2 - .9470 ))
4 5 5 4

((1 - .9470 ) (1 - .8042 ) (1 - .9470 ) + (.8042 ) (1 - .8042) ( 2 - .9470 ))

a
4 5 5 4

.8042 ( .006884 ) + .9470 ( .07875 )

= = .936
(.006884 ) + (.07875 )
12 ( .006884 ) + 3 ( .07875 )
ASN = = 3.72
.085634

æ 100 - 3.72 ö
AOQ = .018 ( .936 ) ç ÷ = .016 with replacement
è 100 ø
.0118 ( .936 ) ( 100 - 3.72 )
AOQ =
100 - 3.72 ( .018 ) - .018 ( 1 - .936 ) ( 100 - 3.72 )
= .016 without replacement

Soundararajan and Vijayaraghavan (1990) have provided a table of unity values for TNT
plans. The unity values are used in the manner of single, double, and multiple plans given
earlier. Their table appears here as Appendix Table T17.1 and covers the specific case in
which the tightened sample size is twice that of the normal sample size. It is for the special
case of k = 2. As an example of its use, suppose it is desired to design a TNT plan having
p1 = .005 and p2 = .11736 for an operating ratio of R = p2/p1 = .11736/.005 = 23.472.

Step Example

1. Select p1 and p2. p1 = .005, p2 = .11736.

2. Determine the operating ratio R = p2/p1. R = p2/p1 = .11736/.005 = 23.472.
3. Find R in Appendix Table T17.3. See table.
4. Obtain corresponding parameters s, t, n2p1 from row of R. s = 4, t = 5, n2p1 = .04905.
5. Calculate n2 = n2p1/p1. n2 = n2p1/p1 = .04905/.005 = 9.81 ~ 10.
6. Determine n1 = kn2 = 2n2. n1 = 2n2 = 2(10) = 20.

So the plan is s = 4, t = 5, n2 = 10, n1 = 20, and c = 0. As in the use of unity values discussed
earlier, there are other parameters, which can be calculated. For example, the table gives
n2(AOQL) = .18707. Hence, the AOQL, for example, is

n2 ( AOQL ) .18707
AOQL = = = .018
n2 10
Compliance Sampling 489

Other measures are handled in a similar manner. This includes the indifference quality (p0),
the point at which the AOQL occurs (pm), and the relative slope of the OC curve at p0. But
remember this table was developed for k = 2.

Quick Switching System

The TNT plans offer protection for the producer in situations in which a zero acceptance
number is required on a stream of lots. A similar procedure will be found in the QSS pro-
posed by Dodge (1967) and studied extensively by Romboski (1969). The system uses
immediate switching to tightened inspection when a rejection occurs under normal inspec-
tion. The QSS-1 plan involves an immediate switch back to normal when a lot is accepted
under tightened inspection. Other plans allow a switch back to normal after two (QSS-2)
or three (QSS-3) lots are accepted under tightened inspection. Figure 17.3 shows how the
quick switching system (QSS-1) is applied.
Romboski (1969) has tabulated unity values for a variety of QSS plans. A table for
QSS-1 is given in Appendix Table T17.1 (Romboski’s Table I). It shows acceptance num-
bers under normal, cN, and tightened, cT, inspection for a fixed sample size n on both
tightened and normal. The unity values are used exactly as those for single, double, and
multiple plans given earlier. For example, the plan n = 20, cN = 1, cT = 0 has an operating
ratio R = 8.213 with

np.95 .308
p1 = = = .015
20 20
np 2.528
p2 = .10 = = .126
20 20

and indifference quality

np.50 1.146
IQ = = = .057
20 20

Lot
rejected

Start Normal Tightened

Lot
accepted

FIGURE 17.3
Quick switching system.
490 Acceptance Sampling in Quality Control

The value h0 given in the table of unity values is the relative slope of the OC curve at the
indifference quality level as defined by Hamaker (1950). Values of probability of accep-
tance for the individual normal PN and the tightened PT plans at the indifference quality
level for the scheme are also given.
Note from the example that p1 = .015 is approximately that of the plan n = 20, c = 1,
which is .018. However p2 = .126 approximates that of the plan n = 20, c = 0, which is .115.
Thus, the QSS plan affects a favorable compromise in protection between its tightened and
­normal constituents.
Romboski has examined a variety of QSS plans including variations in normal and tight-
ened sample sizes, acceptance numbers, and switching rules. The TNT plans offer an inge-
nious application of a QSS procedure for the case when the acceptance number is restricted
to 0. Thus, the QSS plans provide another vehicle for improvement of protection in situations
in which small sample sizes are necessary but high levels of protection must be maintained.
The QSS-1 plan, as introduced by Dodge (1967), works as follows:

1. Utilize a pair of sampling plans—a normal plan (N) and a tightened plan (T)—
with plan T having the tighter OC curve than plan N.
2. Use plan N for sampling the first lot.
3. If the lot is accepted, use plan N for inspection of the next lot and if rejected use
plan T for the next lot.

Romboski (1969) studied the QSS-1 plan using the single-sampling plan as a reference
plan. He then modified the switching rules of the QSS-1 system to create the QSS-2, QSS-3,
and QSS-4 plans. These modified systems produce OC curves that are more d ­ iscriminating
and efficient than the original QSS-1 system. Romboski also considered the QSS(n; kn; c0).
He noted that the operating ratio technique and the h0 procedure for finding match-
ing quick switching systems work in basically the same way for both QSS(n; cN, cT) and
QSS(n, kn; c0) systems. In addition, Romboski found that the sample size for a particular
cN = c0 is larger for the QSS(n, kn; c0) than for the corresponding QSS(n; cN, cT) system.
Soundararajan and Arumainayagam (1990) have described a system of modified QSS
plans designated as QSS-2, 3, 4(n; cN, cT) and QSS-2, 3, 4(n; kn; c0), including the special case
of c0 = 0, that is, the QSS-2, 3, 4(n; kn; 0) plans. Note that for the QSS-2, 3, 4(n; cN, cT) plans,
the acceptance number is tightened. In the case of the QSS-2, 3, 4(n; kn; c0) plans, the sample
size is tightened. We will now look at how these QSS plans work.
The QSS-2(n; cN, cT) plan works as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤cN, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >cN, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size n from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤cT, then accept the lot and continue
inspection until two (2) lots in a row are accepted. Once two accepts in succession
are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >cN, then reject the lot and go to step 4.

A table for QSS-2(n; cN, cT) is given in Appendix Table T17.3 (S&A’s Table 5).
Compliance Sampling 491

The QSS-3(n; cN, cT) plan works similar to that of the QSS-2(n; cN, cT) plan except for step 5.
The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤cN, then accept the lot and repeat
step 1.
3. If the number of defectives d in the sample is >cN, then reject the lot and go to
step 4.
4. From the next lot, take a random sample of size n from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤cT, then accept the lot and continue
inspection until three (3) lots in a row are accepted. Once three accepts in succes-
sion are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >cN, then reject the lot and go to
step 4.

A table for QSS-3(n; cN, cT) is given in Appendix Table T17.4 (S&A’s Table 6).
The QSS-4(n; cN, cT) plan works similar to that of the QSS-2(n; cN, cT) plan except for
step 5. The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤cN, then accept the lot and repeat
step 1.
3. If the number of defectives d in the sample is >cN, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size n from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤cT, then accept the lot and continue
inspection until four (4) lots in a row are accepted. Once four accepts in succession
are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >cN, then reject the lot and go to
step 4.

A table for QSS-4(n; cN, cT) is given in Appendix Table T17.5 (S&A’s Table 7).
Kuralmani and Govindaraju (1992) developed modified tables derived from the work of
Romboski (1969) and Soundararajan and Arumainayagam (1990) for the selection of quick
switching systems of type QSS-2, QSS-3, and QSS-4 for given values of the AQL, LQL,
α, and β. The QSS-2(n; kn, c0) plan works as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤c0, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >c0, then reject the lot and go to step 4.
492 Acceptance Sampling in Quality Control

4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤c0, then accept the lot and continue
inspection until two (2) lots in a row are accepted. Once two accepts in succession
are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >c0, then reject the lot and go to step 4.

A table for QSS-2(n; kn, c0) is given in Appendix Table T17.6 (K&G’s Table 2).
The QSS-3(n; kn, c0) plan works similar to that of the QSS-2(n; kn, c0) plan except for step 5.
The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤c0, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >c0, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤c0, then accept the lot and continue
inspection until three (3) lots in a row are accepted. Once three accepts in succes-
sion are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >c0, then reject the lot and go to step 4.

A table for QSS-3(n; kn, c0) is given in Appendix Table T17.7 (K&G’s Table 4).
The QSS-4(n; kn, c0) plan works similar to that of the QSS-2(n; kn, c0) plan except for step 5.
The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If the number of defectives d in the sample is ≤c0, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >c0, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
5. If the number of defectives D in the sample is ≤c0, then accept the lot and continue
inspection until four (4) lots in a row are accepted. Once four accepts in succession
are reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >c0, then reject the lot and go to step 4.

A table for QSS-4(n; kn, c0) is given in Appendix Table T17.8 (K&G’s Table 6).
The construction of these QSS plans is presented in Soundararajan and Arumainayagam
(1990), including the special case of c0 = 0. Govindaraju (1991) also produced tables for the
construction on the QSS-1(n, kn; 0) system. The operating procedure of the QSS-1(n, kn; 0)
system is as follows:

1. Take a random sample of size n1 (= n) from a lot under normal inspection and
count the number of defectives d. Accept the lot if the number of nonconforming
units or nonconformities is zero. Otherwise, reject the lot.
Compliance Sampling 493

2. If a lot is rejected under the (n1, 0) plan, then apply the single-sampling plan (n2, 0)
to the next lot where n2 = kn and k > 1. If a lot is accepted under the (n1, 0) plan, con-
tinue sampling with the (n1, 0) plan. If a lot is accepted under the (n2, 0) plan, then
switch to the (n1, 0) plan for the next lot. Otherwise, continue to use the (n2, 0) plan.

A table for the QSS-1(n, kn; 0) system is given in Appendix Table T17.9 (Govindaraju’s Table 1).
The QSS-2(n; kn, 0) plan works as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If there are no defectives d in the sample, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >0, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
5. If there are no defectives D in the sample, then accept the lot and continue inspec-
tion until two (2) lots in a row are accepted. Once two accepts in succession are
reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >0, then reject the lot and go to step 4.

A table for QSS-2(n; kn, 0) is given in Appendix Table T17.10 (selected columns from S&A’s
Table 11).
The QSS-3(n; kn, 0) plan works similar to that of the QSS-2(n; kn, 0) plan except for step 5.
The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If there are no defectives d in the sample, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >0, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
5. If there are no defectives D in the sample, then accept the lot and continue inspec-
tion until three (3) lots in a row are accepted. Once three accepts in succession are
reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >0, then reject the lot and go to step 4.

A table for QSS-3(n; kn, 0) is given in Appendix Table T17.11 (selected columns from S&A’s
Table 12).
The QSS-4(n; kn, 0) plan works similar to that of the QSS-2(n; kn, 0) plan except for step 5.
The procedure is as follows:

1. Take a random sample of size n from a lot under normal inspection and count the
number of defectives d.
2. If there are no defectives d in the sample, then accept the lot and repeat step 1.
3. If the number of defectives d in the sample is >0, then reject the lot and go to step 4.
4. From the next lot, take a random sample of size kn from a lot under tightened
inspection and count the number of defectives D.
494 Acceptance Sampling in Quality Control

5. If there are no defectives D in the sample, then accept the lot and continue inspec-
tion until four (4) lots in a row are accepted. Once four accepts in succession are
reached, go to step 1. Otherwise, repeat step 4.
6. If the number of defectives D in the sample is >0, then reject the lot and go to step 4.

A table for QSS-4(n; kn, 0) is given in Appendix Table T17.12 (selected columns from
S&A’s Table 13).
Several Excel templates used for the construction of these plans are discussed later in this
chapter with examples.

MIL-STD-1916
The QSS and TNT procedures illustrate the importance of the sampling scheme approach.
Probably, the broadest application of the concept is contained in MIL-STD-1916 (United
States Department of Defense 1996) issued on April 1, 1996. This standard addresses the
importance of statistical process control (SPC) in modern acceptance control by incorporat-
ing an evaluation of the quality management system along with c = 0 attributes sampling,
variables sampling, and continuous sampling plans as alternate means of acceptance in
one standard. Thus, the standard is unique since not only is there switching among plans,
but different alternate acceptance procedures may be selected from this standard as well.

Structure
The structure of the MIL-STD-1916 standard is outlined in Figure 17.4. The first three
sections are devoted to housekeeping details and definitions. The fourth section
addresses general considerations in the use of the standard. This includes the choice
between an alternate acceptance method, such as ISO 9000 or MIL-Q-9858, and con-
ventional sampling tables. The fifth section presents detailed requirements for devel-
oping and confirming the adequacy of the quality system if it is to be used instead of
sampling. It also presents the preferred sampling inspection tables and procedures
to be used in lieu of the quality systems approach. The sixth section provides some
administrative notes.
The appendix of MIL-STD-1916 includes some excellent examples of the use of the sam-
pling tables that facilitate the implementation of the standard.

Operation
MIL-STD-1916 provides two different and distinct means of product acceptance: accep-
tance by contractor proposed provision and acceptance by tables. The former requires
qualification and verification of the quality management system associated with the prod-
uct. The latter relies on traditional sampling plans for acceptance. The contractor and the
customer must make a decision on which to use at the outset.
If the contractor elects to rely on the quality system to demonstrate acceptabil-
ity of the product, quality system documentation including a quality plan will be
required showing that the system is prevention based with a process focus. Evidence
of the implementation and effectiveness of the quality system will be required.
Compliance Sampling 495

Applicable Detailed Master Examples of

General
Scope Definitions sampling plan
documents requirements requirements tables
in use

Continuous sampling plans

Attribute sampling plans

Variables sampling plans

ANSI/ASQC ISO

Code letters
Q9000 Quality management and quality

Q9004 Quality management and quality

Q9000 Quality management and quality

Q9004 Quality management and quality

B3 Control chart method of controlling
B1 Guide for quality control charts

assurance standards guidelines for

assurance standards guidelines for
B2 Control chart method of

system elements guidelines

8402 Quality—vocabulary
system elements guidelines
quality during production
analyzing the data

selection and use

selection and use

FIGURE 17.4
Structure of MIL-STD-1916.

This includes evidence of systematic process improvement based on process control

and ­demonstrated product conformance.
If it is decided to use tables for the acceptance of product, the approach is more
conventional. Given lot size and verification level (VL), a code letter is selected from
Appendix Table T17.13 (Table I of MIL-STD-1916). The standard provides seven VLs
with level 7 being the most stringent. The VLs play a role similar to the AQLs of MIL-
STD-105E and allow for adjustment of the severity of inspection. If no VL is specified,
the default levels are

Defect Type Default VL

Critical VII
Major VI
Minor I

Tables are provided for three different sampling schemes: attributes, variables, and con-
tinuous. Each is indexed by VL and code letter. They are matched so it is possible to switch
easily from one to another. All attributes plans in the standard have c = 0.
Sampling schemes require switching rules. These are illustrated in Figure 17.5. They are
simple and effective and must be used to ensure that the system will produce the level of
protection desired.
A check sequence for selecting a plan from MIL-STD-1916 is given in Figure 17.6.
496 Acceptance Sampling in Quality Control

Preceding 10 lots or batches

accepted (tables II and III) or
no units rejected in last 10*
Start
na(N) inspections (Table IV)
Production steady
2 out of 5
Contractor’s quality system
consecutive lots or
considered satisfactoty by
batches rejected
government
(Tables II and III) or
Reduced inpsection 2 units rejected in ≤5*
considered desirable by na(N) inspections
government (Table IV)

Reduced Normal Tightened

Lot or batch rejected (Tables II Five consecutive lots

and III) or a unit is rejected or batches accepted
(Table IV) (Tables II and III) or no
Production irregular units rejected in last 5*
Process
na(T) inspections
Contractor’s quality system is remains on
(Table IV)
unsatisfactory tightened
Other conditions warrant
a return to normal inspection
Government
reserves right to
discontinue the
acceptance

FIGURE 17.5
Switching rules.

Implementation
Implementation of the sampling aspect of MIL-STD-1916 is accomplished through four
tables

Appendix Table MIL-STD-1916 Table Description

T17.13 Table I Code letters

T17.14 Table II Attributes sampling plans
T17.15 Table III Variables sampling plans
T17.16 Table IV Continuous sampling plans

which will be illustrated by example. Figures 1 through 4 of MIL-STD-1916 highlight the

use of these tables. Recall for all attributes plans, c = 0.
Assume as in our discussion of MIL-STD-105E that a lot is presented having a lot size
of 500 and (lacking a specified value) verification level (VL) of IV, then the default VL is
to be used. Table I of the standard gives code A. Table II (Appendix Table T17.14) speci-
fies na = 80, c = 0 for the normal inspection plan. The tightened plan is found directly to
the left of the normal plan, namely, nt = 192, c = 0, and the reduced plan to the right of
the normal plan is nr = 32, c = 0. Note that columns T and R in Appendix Table T17.13
Compliance Sampling 497

Start

VL is specified in contract or product spec.

Determine lot size

Obtain code entry from Table I

Determine the type of sampling to be used and select the plan

Attributes Variables Continuous

Tables ΙΙ Tables ΙΙΙ Tables ΙV

Check the operating characteristics

to insure that protection is as desired
that (Appendix D)

Apply switching rules (Section 5.2.1.3)

Normal: VL column
Tightened: column to left of VL col
Reduced: column to right of VL col

FIGURE 17.6
A check sequence for selecting a plan from MIL-STD-1916.

(Table I in MIL-STD-1916) provide for tightened and reduced plans when operating
with normal VLs VII and I. Switching rules are as outlined in the standard as shown in
Figure 17.5.
The variables plans are addressed in a similar manner and are taken from Appendix
Table T17.15 (Table III in MIL-STD-1916). For code A, VL IV we have the normal plan
nV = 29, k = 2.40, and F = .193. Thus, the maximum allowable standard deviation is
smax
£F
U -L

or

smax £ F (U - L )
498 Acceptance Sampling in Quality Control

Also, we have for the tightened plan nV = 44, k = 2.69, F = .174 and for the reduced plan
nV = 18, k = 2.05, F = .222.
If continuous sampling is to be used, plans are taken from Appendix Table T17.16
(Table  IV in MIL-STD-1916). The normal plan in the example is f = 1/12, i = 264. The
­corresponding tightened plan is f = 2/17, i = 527 with a reduced plan of f = 1/17, i = 125.
For example, we have developed a matched set of three schemes:

Attributes Variables Continuous

Tightened na(T) = 192, c = 0 nV(T) = 44, k = 2.69, F = .174 f = 2/17, i = 527

Normal na(N) = 80, c = 0 nV(N) = 29, k = 2.40, F = .193 f = 1/12, i = 264
Reduced na(R) = 32, c = 0 nV(R) = 18, k = 2.05, F = .222 f = 1/17, i = 125

Use of the switching rules is slightly more complicated for the continuous plans. Recall the rule
for going from tightened to normal if five consecutive lots are accepted. This would amount
to no defectives in five times the lot size, since c = 0. Thus, the rule for continuous sampling
would be to switch to normal from tightened if no defectives were found in a span of five
times the sample size of the corresponding matched attribute plans. The rules then become

Normal to tightened Two defectives found in a span of 5(na(N)) units inspected

Tightened to normal No defectives found in a span of 5(na(T)) units inspected
Normal to reduced No defectives found in a span of 10(na(N)) units inspected
Reduced to normal A defective unit is found

For our example, the spans become

Normal to tightened 5(80) = 400

Tightened to normal 5(192) = 960
Normal to reduced 10(80) = 800

Of course, other conditions impact the switching rules as well. Refer to Figure 17.5 for
these requirements.

Measures
MIL-STD-1916 does not present tables of various measures of the sampling plans as does
MIL-STD-105E. However, it is supplemented by a Department of Defense Handbook, “Companion
Document to MIL-STD-1916.” This handbook presents an exhaustive compilation of graphs
and tables of the properties of the MIL-STD-1916 plans. It includes OC curves, AOQ curves,
AFI curves, and associated tables. The handbook provides for the VL defaults cited earlier.
In support of the quality systems approach to acceptance, the handbook discusses in-depth
various aspects of a prevention-based quality system that provides a process focus for the
quality system. Various tools are discussed including SPC and other measures of performance.

Further Considerations
MIL-STD-1916 appears as international standard ISO 21247 (2005). The tables are essentially
the same except for minor editorial changes. Measures of performance are also presented
Compliance Sampling 499

in the form of tables of percentage points and other descriptive material. Schematic dia-
grams of the switching rules are also given. Acceptance by supplier-proposed provisions
is maintained and detail is presented.
E2910 Standard Guide for Preferred Methods for Acceptance of Product represents the ASTM
standard version of MIL-STD-1916. The scope of this standard is to establish lot or batch
and continuous sampling plans using MIL-STD-1916 as a basis. It represents an alterna-
tive sampling approach to attributes sampling (MIL-STD-105E, same as E2234), variables
sampling (MIL-STD-414, E2762), and continuous sampling (MIL-STD-1235B, E2819). This
guide provides the sampling plans of MIL-STD-1916 in ASTM format for use by ASTM
committees and others. It recognizes the continuing usage of MIL-STD-1916 in industries
supported by ASTM. Most of the original text in MIL-STD-1916 is preserved in Sections 4–6
of this guide. The original wording of “Government” in MIL-STD-1916 has been changed
to “consumer,” and “contractor” has been changed to “producer” to make this standard
more generic.

Simplified Grand Lot Procedure

Acceptance inspection and compliance testing often necessitate levels of protection for
both the consumer and the producer that require large sample sizes relative to lot size.
A given sample size can, however, be made to apply to several lots jointly if the lots can be
shown to be homogeneous. This reduces the economic impact of a necessarily large sample
size. Grand lot schemes, as introduced by Simon (1941), can be used to affect such a reduc-
tion. The original grand lot scheme was later modified by Schilling (1979) to incorporate
graphical analysis of means procedures in verifying the homogeneity of a grand lot. The
resulting approach can be applied to attributes or variables data, is easy to use, provides
high levels of protection economically, and can reduce sample size by as much as 80%. It
may be applied to unique “one-off” lots, isolated lots from a continuing series, an isolated
sequence of lots, or to a continuing series of lots.
MIL-STD-105E indicates that lots or batches should be formed in such a way that “each
lot or batch shall, as far as is practicable, consist of units…manufactured under essentially
the same conditions….” This suggests that one way to reduce sample size relative to lot
size is to increase the size of the lot as much as possible within limits of homogeneity of the
product included therein. The grand lot scheme, as originated by Simon, utilizes the power
of the control chart to achieve a drastic reduction in sample size relative to lot size in the
application of an acceptance sampling plan to homogeneous material.
As defined by Simon (1944), “A lot is an aggregation of articles that are essentially alike.”
The control chart is used to distinguish members of what may be considered a grand lot
from a collection of sublots. Thus, the chart becomes the criterion for what is “alike” in
Simon’s definition. Use of a large sample size on the grand lot allows a precision in sam-
pling that would often prove uneconomical if applied to each sublot separately. Thus, use
of plans having low discrimination, such as c = 0 attributes plans, can be avoided.
For example, if the protection desired requires a sample size of 800 on each of eight
sublots, a total of 6400 units would be inspected. If the eight sublots were sufficiently alike
to allow aggregation into a single grand lot, the resulting sample of 800 on the grand lot
would amount to an 87.5% reduction in sample size with no decrease in protection. In this
way, very stringent levels of quality can be assured with great economy.
500 Acceptance Sampling in Quality Control

The concept of the grand lot has application in a number of areas of acceptance sampling:

1. Verification of the quality of large quantities of material on an acceptance sam-

pling basis to very stringent quality levels (e.g., sampling for safety-related defects
when sample sizes required are prohibitively large when applied to sublots).
2. Sampling quarantined or returned material to distinguish aberrant sublots and to
determine disposition of the material (e.g., sampling many skids of material that
have been rejected by a customer an unsatisfactory when it is believed that only a
few skids may, in fact, be nonconforming).
3. Acceptance sampling of unique lots (sometimes called one-off lots) or of isolated
lots from a continuing series, when the lot to be inspected may logically be divided
into a set of sublots (e.g., acceptance of an order that is comprised of material pro-
duced by several different identifiable production units; the order may or may not
be part of a continuing series).
4. Acceptance sampling of an isolated sequence of sublots that may logically be
aggregated into a grand lot (e.g., acceptance of a week’s production on the basis of
results from individual days [or shifts]).
5. Acceptance sampling of a continuing series of sublots that can be shown to con-
sistently comprise a grand lot (e.g., acceptance on an open order from a captive
supplier).

Not only is the procedure useful in routine inspection of individual lots or series of lots of
material, but it is also especially valuable in surveillance inspection and in compliance test-
ing. The experience of the U.S. Army, Chemical Corps Material Command, in this regard
has been set forth by Mandelson (1963).
The original grand lot scheme, as proposed by Simon, utilized a control chart approach
to identifying the grand lot, but was complicated by the necessity of including extra proce-
dures for assessing the compound probabilities associated with a simultaneous compari-
son of many points (representing the sublots) against the limits. Conventional control chart
limits are set up to be used one point at a time. While Burr (1953) modified the application
of the approach somewhat by using approximations in place of Simon’s original use of the
incomplete beta function, the extra steps incurred by this part of the method remained.
The  procedure was simplified still further by Schilling (1979) by providing graphical
control chart limits that automatically account for these compound probabilities and by
incorporating an identical approach for both variables and attributes data. The simplified
procedure is based on the use of the analysis of means limits developed by Ott (1967),
Ott and Lewis (1960), and Schilling (1973a,b). Use of these limits retains the simplicity of
the control chart without recourse to additional steps since they are designed to maintain
specified probability levels when many points are compared to the limits.

Simon’s Approach
The approach suggested by Simon (1941) is essentially as follows:

1. Determine an appropriate sublot sample size.

2. Sample the sublots.
3. Plot the sample results on a control chart of the form shown in Figure 17.7.
Compliance Sampling 501

Extreme suspected mavericks

(E) Upper extreme limit
Suspected mavericks
(M) Upper maverick limit

Grand lot Average level

(M) Lower maverick limit

Suspected mavericks
(E) Lower extreme limit
Extreme suspected mavericks

FIGURE 17.7
Simon grand lot chart. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 117, 1979. With permission.)

4. Identify any points that plot beyond the extreme limits (E) as “extreme suspected
mavericks.” Eliminate these sublots, as outliers, from further consideration as
part of the grand lot. Treat the eliminated lots separately, applying an appropriate
­sampling plan to each. Recompute the limits until there are no further extreme
suspected mavericks outside the extreme limits (E).
5. Identify any points outside the maverick limits (M) as suspected mavericks.
Utilize the incomplete beta function to determine if the number of maverick points
is significantly large on the basis of the compound probabilities inherent in con-
ventional control limits. If it is, reject the grand lot hypothesis. (Note, this step
is unnecessary when using analysis of means limits on the control chart since a
single point beyond the maverick limits is sufficient to reject the grand lot hypoth-
esis when such limits are used.)
6. If the grand lot hypothesis is rejected, test each sublot separately using an appro-
priate sampling plan. This plan would normally be the same as that applied to the
grand lot if the grand lot hypothesis had been accepted.
7. If the grand lot hypothesis is accepted, combine the sublots not determined to be
extreme suspected mavericks into a grand lot and apply an acceptance s­ ampling
plan sufficient to give the consumer and the producer the protection desired,
­taking additional samples as necessary to complete the sample size required.

This procedure allows application of a very discriminating sampling plan to lots made as
large as possible. The sampling plan applied to the grand lot should afford at least LTPD
protection to the consumer with due consideration for the producer’s risk as evidenced by
the OC curve. Application to large lots allows higher acceptance numbers to be used with
larger sample sizes that leads to better protection for both parties.
In discussing the grand lot approach, Simon (1941) points out that, if the grand lot hypoth-
esis is rejected, “… the grand lot judge is called upon to revise the grand-lot grouping, if a
logical basis for regrouping exists, or the grand lot must be abandoned and resort made to
individual sampling.” This provides greater flexibility in application, however, regrouping
should be allowed only on a documented rational basis and only with the concurrence of
both parties to the acceptance decision—the producer and the consumer. Simon (1941) also
states that “very good grand lot judgments are desirable but not essential to the operation
of the system, as very poor ones will almost inevitably be caught. Poor grand lot judg-
ments result in retesting…and serve to decrease the efficiency of the system.”
502 Acceptance Sampling in Quality Control

Simplified Procedure: Attributes

Given a presumptive grand lot made up of k sublots, it is desired to obtain LTPD protection
against a process fraction defective pt.

1. Determine the sublot sample size,* n, as

2.303
n=
pt

Round up.
2. Sample n items from each sublot and determine p, the estimated fraction defective
from the nk units sampled, as

X
p=
N

where
X is the total number nonconforming
N = nk is the total sample size
3. Construct an analysis of means chart in the form of Figure 17.7 where
a. Extreme limits are set at

p (1 - p )
E: p ± H.002
n

b. Maverick limits are set at

p (1 - p )
M : p ± H.05
n

using the Hα factors from Appendix Table T17.17.

4. Eliminate any sublot whose sample proportion plots beyond extreme limits (E)
from further consideration as part of the grand lot. Dispose of such lots separately
using an appropriate sampling plan. Sublots that plot below the lower extreme
limit, however, may be accepted if the grand lot is accepted.
5. Recompute limits on the remaining points until all extreme suspected mavericks
have been eliminated. Then, if any remaining points plot beyond the maverick
limits (M), reject the grand lot hypothesis and test each sublot individually using
an appropriate plan on each.

* As suggested by Simon, this gives roughly 90% probability of obtaining at least one defective in the sublot
­sample if the process fraction defective is, in fact, pt. Alternatively, sublot sample size may be determined by
sampling a fraction of the sublot, as obtained from a lot sensitive sampling plan. This relates sample size directly
to sublot size and maintains protection equivalent to the formula given earlier, with slightly smaller samples.
For large sublots, the results will be the same for both approaches.
Compliance Sampling 503

6. If all points lot within the maverick limits (M), accept the grand lot hypothesis and
group the remaining sublots into a grand lot. Apply a standard sampling plan to the
grand lot to obtain the required LTPD protection, using a sample size-­acceptance
number combination, which will afford reasonable protection for the producer.
Take additional samples as necessary to complete the required sample size.

Example: Attributes
Suppose a shipment consisting of 12 cartons, each containing 5000 parts, for use in an assem-
bly operation, is presented for incoming inspection. The production process can ­tolerate
2.5% defective, but quality of 6% or more must be rejected. Inspection is on a g
­ o-no-go basis.
A grand lot plan is to be used with pt = .06.

1. The sublot sample size is determined to be

2.303
n= = 38.4 ~ 40
.06

2. Sample results are as follows:

Carton Sample Size Defectives Proportion Defective

1 40 1 .025
2 40 2 .050
3 40 2 .050
4 40 5 .125
5 40 0 .000
6 40 4 .100
7 40 3 .075
8 40 1 .025
9 40 7 .175
10 40 2 .050
11 40 1 .025
12 40 1 .025
Total 480 29 .060

Source: Data adapted from Simon, L.E., An Engineer’s Manual of Statistical Methods,

John Wiley & Sons, New York, 1941.

3. Limits are set at

.06 ( .94 )
E: .06 ± 3.60
40
.06 ± .135
.0 to .195
.06 ( .94 )
M : .06 ± 2.74
40
.06 ± .103
.0 to .163

and the resulting analysis of means chart is shown in Figure 17.8.

504 Acceptance Sampling in Quality Control

.20 E = .195
.18 x
.16 M = .163
.14
.12 x

.10 x
p .08 x
.06 p = .06
x x x
.04
.02 x x x x
0 x
1 2 3 4 5 6 7 8 9 10 11 12
Sublot

FIGURE 17.8
Analysis of means chart—attributes data. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 120, 1979. With
permission.)

4. No sublots are identified as extreme suspected mavericks since none plots beyond
the extreme limit (E). It is, therefore, unnecessary to recompute the limits.
5. Sublot 9 is identified as a suspected maverick since it plots beyond the maverick
limit (M). The grand lot hypothesis is rejected and each sublot must be inspected
separately. Using MIL-STD-105E, it is found from Table VI-A that for an AQL of
2.5% and an LTPD of 5.6%, the plan n = 500, c = 21 will give the desired protection
on an isolated lot. Accordingly, an additional sample of 460 must be taken from
each sublot and the plan applied to the samples of 500.
6. If the grand lot hypothesis had been accepted, results from the individual lots
could be aggregated. A further sample of 20 would be taken at random from the
total shipment to reach the sample size of 500 necessary for application of the
MIL-STD-105E plan.

It is interesting to note that rejection of the grand lot hypothesis resulted in inspection of
6000 units; whereas if the grand lot hypothesis had been accepted, inspection of only 500
units would have been required. This could have resulted in a 92% decrease in inspection
effort; however, the procedure identified the lack of homogeneity of the cartons, making
aggregation deceptive and unwarranted.

Simplified Procedure: Variables

Given a presumptive grand lot made up of k sublots, it is desired to apply a variables sam-
pling plan for a measurement characteristic, X.

1. Determine the sublot sample size n as

120
n= +1
k

Round up. In none of the case sample less than five items from a sublot.
Compliance Sampling 505

2. Sample n items from each sublot and compute X and s from each as follows:

åX
1
Xj = ij
n i =1
n

å(X )
2
ij - Xj
sj = i =1

n -1

where Xij is the ith observation of the measurement characteristic from the jth
­sublot and X j and sj are the sample mean and standard deviation of the jth sublot.
Also obtain

åX
1
X= j
k j =1

ås
1
sˆ = 2
j
k j =1

3. Construct an analysis of means chart for s as in Figure 17.7, where

a. Extreme limits are set at

s
E: s ± H.002
2n

b. Maverick limits are set at

s
M : s ± H.05
2n

using the Hα factors of Appendix Table T17.17.

4. Eliminate any sublot whose standard deviation plots beyond the extreme limits
(E) from further consideration as part of the grand lot. Dispose of such sublots
separately using an appropriate sampling plan. Sublots that plot below the lower
extreme limit, however, may be accepted if the grand lot is accepted, provided
they are not disqualified on the basis of their mean.
5. Recompute limits on the remaining points until all extreme suspected mavericks
have been eliminated. Then, if any remaining points plot beyond the maverick
limits (M), reject the grand lot hypothesis and test each sublot individually using
an appropriate sampling plan.
6. If all remaining points plot within the maverick limits (M), accept the grand lot
hypothesis for the standard deviations and proceed to test the means against the
grand lot hypothesis.
506 Acceptance Sampling in Quality Control

7. For the sublots not eliminated as extreme suspected mavericks in testing the stan-
dard deviations and using their estimated grand standard deviation, ŝ, and grand
mean, X an analysis of means chart with
a. Extreme limits are set at

s
E: X ± H.002
n

b. Maverick limits set at

s
M : X ± H.05
n

8. Eliminate any sublot whose mean plots beyond extreme limits (E) from further
consideration as part of the grand lot. Dispose of such lots separately. Recompute
limits on the remaining points. However, do not recompute limits for testing s
against the grand lot hypothesis.
9. Recompute limits on the remaining points until all extreme suspected mavericks
have been eliminated. Then, if any remaining points plot beyond maverick limits
(M), reject the grand lot hypothesis and test each sublot individually using an
appropriate sampling plan.
10. If all points plot within the maverick limits (M), accept the grand lot hypothesis
and group the remaining sublots into a grand lot. Apply a standard sampling plan
to the grand lot to obtain desired protection. Take additional samples as necessary
to complete the required sample size.

Example
An arms wholesaler receives 30 consecutive lots of rounds of ammunition. These lots are
to be tested for muzzle velocity. Specifications require an individual round to be in the
range 1670–1790 ft/s. A grand lot plan is to be employed.

1. The sublot sample size is

120
n= +1= 5
30

2. Sample results are as follows:

Lot X s Lot X s

1 1711 16.9 16 1783 20.6

2 1711 16.1 17 1777 3.6
3 1713 15.7 18 1794 6.0
4 1718 10.5 19 1773 14.9
(Continued)
Compliance Sampling 507

Lot X s Lot X s
5 1735 4.0 20 1789 21.8
6 1739 10.1 21 1798 6.0
7 1723 15.7 22 1789 11.7
8 1741 6.0 23 1788 15.7
9 1738 4.4 24 1799 12.1
10 1725 12.5 25 1807 17.7
11 1731 10.1 26 1784 4.3
12 1721 7.7 27 1775 15.7
13 1719 17.3 28 1787 12.8
14 1735 15.7 29 1770 6.1
15 1741 5.9 30 1796 19.7

Source: Adapted from Simon, L.E., An Engineer’s Manual of Statistical

Methods, John Wiley & Sons, New York, 1941, p. 367.

52710
X= = 1757
30
5123.33
sˆ =
30
= 13.1
3. Limits for s are set at
13.1
E: 13.1 ± 3.92
10
13.1 ± 16.2
0 to 29.3
13.1
M : 13.1 ± 3.09
10
13.1 ± 12.8
0.3 to 25.9
and the resulting analysis of means chart is shown in Figure 17.9.
4. The analysis of means plot for s shows no extreme suspected mavericks, so the
limits need not be recomputed.
5. There are no suspected maverick lots on the basis of the analysis of means plot for s.
6. The grand lot hypothesis is accepted for standard deviations, and so the means
are analyzed next.
7. Limits for X are set at
13.1
E: 1757 ± 3.92
5
1757 ± 23.0
1734.0 to 1780.0
13.1
M: 1757 ± 3.09
5
1757 ± 18.1
1738.9 to 1775.1

and the resulting analysis of means chart is shown in Figure 17.10.

508 Acceptance Sampling in Quality Control

30
E = 29.3
28
26 M = 25.9
24
22
x x
20 x
18 x
x x
Velocity

16 x
x x x x x x
14
ˆs = 13.1
12 x x
x x
10 x x x
8 x
6 x x x x x
4 x x x
x
2
0
M = 0.3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Sublot

FIGURE 17.9
Analysis of means chart—standard deviation. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 122, 1979.
With permission.)

1810 x
1800 x x
x
1790 x x x x x
x x x
1780 E = 1780
x x M = 1775.1
1770 x
Velocity

1760 X = 1757
1750
1740 x x x M = 1738.9
xx x E = 1734
1730 x
x
1720 x x
x x
1710 x x x
1700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Sublot

FIGURE 17.10
First analysis of means chart—means. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 122, 1979. With
permission.)

8. The analysis of means chart immediately shows a shift at sublot 16. Also, 21 of
the 30 points are extreme suspected mavericks.
9. Clearly, the grand lot hypothesis must be rejected. However, it is also evident
that the shipment may be composed of two potential grand lots consisting of
sublots 1–15 and 16–30, respectively.
10. The limits may be recalculated for these two groups as follows:
Lots 1–15 (k = 15), X = 1726.7
Compliance Sampling 509

13.1
E: 1726.7 ± 3.69
5
1726.7 ± 21.6
1705.1 to 1748.3
13.1
M: 1726.7 ± 2.84
5
1726.7 ± 16.6
1710.1 to 1743.3

Lots 16–30 (k = 15), X = 1787.3

13.1
E: 1787.3 ± 3.69
5
1787.3 ± 21.6
1765.7 to 1808.9
13.1
M: 1787.3 ± 2.84
5
1787.3 ± 16.6
1770.7 to 1803.9

The resulting analysis of means plot is shown in Figure 17.11.

11. Figure 17.11 reveals that sublots 1–15 can be considered a grand lot, aggregated,
and tested accordingly. Sublots 16–30, however, cannot be considered to form a
grand lot since lots 25 and 29 are suspected mavericks.
12. Assuming the normality of the underlying distribution of measurements, a
variables plan from MIL-STD-414 (United States Department of Defense 1957)
may be selected from the OC curves to give an AQL of 0.1% and a consumer

1810 E = 1808.9
M = 1803.9
1800 x x x
x
1790 x x
x x X = 1787
1780 x
x x
1770 x M = 1770.7
E = 1765.7
1760
Velocity

1750 E = 1748.3
1740 x x M = 1743.3
x x x x
1730 x
x X = 1726.7
1720 x x x
x
x
1710 xx M = 1710.1
E = 1705.1
1700
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Sublot

FIGURE 17.11
Second analysis of means chart—means. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 123, 1979. With
permission.)
510 Acceptance Sampling in Quality Control

quality level of 1.0% with 10% probability of acceptance. Such a plan is Code N,
0.1% AQL, with standard deviation unknown. This requires a sample size
of n =  75 with an acceptance constant k = 2.66, so no additional samples are
needed. Standard variables acceptance procedures may then be applied sepa-
rately to both specification limits, 1670 and 1790, respectively, since they are
estimated to be more than 9 standard deviations apart, allowing a maximum
standard deviation of 20.88. The acceptability criterion is
U-X
>k
sˆ
1790 - 1726.7
> 2.66
13.1
4.83 > 2.66
and
X-L
>k
sˆ
1726.7 - 1670
> 2.66
13.1
4.33 > 2.66

The acceptability criterion is met and so the grand lot consisting of sublots 1–15
is accepted. Note that the remaining lots 16–30 must be inspected separately,
requiring an additional sample of 1050 if equivalent protection is to be main-
tained on each of them. This illustrates the leverage possible from the formation
of a grand lot.

Continuing Series of Lots

In introducing the grand lot plan, Simon (1941, p. 33) pointed out that it can easily be
applied to a continuing series of lots. He also suggested the following approach:
From the first few lots, or at least from the first lot, one must take a large sample in order
to have a reliable estimate of the manufacturer’s general level of quality as measured by
the lot fraction defective…From then on, one can treat his successive lots as additional
members of the grand lot, testing each suspected maverick by a large sample to see if its
quality is really satisfactory. However, the occurrence of an extreme suspected maverick
or an excessive number of suspected mavericks should result in terminating the manu-
facturer’s grand lot and in making him qualify all over again.

This can be carried out using the simplified procedure as follows:

1. Qualify the first 10 lots* using a standard sampling plan with a sample size-­
acceptance constant combination sufficient to protect both the consumer and the
producer. Sample size for the qualification must equal or exceed that determined
from the sublot sample size formulas given in the simplified method for attributes
or for variables with k = 10 lots.
2. Test the grand lot hypothesis on the first 10 lots using an analysis of means plot,
as set forth in the simplified procedure.

* This conforms to the criterion for switching from normal to reduced inspection under MIL-STD-105E and with
the control chart approach of MIL-STD-105A.
Compliance Sampling 511

3. If any of the 10 lots fail the standard sampling plan or if the grand lot hypothesis
is rejected, the producer must requalify subsequent lots.
4. If the grand lot hypothesis is accepted, construct a control chart as in Figure 17.7
using probability limits to test subsequent lots. Use the overall values of p  attri-
butes or X and ŝ for variables obtained from the 10 qualification lots to set up the
limits. Sample size from each subsequent lot is

2.303
Attributes: n = Variables: n = 5
pt

Limits are set using the formulas given in the simplified procedure employing
standard control limits, so that

H.002 = 3.09

H.05 = 1.96

5. Lots that plot within the extreme limits are accepted. Lots that plot outside the
extreme limits must be tested individually using an appropriate acceptance sam-
pling plan.
6. The producer must requalify if any lot plots outside the extreme limits or if two
out of any successive five points plot outside the maverick limits* in an undesir-
able direction.

Example
Consider the attributes data given earlier. Suppose these constitute the next 12 from a
continuing series of lots. The producer and the consumer agree to use an AQL of 2.5%
and an LTPD of 6%. Table VI-A of MIL-STD-105E shows that for isolated lots, the plan
n = 500, c = 21 is appropriate. Suppose sample results on the preceding 10 lots were

Lot Sample Size Number Defective Proportion Defective

~9 500 15 .030
~8 500 10 .020
~7 500 13 .026
~6 500 18 .036
~5 500 15 .030
~4 500 12 .024
~3 500 15 .030
~2 500 13 .026
~1 500 20 .040
0 500 19 .038
Total 5000 150 .030

* This is essentially the same as the criterion for switching to tightened inspection under MIL-STD-105E.
512 Acceptance Sampling in Quality Control

1. All 10 qualification lots pass the standard plan.

2. The analysis of means plot to test the grand lot hypothesis has limits

.03 (.97 )
E: .03 ± 3.53
500
.03 ± .027
.003 to .057
.03 (.97 )
M: .03 ± 2.66
500
.03 ± .020
.010 to .050

and the resulting analysis of means chart is shown in Figure 17.12.

3. The analysis of means plot shows that the grand lot hypothesis is accepted, and
since the 10 lots passed the standard plan, a control chart can be instituted.
4. Subsequent lots are sampled using a sample size of

2.303
n= = 38.4 ~ 40
.06

.13 Qualification Control chart

.12
.11 E = .113
.10
.09
.08 M = .083

.07
.06
p

E = .057
.05 M = .050 x x
.04 x
x x
.03 x x x p = .03
x x x x
.02 x
.01 M = .010
0 E = .003
–9 –8 –7 –6 –5 –4 –3 –2 –1 –0 1 2 3 4
Lot

FIGURE 17.12
Charts for continuing series of lots. (Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 125, 1979. With
permission.)
Compliance Sampling 513

and the resulting fractions defective plotted on a control chart with limits

.03 (.97 )
E: .03 ± 3.09
40
.03 ± .083
0 to .113
.03 (.97 )
M: .03 ± 1.96
40
.03 ± .053
0 to .083
The control chart for lots 1–4 is also shown in Figure 17.12.
5. Lot 4 plots outside the extreme limit and so it must be subjected to further
testing.
6. The grand lot hypothesis is rejected at lot 4 and the producer would now have
to requalify from the beginning of the procedure.

Further Considerations
Simon’s (1941) original approach to testing the grand lot hypothesis after the extreme
­suspected mavericks were eliminated was as follows:

1. Prepare a control chart for the property being tested (p, s, or X) with a specified
probability, Q, of exceeding the maverick limits.
2. Count the actual number of points outside the maverick limits.
3. Test the null hypothesis that the probability of exceeding the limits is equal to that
specified, against an alternate hypothesis that it is greater. This is done by compar-
ing the actual number of exceedances found against a critical value obtained from
the incomplete beta function with parameter Q, sample size k, and Type I risk α = .10.
4. Accept or reject the grand lot hypothesis as this null hypothesis is accepted or
rejected.

This part of the procedure was intended to account for the degradation of the Type I risk
of a control chart as the increasing number of points is compared to the limits. It is well
known, for example, that if the probability of one point exceeding the limits is .05, the
probability of two points plotted outside the limits is

1 - ( .95 ) = .0975
2

and three points plotted outside the limits is

1 - ( .95 ) = .1426
3

and so on.
Schilling (1979) pointed out that the advent of analysis of means procedures for con-
structing decision limits, which adjust the limits to take account of the number of points
involved in the comparison, made this part of the original procedure unnecessary.
514 Acceptance Sampling in Quality Control

The operation of the grand lot plan was greatly simplified by straightforward comparison
of the points against the analysis of means limits. The Hα factors are upper bounds for the
studentized maximum absolute deviate as derived by Halperin et al. (1955), which were
incorporated into the analysis of means procedure by Ott (1967), Ott and Lewis (1960),
and Schilling (1973a,b). They were computed using a result attributed to Tukey (1953)
in the manner described by Schilling (1973a). As such, the limits apply regardless of any
correlation that might exist between the lots. The values presented are for a Type I risk of
.002 and .05, respectively. This corresponds to the British system of probability limits for
control charts, as presented, for example, by Pearson (1935), which employ these levels of
risk for action and warning limits. The risks are very close to those used by Simon (1941)
in his approach to variables data. They are also used in the simplified procedure with
attributes data for reasons of consistency and uniformity.
The attributes sample size formula for sublots was chosen to conform to Simon’s original
recommendation to allow a 90% probability for at least one defect to occur when product
quality is at the critical level. The formula is based on the Poisson distribution and so is
conservative in cases where the binomial or hypergeometric distributions should apply.
Chart limits for attributes employ the normal approximation to the binomial distribution
but should be adequate in practice.
Sample size for variables data was chosen to give an estimate of the standard devia-
tion with 120 degrees of freedom. A study of tables of Hα indicates that 120 degrees of
freedom is sufficient to allow use of Hα factors for standard deviation known (i.e., using
df = ∞ as an approximation). This corresponds to the practice of starting a control chart for
subgroups of five after 30 points have been plotted (120 df) and using the estimate of the
standard deviation as if it were known. In this way, one set of Hα factors could be presented
for use with both variables and attributes data. Note that the values given in Appendix
Table T17.17 are for infinite degrees of freedom and correspond to values of the extreme
standardized deviate from the sample mean.
The method for dealing with a continuing series of lots was, of course, initially suggested
by Simon (1941). A straightforward analysis of means is performed on the first 10 lots since
they are simultaneously compared to the decision limits. Thereafter, a control chart with
British probability limits is utilized since subsequent lots will be compared to the limits
individually, one at a time.
The simplified grand lot procedure suggested earlier can easily be modified to cor-
respond to operating conditions. Sublots can be recombined to form new presumptive
grand lots, where justified, when the grand lot hypothesis is rejected. Risks can be altered
as appropriate and analysis of means limits computed using factors such as those given
by Nelson (1974) for Schilling’s hα at other probability levels. These may be converted to
­values of Hα by the relation

k -1
H a = ha
k

While analysis of means is relatively insensitive to nonnormality, as shown by Schilling and

Nelson (1976), special procedures are available for analysis of means when the variate departs
substantially from the normal form and are given in Schilling (1973b). Also, the assumption
of shape in a variables sampling plan can be minimized by performing the analysis of means
procedure with variables data and using attributes to inspect lots not belonging to the grand
lot, or indeed, to inspect the grand lot itself. Thus, the grand lot scheme can employ mixed
variables–attributes procedures as presented by Schilling and Dodge (1969).
Compliance Sampling 515

The grand lot approach has great potential for increasing the efficiency and economy of
acceptance sampling. The simplified graphical procedure facilitates its use in achieving the
wide application Simon intended.

Nomograph for Samples Having Zero Defectives

Nelson (1978) presented a simple nomograph for obtaining the upper confidence limit on
percent defective for a sample that contained no defectives. This nomograph relies on the
Poisson distribution to compute the upper confidence limit for a given sample size and
confidence level. The relationship is

g = 1 - e - npu

where
γ is the confidence level
n is the sample size
pu is the upper confidence limit on proportion defective

The nomograph can be found in Appendix Table F17.18. It can be used to obtain an
approximate solution for n, γ, or 100pu (given the other two).
Suppose that a sample of n pieces is taken from a lot of size N that contains more than
1000 pieces, but having no defectives. What would be the upper 90% confidence limit on
percent defective? Using the nomograph, the percent defective would be no greater than
2.3% with 90% confidence.

Accept on Zero Plans

U.S. Department of Defense Approach
It is appropriate that this chapter began with a discussion of c = 0 plans, also known as zero
defective plans, and ends with a discussion of their pros and cons. What is so attractive
about these plans that so many companies require their use? Browsing the Internet with
the simple search “c = 0 sampling plans” yields thousands of references including their
use in the food industry, pharmaceutical companies, and fisheries. Liuzza and Pap (1999)
stated that the U.S. Department of Defense (DOD) advocated that “only ‘accept on zero’
(AoZ) plans are to be used for attributes sampling.” Their argument is that many compa-
nies have worked hard to recapture markets, regain profitability, and won quality awards
using total quality management (TQM) and continuous process improvement.
In other words, “customers have the right to expect quality” so AQL-based sampling
plans that allow for nonzero nonconformities are contrary to the message that the DOD
wanted people to receive. This message was intended to encourage vendors to “moni-
tor, control, and continuously improve their processes; minimize variability; achieve high
516 Acceptance Sampling in Quality Control

capability; and prevent rather than inspect for nonconformities.” The DOD was not saying
that MIL-STD-105 did not work, but rather that it caused users to create a culture where
production is judged by the rate at which lots are accepted rather than striving for 100%
compliance. Their argument is that a vendor would typically argue for a waiver for a failed
lot as the result of a statistical aberration and avoid continuous process improvement.
The DOD initially moved to recommend that product lots be accepted when quality
levels were in parts per million (ppm). Usually, these levels were understood to be from
1 to 100 ppm. The reaction of many people, particularly those used to AQL levels in percent-
ages, interpreted this move as no conformities would be allowed for any production lots
made for the DOD. Consequently, the DOD adopted the AOZ plans from MIL-STD-1916,
which was developed to show its new view of the business process. MIL-STD-1916 states
that process control-based acceptance is the preferred method and claims that AoZ sam-
pling plans are used until the vendor can demonstrate quality in the ppm range. Note that
Cross (1984) provided a set of AOQL plans for ppm applications.
Liuzza and Pap also discuss some misconceptions regarding AoZ plans:

• Zero nonconformities in a sample imply zero nonconformities in the lot. This is incorrect
since a single nonconformity in the lot will not necessarily end up in the sample.
• AoZ sampling requires that the entire lot must be perfect. AoZ sampling requires ppm
levels, not a flawless population. In fact, if a nonconformity does appear in a sam-
ple that is not large, the quality level is probably not in the ppm range anyway.
• AoZ sampling plans are inferior because they are not as discriminating as non-AoZ plans.
This is true in the sense that AoZ plans are not AQL-based plans. In fact, the DOD
developed AoZ plans that would not utilize an AQL or LTPD. Simply stated, AoZ
plans stated that for any sample size no nonconforming unit will be allowed for
a  lot to be accepted. Furthermore, the DOD recognizes that it does not want to
compare OC curves as it does for non-AoZ plans based on AQLs.
• Use of AoZ plans for some contracts will result in too many lot rejections. According to
MIL-STD-1916, if a lot fails to pass the AoZ plan, it is not necessarily rejected, but
rather lot acceptance is withheld and the vendor is required to take several actions
as stated in the standard.

It is the intent of the DOD to encourage its vendors to pursue continuous improvement
and process controls such that AoZ plans are unnecessary. A 100% conformance is the
end goal, and MIL-STD-1916 makes it clear that any AoZ plan will not be the final word
on product acceptance. Of course, if the vendor does not improve their process then AoZ
plans will prevail and result in improved quality levels due to their fear of excessive lot
rejections. Thus, MIL-STD-1916 (and its companion document MIL-HDBK-1916 [United
States Department of Defense 1999]) puts the utmost importance on process control to
achieve quality, while AoZ plans play a secondary role. The DOD did not popularize AoZ
plans. Their development came many years ago.

Squeglia Plans
In 1961, Squeglia developed a set of c = 0 plans in an effort to create an alternative to the
widespread use of MIL-STD-105C. In 1963, MIL-STD-105D was created and the c = 0 plans
were updated and revised. The plans were based on a match of the c = 0 plan LTPD with the
LTPD of the MIL-STD-105D plan with some adjustments. The hypergeometric distribution
Compliance Sampling 517

was used. Of course, this results in the necessity for the producer to operate at quality lev-
els 1/45 of the LTPD to have 95% of the lots accepted. Thus, for the lot size (91–150), the
MIL-STD-105E plan for Code F, 2.5% AQL (n = 20, c = 1), shows an LTPD = 18.1% that
corresponds with the Squeglia AOZ plan (n = 11, c = 0) that has an LTPD = 18.7%. In 1965,
Squeglia wrote the seminal paper Sampling Plans for Zero Defects, which led to the publica-
tion of his book Zero Acceptance Number Sampling Plans in 1969. In the book, which is not a
true standard, zero defective plans are presented in detail along with their respective OC
curves. According to Squeglia, a survey conducted in 1983 indicated that the majority of
users of c = 0 plans reported an average of 18% savings. In 1989, MIL-STD-105E superseded
MIL-STD-105D.
In the fourth and later editions of his book, Squeglia (1994) presents his table of c = 0
plans along with another table for small lot sizes. These tables are shown in Tables 17.3
and 17.4, respectively. His primary argument for the use of c = 0 plans is that they result
in a lower sample size than the corresponding MIL-STD-105 plan for the same AQL. For
example, suppose a supplier is inspecting lots of size N = 2000 to general inspection level II
with a 1.0% AQL under normal inspection. MIL-STD-105E (or ANSI/ASQC Z1.4) requires
a sample size of n = 125 with c = 3 nonconformities allowed. Squeglia’s corresponding
c = 0 plan from Table 17.3 for a lot of N = 2000 and an associated AQL of 1.0% requires only
n = 42 samples to be taken with c = 0 nonconformities allowed.
Squeglia provided a special sampling plan table in his book for users who work with
small lots, and when the associated AQL they are using is 1.5% and below. This table is pre-
sented here as Table 17.4. Note that for AQLs above 1.5%, Table 17.3 is sufficient for small
lot sizes. Squeglia suggests that if the user works with a broad range of lot sizes along with
associated AQLs between 0.25% and 1.5%, the small lot sampling Table 17.4 should be used.
According to Squeglia, the DOD issued a notice in the year 2000 that authorized the use
of c = 0 plans and promoted his book as the state of the art in zero-based sampling plans.
However, based on the paper by Liuzza and Pap (1999), the recommendation was to use
a zero-based (AoZ) sampling plan only on an interim basis—not as an exclusive approach
to the control of quality.

AoZ and AQL Plans

While there are people who endorse the use of c = 0 plans, there are just as many others
who condemn their use. After all, the OC curve is always convex when c = 0, whereas
nonzero sampling plans have a shoulder near p = 0. Why is such an OC curve a concern?
While such plans are attractive to the consumer, they carry excessive risks for the producer
who must pass on these higher costs to the consumer. Furthermore, there may be delays in
shipment since there will be a large number of acceptable lots that will not pass inspection
when one nonconformity appears in the sample of an otherwise acceptable lot. As stated
previously, when no nonconformities are found, the consumer is led to the belief that the
lot is defect-free that is not necessarily true. Baker (1988) evaluated producer costs associ-
ated with Squeglia’s c = 0 plans. According to Baker, “if one were to use the Squeglia c = 0
plans, unless the true process percent defective was substantially better than the specified
AQL, total inspection cost would be significantly higher.”
Gershon and Christobek (2006) compared the cost of quality between c = 0 acceptance
plans and MIL-STD-105E plans. Whereas most people had only evaluated the effect on
producer costs, as stated earlier, these authors looked at the costs to both the producer and
consumer, that is, the total costs. These authors identified six factors that affect total costs.
These factors are given in Table 17.5.
 518

TABLE 17.3
c = 0 Sampling Plan Table
Index Values (Associated AQLs)

0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10.0

Lot Size Sample Size

2–8 * * * * * * * * * * * * 5 3 2 2
9–15 * * * * * * * * * * 13 8 5 3 2 2
16–25 * * * * * * * * * 20 13 8 5 3 2 2
26–50 * * * * * * * * 32 20 13 8 5 3 2 2
51–90 * * * * * * 80 50 32 20 13 8 7 6 5 4
91–150 * * * * * 125 80 50 32 20 13 12 11 7 6 5
151–280 * * * * 200 125 80 50 32 20 20 19 13 10 7 6
281–500 * * * 315 200 125 80 50 48 47 29 21 16 11 9 7
501–1,200 * 800 500 315 200 125 80 75 73 47 34 27 19 15 11 8
1,201–3,200 1250 800 500 315 200 125 120 116 73 53 42 35 23 18 13 9
3,201–10,000 1250 800 500 315 200 192 189 116 86 68 50 38 29 22 15 9
10,001–35,000 1250 800 500 315 300 294 189 135 108 77 60 46 35 29 15 9
35,001–150,000 1250 800 500 490 476 294 218 170 123 96 74 56 40 29 15 9
150,001–500,000 1250 800 750 715 476 345 270 200 156 119 90 64 40 29 15 9
500,001 and over 1250 1200 1112 715 556 435 303 244 189 143 102 64 40 29 15 9

Source: Squeglia, N.L., Zero Acceptance Number Sampling Plans, 4th ed., ASQ Quality Press, Milwaukee, WI, 1994.
* Indicates that the entire lot must be inspected.
Acceptance Sampling in Quality Control
Compliance Sampling 519

TABLE 17.4
Small Lot Size Supplement Table to Table 17.3
Associated AQLsa
Lot Size 0.25 0.40 0.65 1.00 1.50

5–10 b b b 8 5
11–15 b b 11 8 5
16–20 b 16 12 9 6
21–25 22 17 13 10 6
26–30 25 20 16 11 7
31–35 28 23 18 12 8

Source: Squeglia, N.L., Zero Acceptance Number Sampling Plans, 4th ed.,
ASQ Quality Press, Milwaukee, WI, 1994.
a Used for small lots when the associated AQL values are 1.5 and below.

b Indicates that the entire lot must be inspected.

TABLE 17.5
Summary of Input Parameters to Total Cost of Acceptance Sampling
Input Parameter Effect on Total Cost of Sampling Plan

AQL A lower AQL will result in a lower total cost

Sample size Increasing the sample size increases the total cost
Acceptance number Reduction of acceptance sampling number reduces total costs
Producer inspection cost Reducing the producer cost reduces the total costs
Consumer rectification cost Reducing the consumer cost reduces the total costs
Acceptance sampling plan used c = 0 acceptance sampling plans show a reduction of total cost
over the comparable MIL-STD-105E (ANSI/ASQC Z1.4)
acceptance sampling plan

Source: Reprinted from Gershon, M. and Christobek, M., Int. J. Productivity Qual. Manage., 1(3), 272, 2006.
With permission.

Gershon and Christobek computed total quality costs using the AOQ and average total
inspection (ATI) curves under the assumption of rectification, that is, rejected lots are 100%
inspected and conforming units replace all nonconforming units. The producer must know
the cost of inspecting and reinspecting a unit of product. The producer cost is determined
by multiplying this inspection/reinspection cost per unit by the ATI. The consumer must
know the cost of finding and removing a nonconforming unit. The consumer cost is deter-
mined by multiplying the cost for the receipt of a nonconforming unit by the AOQ and the
lot size. Using this approach of computing total costs, these authors conclude that despite
the lower maximum cost of the Squeglia c = 0 plan the modified shape and location of the
total cost curve are also affected by other input parameters such as the producer and con-
sumer cost. These authors did acknowledge that the c = 0 plans adequately matched the
quality of the MIL-STD-105E plans.

Chain Sampling Alternative

So, is there a compromise position between the Squeglia c = 0 plan and the correspond-
ing MIL-STD-105E plan with a nonzero acceptance number? The main concern regarding
520 Acceptance Sampling in Quality Control

the c = 0 plan is the shape of its OC curve, which shows a large producer risk particularly
for small percent defective levels, that is, low values of the AQL. The corresponding MIL-
STD-105E plan can closely match the c = 0 plan relative to the consumer risk, but has a
much lower associated producer risk. If we wish to reduce the producer risk associated
with the c = 0 plan, the answer is simple—use a chain sampling approach.
As an example, consider the supplier who is inspecting lots of size N = 2000 to general
inspection level II with a 1.0% AQL and 5.0% LTPD. MIL-STD-105E (or ANSI/ASQC
Z1.4) required a sample size of n = 125 with c = 3 nonconformities allowed. Squeglia’s
corresponding c = 0 plan from Table 17.3 for a lot of N = 2000 and an associated AQL of
1.0% required only n = 42 samples to be taken with c = 0 nonconformities allowed. Now,
suppose that a ChSP-1 plan will be used with the c = 0 plan to improve the shape of its
OC curve. The MIL-STD-105E plan of n = 125, c = 3 yields a producer risk of α = .037 and
a consumer risk of β = .124 at a producer’s quality level p1 = .01 and a consumer’s qual-
ity level of p2 = .05, respectively. Using these risks and n = 42 produces the ChSP-1 plan
of n = 42 with i = 1. Figure 17.13 shows a comparison of OC curves for the MIL-STD-
105E, c = 0, and ChSP-1 sampling plans. The ChSP-1 plan does provide a compromise
between the c = 0 and MIL-STD-105E plans relative to the producer risk. At the AQL
of 1%, the producer risks for the MIL-STD-105E, c = 0, and ChSP-1 plans are .0374, .3443,
and .1620, respectively. At an LTPD of 5%, all three sampling plans are matched at
.1238, .1160, and .1457, respectively.
Under rectification, the AOQ and ATI curves can be generated to compare these plans.
The AOQL values for the MIL-STD-105E, c = 0, and ChSP-1 plans are .0155, .0086, and
.0119, respectively, as shown in Figure 17.14. Here, it can be seen that the n = 125, c = 3 plan
does perform better than the other plans at defective levels above the LTPD.
The ATI curves for these plans are shown in Figure 17.15. Due to the higher producer
risk of rejecting acceptable lots, the c = 0 plan quickly overtakes the MIL-STD-105E plan.
Above the LTPD, the MIL-STD-105E plan overtakes the other plans but all plans are at
100% inspection for defective levels above 15%.

100%
80% n = 125, c = 3
n = 42, c = 0
Probability of accepting lot (Pa)

80%
n = 42, i = 1, c = 0
70%
60%
50%
40%
30%
20%
10%
0%
0.0000 0.0500 0.1500 0.1500 0.2000 0.2500
Fraction nonconforming in lot (p)

FIGURE 17.13
OC curves for the n = 42, c = 0 plan; n = 125, c = 3 plan; and n = 42, i = 1, c = 0 plan.
Compliance Sampling 521

1.800%

Average outgoing quality (% defective in accepted lots)

n = 125, c = 3
1.600%
AOQL = 0.0155
1.400% n = 42, c = 0
AOQL = 0.0086
1.200% n = 42, i = 1, c = 0
AOQL (ChSP-1) = 0.0119
1.000%

0.800%

0.600%

0.400%

0.200%

0.000%
0.0000 0.0500 0.1500 0.1500 0.2000 0.2500
Fraction defective in lots coming to inspection station (p)

FIGURE 17.14
AOQ curves for the n = 42, c = 0 plan; n = 125, c = 3 plan; and n = 42, i = 1, c = 0 plan.

2500

2000
Average total inspection

1500

1000
n = 125, c = 3
n = 42, c = 0
500 n = 42, i = 1, c = 0
Lot size, N = 2000

0
0.0000 0.0500 0.1000 0.1500 0.2000 0.2500
Fraction defective in lots coming to inspection station (p)

FIGURE 17.15
ATI curves for the n = 42, c = 0 plan; n = 125, c = 3 plan; and n = 42, i = 1, c = 0 plan.

The Zero Acceptance Number Chained Quick Switching System

It seems inevitable that ideas of the AoZ c = 0 plan, QSS plan, and chain sampling plan
should come together. Govindaraju (2011) presented a ChQSS as a means of chaining
past lot results together within a quick switching system to improve the discrimination
power of a zero acceptance number plan. Govindaraju (1991) had studied the design
522 Acceptance Sampling in Quality Control

of a quick switching system using two zero acceptance number single-sampling plans
designated as QSS-1(nN, nT, 0).
In this QSS-1 system, a c = 0 plan with sample size of nN is used for normal inspection,
that is, (nN, 0), and a larger sample size nT is used for tightened inspection, that is, (nT, 0).
As long as lots continue to be accepted under the (nN, 0) plan, normal inspection continues
with the next lot. Upon the first rejection, the plan changes to the (nT, 0) plan under tight-
ened inspection.
Utilizing the ChSP-1 plan, a single nonconformity is considered as an occasional one when
the preceding i successive samples are free of nonconforming units. The QSS-1(nN, nT, 0)
system does not allow for a single nonconformity to occur after a succession of i preceding
accepted lots. Fortunately, it is possible to employ such a rule during the normal inspec-
tion stage of the QSS-1(nN, nT, 0) system. This results in the chained quick switching system
with zero acceptance numbers called the ChQSS-1(nN, nT, i) system, which is shown in
Figure 17.16.
Govindaraju (2011) noted that the ChQSS-1(nN, nT, i) system is not exactly equivalent to
a QSS-1 system using the ChSP-1 plan for the normal level of inspection and the c = 0 plan
for the tightened level of inspection. Why? Because the clearing interval i may include the
result of the last lot inspected under the tightened inspection plan (nT, 0) just prior to the
change back to normal inspection. This will affect the plan’s transitive OC curve. The idea
of a transitive OC curve was first discussed by Taylor (1996), who used it to quantify the
protection provided by the plan during periods of changing quality.
The steady-state OC curve of the ChQSS-1(nN, nT, i) system is the same as the QSS-1
system having the (nT, 0) plan for tightened inspection and the ChSP-1 plan for the normal
level of inspection. The OC function of the tightened inspection level (nT, 0) plan under the
binomial distribution is

PT ( p ) = ( 1 - p )
nT

where PT is the probability of accepting the lot, nT is the sample size under tightened
inspection, and p is the fraction nonconforming in the incoming lots. This probability of lot
acceptance under the Poisson distribution is

PT ( p ) = e - xT

where xT = nTp, and is valid when nTp < 5. This probability is particularly well suited when
the quality measure is the number of independent nonconformities.
In the case of nonconforming items, the OC function of the ChSP-1(nN, i) plan for the
­normal level of inspection under the binomial distribution was shown by Dodge (1955) to be

PN ( p ) = ( 1 - p ) + nN p ( 1 - p )
nN nN + nN i - 1

For the case of the number of nonconformities under the Poisson distribution, the corre-
sponding OC function is

PN ( p ) = e - xN + x N e
- xN ( i + 1)

where xN = nNp.
Compliance Sampling 523

Start

Tightened plan
Sample nT items and observe the
number of nonconforming items dT

Reject the current lot and

No continue to apply the
dT = 0?
tightened plan for the
next lot
Yes

Accept the current lot

and invoke the normal
plan for the next lot

Normal plan
Sample nN items and observe the
number of nonconforming items dN

Accept the current lot and

Yes
continue to apply the dN = 0?
normal plan for the next lot

No

Yes Reject the current lot and

dN > 1? invoke the tightened plan
for the next lot
No

No Was dN or dT > 0 in any of Yes

the i preceding samples?

FIGURE 17.16
Operation of the ChQSS-1(nN, nT, i) chained quick switching system with zero acceptance numbers plan.

The expected number of lots under tightened inspection under the QSS-1 system before
a switch to normal inspection is given by Romboski (1969) to be
1
cT ( p ) =
PT ( p )

and the expected number of lots under normal inspection before a switch to tightened
inspection is given by
1
cN ( p ) =
1 - PN ( p )
524 Acceptance Sampling in Quality Control

These formulas for cN and cT lead to the steady-state OC function of the QSS-1 system
shown by Romboski (1969) as

cT ( p ) PT ( p ) + cN ( p ) PN ( p ) PT ( p )
Pa ( p ) = =
cT ( p ) + cN ( p ) 1 - PN ( p ) + PT ( p )

The average sample number (ASN) of the QSS-1 system was shown by Govindaraju (2011)
to be

ASN ( p ) =
( )
cT ( p ) nT + cN ( p ) nN nT 1 - PN ( p ) + nN PT ( p )
=
cT ( p ) + cN ( p ) 1 - PN ( p ) + PT ( p )

Using the formulas based on the Poisson distribution for the probabilities PT and PN,
we have the steady-state OC function of the ChQSS-1(nN, nT, i) system shown as

e - xT
Pa ( p ) =
- rxT e T ( ) + e - xT
- rx i + 1
1 - e - rxT

where r = nN/nT. Govindaraju (2011) shows that the steady-state OC curve of the ChQSS-
1(nN, nT, i) system is closer to the OC curve of the ChSP-1(nN, i) plan at good quality levels
but shifts toward the OC curve of the tightened inspection plan (nT, 0) when incoming
quality deteriorates. When quality is good, the ChQSS-1 system uses a smaller sample size
nN and behaves like a ChSP-1 plan in allowing only occasional nonconforming units. When
quality is poor, the ChQSS-1 system uses a larger sample size nT and behaves like an AoZ
plan in not allowing any nonconforming units.
While it may appear that the ChQSS-1(nN, nT, i) system is always preferable, it is not as
advantageous at the moderate quality levels between the AQL and LTPD because the OC
curve of the system lies between the OC curves of the normal and tightened plans. This
will be demonstrated with an example in the software section at the end of this chapter.
It is interesting to note that when i = 0, the ChSP-1(nN, 0) plan is equivalent to a c = 1 single-
sampling plan. Thus, the ChQSS-1(nN, nT, 0) system is a QSS-1 system having a single-sampling
plan (nN, 1) for the normal inspection level and a single-sampling plan (nT, 0) at the tightened
inspection level. So, when 1 approaches infinity, the ChSP-1 plan becomes the c = 0 (AoZ)
single-sampling plan, and the ChQSS-1(nN, nT, ∞) system becomes the QSS-1(nN, nT, 0) system.

Summary
The strength of c = 0 plans is in their small sample size. In an increasingly litigacious society,
jury members may not understand the deceptive logic of AoZ plans. No defect in the sam-
ple does not imply that there are no defects in the lot. A defect in the sample does not mean
the lot must be rejected—it depends upon the sample size. The quality engineer would do
well to check with the OC curves to devise a plan appropriate to the problem at hand.
It has been shown in this chapter that the reader has some options other than the AoZ
plan. TNT and QSS plans can be effective at reducing the producer risk of the AoZ plan
and improve its discriminatory power. The use of chain sampling in the form of a ChQSS
plan can be a more attractive option if its additional complexity can be managed.
The reader is urged to study the plans discussed in this chapter to make an informed
decision on the type of compliance sampling plan that best fits their situation. Use
of ­software is encouraged to evaluate these plans to determine which ones offer the
Compliance Sampling 525

desired protection to both the producer and consumer. In the next section, software
options, including several Excel templates, are discussed.

Software Applications
Excel
LSP Plans
The Excel template LSP Plan.xlsm can be used to generate and evaluate an LSP. To illustrate its
use, suppose we desire an LSP plan based on the earlier example of a part received at incom-
ing inspection in lots of 100 items. Protection against a fraction defective of 10% is desired.
Enter the lot size N of 100 into cell B3 and the limiting quality level of 0.10 into cell B5. Click on
the “Find LSP Plan” button to generate the plan as shown in Figure 17.17. The OC and AOQ
curves for the LSP plan can be found on the “OC Curve” and “AOQ curve” tabs, respectively.

TNT Plans
The Excel template TNT Plan.xlsm can be used to generate and/or evaluate a TNT sampling
plan for the special case of k = 2 using the tab entitled “Design TNT Plan—Calcs.” As an
example of its use, suppose it is desired to design a TNT plan having p1 = .005 (cell C17) and
p2 = .11736 (cell C18) for an operating ratio of R = p2/p1 = .11736/.005 = 23.472 (cell C19).
Click on the “Generate TNT Plan” button to generate the plan as shown in Figure 17.18. The
solution is n1 = 20, n2 = 10, t = 5, and s = 4. The OC, AOQ (assuming a lot size N is entered

FIGURE 17.17
Excel template LSP Plan.xlsm for generating a lot sensitive sampling plan.
 526

FIGURE 17.18
Excel template TNT Plan.xlsm used to generate a TNT sampling plan.
Acceptance Sampling in Quality Control
Compliance Sampling 527

into cell C6), and ASN curves are shown on subsequent tabs. This Excel template can also
evaluate a given TNT plan using the “Specify TNT Plan—Calcs” tab (see Figure 17.19) or
evaluate multiple TNT plans using the “Evaluate Multiple TNT Plans” tab. Again, the OC,
AOQ (assuming a lot size N is entered), and ASN curves are shown on subsequent tabs.

QSS Plans
The Excel template QSS-1(n; cN; cT) Plan.xlsm can be used to generate and/or evaluate
a Romboski QSS-1 sampling plan based on specifying p1 and p2 (“Design QSS-1(n; cN; cT)
Plan (1)” tab), or by specifying n, cN, and cT (“Design QSS-1(n; cN; cT) Plan (2)” tab). Using
the “Design QSS-1(n; cN; cT) Plan (1)” tab, suppose that we want to generate a QSS-1 plan
assuming that p1 = 0.015 (cell C17) and p2 = 0.126 (cell C18) for an operating ratio of 8.4 (cell
C19). Click on the “Generate Romboski QSS Sampling Plan based on PQL and CQL” button
to get the solution of n = 21 with cN = 1 and cT = 0 as shown in Figure 17.20. The OC and AOQ
curves are given on subsequent tabs. Using the “Design QSS-1(n; cN; cT) Plan (2)” tab, we
can evaluate a given QSS-1 plan by entering the lot size N, if known, into cell C6, the sample
size n into cell C8, and the acceptance numbers cN and cT into cells E8 and E9, respectively,
as shown in Figure 17.21. The OC and AOQ curves are given on subsequent tabs. The Excel
template also allows a user to evaluate multiple QSS-1 plans on the tab entitled “Evaluate
Multiple QSS-1 Plans.” The OC and AOQ curves are given on subsequent tabs.
The Excel template QSS-r (n; cN; cT) Plan.xlsm can be used to generate and/or evaluate
Romboski QSS-1/-2/-3/-4 sampling plans based on specifying p1 and p2 (“Generate Mult
QSS-r Plans (1)” tab), or by specifying n, cN, and cT for each plan (“Evaluate Mult QSS-r
Plans (2)” tab). Using the “Generate Mult QSS-r Plans (1)” tab, suppose we wish to gener-
ate the corresponding QSS-1/-2/-3/-4 plans for p1 = 0.020 (cell C27) and p2 = 0.050 (cell
C28) for an operating ratio of 2.5 (cell C29). Click the “Generate QSS-r Sampling Plans
based on PQL and CQL” button to obtain the solution of QSS-1 (n = 193, cN = 7, cT = 5),
QSS-2 (n = 108, cN = 5, cT = 2), QSS-3 (n = 184, cN = 7, cT = 5), and QSS-4 (n = 77, cN = 5, cT = 1)
as shown in Figure 17.22. The OC and AOQ curves are given on subsequent tabs.
Using the “Evaluate Mult QSS-r Plans (2)” tab, we can evaluate a given series of QSS plans.
Suppose that it is of interest to evaluate the QSS-1/-2/-3/-4 plans with n = 100 and cN = 5 for
cT values of 4, 3, 2 and 1, respectively. Figure 17.23 shows the template for entry of the lot size
N, if known, into cell C7; n = 100 into cells C9, C12, C15, and C18; cN = 5 into cells E9, E12, E15,
and E18; and values of cT = 4, 3, 2, and 1 into cells E10, E13, E16, and E19, respectively. Once
the values are entered, the corresponding probabilities are computed, and the OC and AOQ
curves are given on subsequent tabs. Figure 17.24 shows the OC curves for this example.
The Excel template QSS-r (n; kn; c0) Plan.xlsm can be used to generate and/or evaluate
Soundararajan and Arumainayagam (1990) QSS-1/-2/-3/-4 sampling plans based on speci-
fying p1 and p2 (“Design QSS-r (n; kn; c0) Plan” tab). An example from this paper supposes
that with p1 = 0.01, α = 0.05, p2 = 0.04, and β = 0.10, what would be the QSS-4 (n; cN, cT) plan?
Using this template, we must compute the operating ratio, which is 4.0 for this example. The
Soundararajan and Arumainayagam unity values assume that α = 0.05 and β = 0.10, so we
enter p1 = 0.01 into cell C27 and p2 = 0.04 into cell C28 as shown in Figure 17.25. Click on the
“Evaluate Soundararajan and Arumainayagam QSS-r (n, kn, c0) Sampling Plan based on p1
and p2 for α = 0.05 and β = 0.10” button to generate the QSS-1/-2/-3/-4 sampling plans. In the
paper, the solution for the QSS-4 plan is n = 130, kn = 163, and c0 = 3. The template solution for
this plan is n = 130, kn = 162, and c0 = 3. The OC and AOQ curves are given on subsequent tabs.
The Excel template QSS-1(n; kn; 0) Plan.xlsm can be used to generate and/or evaluate a
Govindaraju (1991) QSS-1 sampling plan based on specifying p1 and p2 (“Design QSS-1(n; kn; 0)
 528

FIGURE 17.19
Excel template TNT Plan.xlsm used to evaluate a TNT sampling plan.
Acceptance Sampling in Quality Control
 Compliance Sampling

FIGURE 17.20
Excel template QSS-1(n; cN; cT) Plan.xlsm used to generate and evaluate a QSS-1 sampling plan where p1 and p2 are specified.
529
 530

FIGURE 17.21
Excel template QSS-1(n; cN; cT) Plan.xlsm used to evaluate a QSS-1 sampling plan where n, cN, and cT are specified.
Acceptance Sampling in Quality Control
 Compliance Sampling

FIGURE 17.22
Excel template QSS-r (n; cN; cT) Plan.xlsm used to generate and evaluate QSS-r sampling plans where p1 and p2 are specified.
531
 532

FIGURE 17.23
Excel template QSS-r (n; cN; cT) Plan.xlsm used to evaluate a series of QSS-1/-2/-3/-4 sampling plans where n, cN, and cT are specified for each plan.
Acceptance Sampling in Quality Control
Compliance Sampling 533

Operating characteristic (OC) curve for multiple QSS–r (n; cN; cT) sampling plans
Assumed distribution = Binomial
100%
Pa, Scheme Pa of QSS-1 (100;5;4) plan
Pa, Scheme Pa of QSS-2 (100;5;3) plan
90%
Pa, Scheme Pa of QSS-3 (100;5;2) plan
Pa, Scheme Pa of QSS-4 (100;5;1) plan
80%
Probability of lot acceptance, Pa

70%

60%

50%

40%

30%

20%

10%

0%
0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070 0.080 0.090 0.100
Proportion defective in lot, p

FIGURE 17.24
OC curves from Excel template QSS-r (n; cN; cT) Plan.xlsm used to evaluate a series of QSS-1/-2/-3/-4 sampling
plans where n = 100, cN = 5, and cT = 1, 2, 3, and 4 specified for each plan.

Plan” tab) for the special case where c0 = 0. An example from this paper supposes that with
p1 = 0.001, α = 0.05, p2 = 0.019, and β = 0.10, what would be the QSS-1(n; kn; 0) plan? Using
this template, we must compute the operating ratio, which is 19.0 for this example. We
enter p1 = 0.001 into cell C17 and p2 = 0.019 into cell C18 as shown in Figure 17.26. Click on
the “Evaluate Govindaraju QSS-1(n, kn, 0) Sampling Plan based on p1 and p2 for α = 0.05
and β = 0.10” button to generate the QSS-1 sampling plan. In the paper, the solution for
the QSS-1 plan is n = 47, kn = 140, and c0 = 0. The template solution for this plan is n = 47,
kn = 140, and c0 = 0. The OC and AOQ curves are given on subsequent tabs.
The Excel template QSS-r (n; kn; 0) Plan.xlsm can be used to generate and/or evalu-
ate Soundararajan and Arumainayagam (1990) QSS-1/-2/-3/-4 sampling plans based on
specifying p1 and p2 (“Design QSS-r (n; kn; 0) Plan” tab) for the special case, where c0 = 0.
Another example from the Soundararajan and Arumainayagam (1990) paper supposes
that with p1 = 0.005, α = 0.05, p2 = 0.08, and β = 0.10, what would be the QSS-2 (n; kn; 0) plan?
Using this template, we must compute the operating ratio, which is 16.0 for this example.
We enter p1 = 0.005 into cell C27 and p2 = 0.08 into cell C28 as shown in Figure 17.27.
Click on the “Evaluate Soundararajan and Arumainayagam QSS-r (n, kn, 0) Sampling Plan
based on p1 and p2 for α = 0.05 and β = 0.10” button to generate the QSS-1/-2/-3/-4 sam-
pling plans. In the paper, the solution for the QSS-2 plan is n = 8, kn = 28, and c0 = 0. The
template solution for this plan is n = 7, kn = 29, and c0 = 0. The OC and AOQ curves are
given on subsequent tabs.
 534

FIGURE 17.25
Excel template QSS-r (n; kn; c0) Plan.xlsm used to generate and evaluate QSS-r sampling plans where p1 and p2 are specified for α = 0.05 and β = 0.10.
Acceptance Sampling in Quality Control
 Compliance Sampling

FIGURE 17.26
Excel template QSS-1(n; kn; 0) Plan.xlsm used to generate and evaluate a QSS-1 sampling plan where p1 and p2 are specified for α = 0.05 and β = 0.10.
535
 536

FIGURE 17.27
Acceptance Sampling in Quality Control

Excel template QSS-r (n; kn; 0) Plan.xlsm used to generate and evaluate QSS-r sampling plans where p1 and p2 are specified for α = 0.05 and β = 0.10.
Compliance Sampling 537

MIL-STD-1916 Plans
The Excel template Mil-Std-1916 Sampling Plan.xlsm can be used to generate the various
MIL-STD-1916 sampling plans based on a given lot size N and verification level (VL).
Revisiting the earlier example, assume as in our discussion of MIL-STD-105E that a lot is
presented having a lot size of 500 and (lacking a specified value) VL of IV, then the default
VL is to be used. Table I of the standard gives code A. For this example, we have developed
a matched set of three schemes:

Attributes Variables Continuous

Tightened na(T) = 192, c = 0 nV(T) = 44, k = 2.69, F = .174 f = 2/17, i = 527

Normal na(N) = 80, c = 0 nV(N) = 29, k = 2.40, F = .193 f = 1/12, i = 264
Reduced na(R) = 32, c = 0 nV(R) = 18, k = 2.05, F = .222 f = 1/17, i = 125

Using the Excel template, we enter the lot size N into cell C3 and the verification level of
IV into cell G3. Click on the “Generate MIL-STD-1916 Sampling Plans” button to generate
the plans in the table presented earlier as shown in Figure 17.28. The result agrees with the

FIGURE 17.28
Excel template Mil-Std-1916 Sampling Plan.xlsm used to generate MIL-STD-1916 sampling plans where N and the
verification level are specified.
538 Acceptance Sampling in Quality Control

plans shown in the preceding table. OC, AOQ, and AFI curves are available for the attri-
bute, variable, and continuous sampling plans on separate tabs.

Chain Sampling Alternative Plans

The Excel template Chain Sampling Plan Alternative.xlsx can be used to compare an MIL-
STD-105E single-sampling plan, AoZ sampling plan, and a chain sampling plan for a
specified p1 (AQL), p2 (LTPD), and lot size N. As an example, we can use this template
to re-create the figures seen in this chapter. Enter the MIL-STD-105E sampling plan by
entering 125 for the sample size n into cell B8 and c into cell B9 as shown in Figure 17.29.
Enter the AoZ sampling plan by entering 42 for n into cell B14. The chain sampling plan
is found by entering p1 = 0.01 and p2 = 0.05 into cells B26 and B27, respectively. On a
separate tab entitled “ChSP-1 Calculations,” the AoZ sample size n (cell B14) along with
the p1 and p2 values produces an optimal value of i = 1, which appears in cell B22. The
lot size N = 2000 is entered into cell B28. The OC, AOQ, and ATI curves are available on
separate tabs.

Zero Acceptance Number Chained Quick Switching System Plans

The Excel template ChQSS-1 Plan.xlsm can be used to generate and evaluate a ChQSS-1
zero acceptance number chained quick switching system set of plans based on p1 (AQL)
and p2 (LTPD) and unity values by Govindaraju (2011) for α = 0.05 and β = 0.10. As an
example, suppose that p1 = 1% and p2 = 4% with α = 0.05 and β = 0.10. Using the “Design
ChQSS-1(nN; nT; i) Plan” tab of the template as shown in Figure 17.30, we enter .01 into
cell C17 for p1 and 0.04 into cell C18 for p2 yielding an OR = 4.0. Click on the “Generate
Govindaraju ChQSS-1 Sampling Plan based on p1 and p2 for α = 0.05 and β = 0.10” button
to generate the plan. The result is a ChQSS-1(9; 92; 3) plan that achieves a producer risk
of 0.058 (cell E30) at p1 = 1% and a consumer’s risk of 0.10 (cell E31) at p2 = 4% using the
Poisson distribution as assumed by Govindaraju (2011).
This Excel template can also be used to evaluate a given plan using the “Evaluate ChQSS-
1(nN; nT; i) Plan” tab. Govindaraju (2011) evaluated the ChQSS-1(20; 80; 3) plan along with
the (80, 0) single-sampling plan for the tightened inspection level and the ChSP-1 (20, 3)
plan for the normal inspection level. Using this template, enter the ChSP-1 plan sample
size nN = 20 into cell C8 (the acceptance number cN = 0 is already provided in cell E8) and
i = 3 into cell G8 as shown in Figure 17.31. For the AoZ tightened inspection plan, enter
the sample size nT = 80 into cell C9 (the acceptance number cT = 0 is already provided in
cell E9).
As Govindaraju (2011) points out, it is clear that the ChQSS-1(20; 80; 3) system is closer
to the OC curve of the ChSP-1 (20; 3) plan at good quality levels and shifts toward the OC
curve of the AoZ tightened plan at the poor quality levels when the incoming quality of
lots deteriorates. The OC curves of these plans are given in Figure 17.32.
When incoming quality is good, the ChQSS-1 system uses a smaller sample size nN and
thus behaves more like a ChSP-1 plan allowing only occasional nonconforming units.
When incoming quality of lots deteriorates, the ChQSS-1 system uses a larger sample size
nT making it behave more like the AoZ plan disallowing any nonconforming units. The
ASN curve for the ChQSS-1 plan is shown in Figure 17.33.
 Compliance Sampling

FIGURE 17.29
Excel template Chain Sampling Plan Alternative.xlsx used to compare an MIL-STD-105E single-sampling plan (125; 3), AoZ sampling plan (42; 0), and a chain sampling
plan (42; 1) for a specified p1 = 0.01, p2 = 0.05, and lot size N = 2000.
539
 540

FIGURE 17.30
Excel template ChQSS-1 Plan.xlsm used to generate and evaluate a ChQSS-1 zero acceptance number chained quick switching system set of plans based on p1 = 0.01
and p2 = 0.04 for α = 0.05 and β = 0.10.
Acceptance Sampling in Quality Control
 Compliance Sampling

FIGURE 17.31
Excel template ChQSS-1 Plan.xlsm used to evaluate the ChQSS-1(20; 80; 3) system against the ChSP-1 (20; 3) and AoZ (80; 0) plans.
541
542 Acceptance Sampling in Quality Control

Operating characteristic (OC) curve for ChQSS–1 (20;80;3) plan

Assumed distribution = Poisson
100%
PN, Pa of ChSP–1 normal plan (nN = 20, cN = 0, i = 3) Sampling
PT, Pa of AoZ tightened plan (nT = 80, cT = 0) plan:
90% N = 1000
Pa, Scheme Pa of ChQSS-1 (20;80;3) plan
nN = 20
80% i=3
cN = 0
Probability of lot a cceptance (Pv)

nT = 80
70% cT = 0

60%

50%

40%

30%

20%

10%

0%
0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800 0.0900 0.1000
Proportion defective in lot (p)

FIGURE 17.32
The OC curves of the ChQSS-1(20; 80; 3) system, ChSP-1 (20; 3), and AoZ (80, 0) plans using the Excel template
ChQSS-1 Plan.xlsm.

Average sample number (ASN) curve for ChQSS-1 (20;80;3) plan

Assumed distribution = Poisson
90

80
Average sample number inspected (ASN)

70

60

50

40
ASN
Sampling
30 plan:
N = 1000
nN = 20
20
i=3
cN = 0
10 nT = 8 0
cT = 0
0
0.0000 0.0100 0.0200 0.0300 0.0400 0.0500 0.0600 0.0700 0.0800 0.0900 0.1000
Proportion defective in lot (p)

FIGURE 17.33
The ASN curve of the ChQSS-1(20; 80; 3) system using the Excel template ChQSS-1 Plan.xlsm.
Compliance Sampling 543

Minitab
It is possible to generate an AoZ plan in Minitab though not by specifying an acceptance
number of 0. Choose Stat > Quality Tools > Acceptance Sampling by Attributes to open the
dialog box. As an example, assume that the quality level is in percent defective, the RQL is
0.5%, the consumer risk, β, is 0.10, and the lot size is N = 5000. Since the AQL is not mean-
ingful for an AoZ plan, it is simply set to a small value, such as 0.1. The producer risk, α, is
set to a value nearly equal to 1-β = 0.90, or 0.89. Select “Create a Sampling Plan” from the
topmost dropdown box and Go/no-go (defective) from the Measurement type dropdown
box. The completed dialog box is shown in Figure 17.34. Click on Options and check the
box to use the hypergeometric distribution for isolated lot and click OK. Click OK to gener-
ate the plan and associated curves (Figure 17.35).

Acceptance Sampling by Attributes

Measurement type: Go/no-go
Lot quality in percent defective
Lot size: 5000
Use hypergeometric distribution to calculate probability of acceptance
Acceptable Quality Level (AQL) 0.1
Producer’s Risk (α) 0.89
Rejectable Quality Level (RQL or LTPD) 0.5
Consumer’s Risk (β) 0.1
Generated Plan(s)
Sample Size 439
Acceptance Number 0

FIGURE 17.34
Minitab dialog box to generate an AoZ plan.
544 Acceptance Sampling in Quality Control

Operating characteristic (OC) curve Average outgoing quality (AOQ) curve

0.08

AOQ (percent defective)

1.0
0.06

0.04
0.8
0.02
Probability of acceptance

0.00
0.6 0.00 0.25 0.50 0.75 1.00
(b) Incoming lot percent defective
Average total inspection (ATI) curve
0.4
5000

Average total inspection

4000
0.2 3000

2000

0.0 1000

0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
Lot percent defective (c) Lot percent defective
(a) Sample size = 460, Acceptance number = 0

FIGURE 17.35
Minitab (a) OC, (b) AOQ, and (c) ATI curves for the AoZ plan.

Accept lot if defective items in 439 sampled ≤ 0; Otherwise reject.

 Percent Probability Probability
Defective Accepting Rejecting AOQ ATI
 0.1 0.631 0.369 0.058 2119.8
 0.5 0.100 0.900 0.046 4544.2
Average outgoing quality limit (AOQL) = 0.073 at 0.220 percent defective.

The AoZ plan states that a random sample of 439 units needs to be inspected and no
­nonconforming units are allowed in order to pass the lot. Otherwise, the lot is rejected.

Snap Sampling Plans!

This program offers two options for compliance acceptance sampling plans—Schilling
zero acceptance (LSP) plans and Squeglia zero acceptance plans (per his book).

LSP Plans
Suppose the lot size is N = 100 and the AQL is 10%. We can generate a Schilling sam-
pling plan from Snap Sampling Plans! by clicking the Schilling Zero Acceptance button
from the main menu and completing the dialog box as shown in Figure 17.36 using the
lot size and AQL in %. Click the Sample Plan button to generate the plan. The result is
n = 22, c = 0.
Compliance Sampling 545

FIGURE 17.36
Snap Sampling Plans! selection of a Schilling zero acceptance plan.

FIGURE 17.37
Snap Sampling Plans! selection of a Squeglia zero acceptance plan.

Squeglia Plans
As an example, suppose lots are of size N = 2000 and the AQL (p1) is 1%. The Squeglia
c = 0 plan is seen in Table 17.3 to be n = 42. We can generate a Squeglia sampling plan
from Snap Sampling Plans! by clicking the Squeglia Zero Acceptance button from the
main menu and completing the dialog box as shown in Figure 17.37 using normal
inspection as the sampling level. Click the Sample Plan button to generate the plan. The
result is n = 42, c = 0.

Statgraphics
MIL-STD-1916 Plans
Statgraphics can generate an MIL-STD-1916 sampling plan by choosing from the main
menu SPC > Acceptance Sampling > Attributes > MIL-STD-1916. As in our earlier exam-
ple, suppose that a lot is presented that has a lot size of 100 and (lacking a specified
value) verification level IV, the default VL to be used. Table I of the standard gives code A.
For this example, we have developed a matched set of three schemes:

Attributes Variables Continuous

Tightened na(T) = 192, c = 0 nV(T) = 44, k = 2.69, F = .174 f = 2/17, i = 527

Normal na(N) = 80, c = 0 nV(N) = 29, k = 2.40, F = .193 f = 1/12, i = 264
Reduced na(R) = 32, c = 0 nV(R) = 18, k = 2.05, F = .222 f = 1/17, i = 125
546 Acceptance Sampling in Quality Control

FIGURE 17.38
Statgraphics selection of MIL-STD-1916 attributes acceptance plans.

The dialog box is completed as shown in Figure 17.38. Clicking OK produces the attribute
and continuous sampling plans that agree with the previous table along with an OC curve.

MIL-STD-1916 (Attributes)
Lot size: 2–170
Verification Level: IV
Type of inspection: Normal
Sampling plan: Single (code A)

Sample Size Acceptance Number Rejection Number

80 0 1

Alternative Continuous Sampling Plan

Units
Screening phase clearance number 264
Sampling phase sampling fraction 1/12

The StatAdvisor
The derived sampling plan has a single stage. Take a sample of 80 units. If the number of
nonconforming units is no more than 0, accept the lot. If the number of nonconforming
units is 1 or more, reject the lot.
A continuous sampling plan is also available. It consists of a screening phase in which
100% inspection is applied to the first 264 units. If all units in the screening phase are
acceptable, a sampling phase is initiated in which 1 of every 12 units are inspected. If any
units inspected during the sampling phase are not acceptable or other conditions occur,
the screening phase must be repeated.
In order to generate the MIL-STD-1916 variables plan, we must choose from the main
menu SPC > Acceptance Sampling > Variables > MIL-STD-1916. Using the lot size,
Compliance Sampling 547

verification level, and type of inspection used to generate the attribute plans, complete
the dialog box as done before (the dialog box looks the same). Click OK in this dialog box
and the next one to generate the variables plan with its associated OC curve.

MIL-STD-1916 (Variables)
Lot size: 2–170
Verification Level: IV
Type of inspection: Normal
Sampling plan: (code A)

Sample size n = 29
Minimum allowable k-criterion k = 2.400
Maximum allowable F-criterion (double-sided only) F = 0.193

The StatAdvisor
The derived sampling plan requires taking a sample of 29 units. From measurements
made on each unit, the mean and standard deviation are calculated. The program then
computes quality indices that indicate how well the process behaves with respect to the
specification limits. If the calculated quality indices meet the criteria of the plan and all
measurements are within the specification limits, the lot is accepted.
Statgraphics confirms the sampling plans shown in the preceding table.

Problems
1. A lot of 15,000 units is to be tested for a potential safety hazard. An LTPD of .065% is
to be used. Construct an LSP. What is its AOQL?
2. Draw the OC curve of the LSP plan in Problem 1.
3. If the level of .065% defective is to be guaranteed with 95% probability, what should
be the sample size in Problem 1?
4. A series of lots is to be inspected using a TNT plan and it is desired that p.95 = .01 and
p.10 = .08. Construct the plan.
5. Compute the probability of acceptance, the ASN, and the AOQ when defects are
replaced for the TNT plan in Problem 4 at p = .05 when lots are very large.
6. Derive a QSS-1 plan matching the criteria of Problem 4. What is its indifference
quality?
7. Product is sold in lots of 10,000. It is necessary to verify that the quality is bet-
ter than .005 proportion defective so that LTPD = 0.5%. However, when p = 0.1%
defective, 95% of the lots should be accepted. These criteria would require use of
the single-sampling plan n = 1366, c = 3. Construct a simplified grand lot plan to
achieve these ends.
548 Acceptance Sampling in Quality Control

8. Samples of 9 units have been taken from each of 15 lots giving X = 500 and ŝ = 45.
Construct extreme limits and maverick limits to test if the 15 lots constitute a grand
lot. Assuming all sublots fall within the limits, test the grand lot against a lower speci-
fication limit of 400 using the plan n = 135, k = 2.0.
9. What should be the limits for a test of the grand lot hypothesis on a continuing source
of supply from the vendor in Problem 8?
10. Labels on bottles for a government installation are inspected for missing informa-
tion. A VL of V (VL-V) has been specified. During the month of August, the producer
makes varying lot sizes of 500, 1,200, 4,500, 12,000, 2,100, 7,500, 750, 3,200, 8,900,
10,000, and 25,000 bottles and chooses to use attributes sampling. For each lot, specify
the code letter to be used and sample size. If the number of nonconforming bottles for
these lots (in the previous sequence) was 0, 1, 0, 0, 0, 0, 0, 3, 1, 0, and 0, and then give
the corresponding lot disposition and stage of inspection (tightened, normal, and
reduced) for each lot.

References
American National Standards Institute, 2005, Combined Accept Zero Sampling System and Process
Control Procedures for Product Acceptance, ISO 21247, American National Standards Institute,
New York.
American Society for Testing and Materials, Standard Guide for Preferred Methods for Acceptance of
Product, ASTM Standards E2910, Vol. 14.02, West Conshohocken, PA.
Baker, R. C., 1988, Zero acceptance sampling plans: Expected cost increases, Quality Progress, 21(1):
43–46.
Burr, I. W., 1953, Engineering Statistics and Quality Control, McGraw-Hill, New York.
Calvin, T. W., 1977, TNT zero acceptance number sampling, in American Society for Quality Control
Thirty-First Annual Technical Conference Transactions, American Society for Quality Control,
Philadelphia, PA, p. 37.
Cross, R., 1984, Parts per million AOQL sampling plans, Quality Progress, 17(11): 28–34.
Dodge, H. F., 1955, Chain sampling inspection plan, Industrial Quality Control, 11(4): 10–13.
Dodge, H. F., 1967, A new dual system of acceptance sampling, Technical Report No. 16, The Statistics
Center, Rutgers—The State University, New Brunswick, NJ.
Duncan, A. J., 1977, Addendum to proposed standard for small lot sampling plans based on the
hypergeometric probability distribution, The Relia-Com Review, 2(2): 6.
Duncan, A. J., 1979, In the federal arena: E-11 and regulatory agencies, ASTM Standardization News,
7(1): 20–21, 39.
Gershon, M. and M. Christobek, 2006, Comparing the cost of quality between c = 0 acceptance plans
and MIL-STD-105E plans, The International Journal of Productivity and Quality Management, 1(3):
272–289.
Govindaraju, K., 1991, Procedures and tables for selection of zero acceptance number quick
­switching system for compliance sampling, Communication in Statistics—Theory and Methods,
20(1): 151–171.
Govindaraju, K., 2011, Zero acceptance number chained quick switching system, Communication in
Statistics—Theory and Methods, 40(12): 2104–2116.
Halperin, M., S. W. Greenhouse, J. Cornfield, and J. Zalokar, 1955, Tables of percentage points for the
studentized maximum absolute deviate in normal samples, Journal of the American Statistical
Association, 50(269): 185–195.
Compliance Sampling 549

Hamaker, H. C., 1950, Lot inspection by sampling, Philips Technical Review, 11: 176–182.
Hawkes, C. J., 1979, Curves for sample size determination in lot sensitive sampling plans, Journal of
Quality Technology, 11(4): 205–210.
Kuralmani, V. and K. Govindaraju, 1992, Modified tables for the selection of quick switching systems
for values of the given acceptable quality level and limiting quality level, Journal of Applied
Statistics, 19(2): 263–271.
Liuzza, C. and G. M. Pap, 1999, Accept-on-zero sampling plans improve quality at the DOD, Quality
Digest, August, 43–45.
Mandelson, J., 1963, Use of the grand lot in surveillance, Industrial Quality Control, 19(8): 10–12.
Muehlhause, C. O., V. L. Broussalian, A. J. Farrar, J. W. Lyons, M. G. Natrella, J. R. Rosenblatt,
R. D. Stiehler, and J. H. Winger, 1975, Considerations in the use of sampling plans for effect-
ing compliance with mandatory safety standards, United States Department of Commerce,
National Bureau of Standards, Washington, D.C., Report 75-697, pp. 42–43.
Nelson, L. S., 1974, Factors for the analysis of means, Journal of Quality Technology, 6(4): 175–181.
Nelson, L. S., 1978, Nomograph for samples having zero defectives, Journal of Quality Technology,
10(1): 42–43.
Ott, E. R., 1967, Analysis of means—A graphical procedure, Industrial Quality Control, 24(2):
101–109.
Ott, E. R. and S. S. Lewis, 1960, Analysis of means applied to percent defective data, Technical Report
No. 2, The Statistics Center, Rutgers—The State University, New Brunswick, NJ.
Pearson, E. S., 1935, The Application of Statistical Methods to Industrial Standardization and Quality
Control, British Standard 600: 1935, British Standards Institution, London, U.K.
Romboski, L. D., 1969, An investigation of quick switching acceptance sampling systems, PhD
­­dissertation, Rutgers—The State University, New Brunswick, NJ.
Schilling, E. G., 1973a, A systematic approach to the analysis of means. Part I: Analysis of treatment
effects, Journal of Quality Technology, 5(3): 93–108.
Schilling, E. G., 1973b, A systematic approach to the analysis of means. Part II: Analysis of contrasts;
Part III: Analysis of non-normal data, Journal of Quality Technology, 5(4): 147–159.
Schilling, E. G., 1978, A lot sensitive sampling plan for compliance testing and acceptance inspection,
Journal of Quality Technology, 10(2): 47–51.
Schilling, E. G., 1979, A simplified graphical grand lot acceptance sampling procedure, Journal of
Quality Technology, 11(3): 116–127.
Schilling, E. G. and H. F. Dodge, 1969, Procedures and tables for evaluating dependent mixed accep-
tance sampling plans, Technometrics, 11(2): 341–372.
Schilling, E. G. and P. R. Nelson, 1976, The effect of non-normality on the control limits of X charts,
Journal of Quality Technology, 8(4): 183–188.
Schilling, E. G. and J. H. Sheesley, 1978, The performance of MIL-STD-105D under the switching
rules, Journal of Quality Technology, Part 1, 10(2): 76–83; Part 2, 10(3): 104–124.
Simon, L. E., 1941, An Engineer’s Manual of Statistical Methods, John Wiley & Sons, New York.
Simon, L. E., 1944, The industrial lot and its sampling implications, Journal of the Franklin Institute,
237: 359–370.
Soundararajan, V. and S. D. Arumainayagam, 1990, Construction and selection of modified quick
switching systems, Journal of Applied Statistics, 17(1): 83–114.
Soundararajan, V. and R. Vijayaraghavan, 1990, Construction and selection of tightened-normal-
tightened (TNT) plans, Journal of Quality Technology, 22(2): 146–153.
Squeglia, N. L., 1994, Zero Acceptance Number Sampling Plans, 4th ed., ASQ Quality Press,
Milwaukee, WI.
Taylor, W.A., 1996, Quick switching system, Journal of Quality Technology, 28(4): 460–472.
Tukey, J. W., 1953, The problem of multiple comparisons, Unpublished manuscript, Princeton
University, Princeton, NJ.
United States Department of Defense, 1957, Military Standard, Sampling Procedures and Tables for
Inspection by Variables for Percent Defective (MIL-STD-414), U.S. Government Printing Office,
Washington, DC.
550 Acceptance Sampling in Quality Control

United States Department of Defense, 1989, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
United States Department of Defense, 1996, Department of Defense Test Method Standard, DOD
Preferred Methods for Acceptance of Product (MIL-STD-1916), U.S. Government Printing Office,
Washington, DC.
United States Department of Defense, 1999, Department of Defense Handbook, Companion Document
to Mil-Std-1916 (MIL-HDBK-1916), U.S. Government Printing Office, Washington, DC.
18
Reliability Sampling

Historically, sampling plans have been used to assess the present quality of the material
examined. They are employed to determine the acceptability of the product against speci-
fications at a given time. This has usually been the time of sale. Of course, the implication
is that items presently acceptable will retain their utilitarian properties upon reaching the
consumer. An important quality characteristic of some products, however, is degradation
in use, that is, the useful life of the product with regard to some property.
The advent of considerations of reliability imposed by high technology programs,
such as space and atomic power, consumerism, and conformance testing to government
mandatory standards, has placed a new dimension on the sampling problem, that of
time. Reliability sampling plans are used to determine the acceptability of the product
at some future point in its effective life. This usually involves some form of life testing.
An Advisory Group on Reliability of Electronic Equipment (AGREE) was formed in 1952
under the assistant secretary of defense to “monitor and stimulate interest in reliability
matters and recommend measures which would result in more reliable electronic equip-
ment.” AGREE (1957) defined reliability as follows: “Reliability is the probability of per-
forming without failure a specified function under given conditions for a specified period
of time.” Reliability testing is to provide assurance of reliability. In this sense, it is not
testing what the product is, but rather how it will operate, over time, in the hands of the
consumer. The standard plans discussed so far determine whether the product is made to
specifications. Reliability plans assess how it will perform.
The time dimension implicit in reliability testing is superimposed on the sampling
­problem as an additional criterion. Samples must be tested for a specified length of time.
When all units are tested to failure, the standard plans can be utilized to assess the results
against specified requirements. If lifetimes are measured, these results can be used in
a variables sampling plan, such as MIL-STD-414 (1957) or its derivatives, provided the
­distributional assumption of the plan is satisfied. Also, the number failing before a required
time can be used with standard attributes plans in determining the disposition of the mate-
rial, for example, MIL-STD-105E (1989).
In reliability and safety testing, extremely low levels of probability of acceptance are
often used. When a test based on a two-point plan (p1, p2, α, β) has been passed, it is often
said that a reliability of at least π = 1 − p2 has been demonstrated with γ = 1 − β confidence.
Specifications are often written in this way. Clearly, a variety of plans could satisfy such
a requirement on what amounts to limiting quality (LQ). For example, it follows from
Schilling–Johnson Appendix Table T5.2 that to demonstrate .9995 confidence of .99 reliabil-
ity, the plan n = 1000, c = 1 could be used since for c = 1, when Pa = .0005, np = 10.000, and

np 10.000
= = 1000
1- p .01

For a discussion of this type of specification for sampling plans in reliability, see Lloyd and
Lipow (1962, p. 280).

551
552 Acceptance Sampling in Quality Control

When testing of the sample is terminated before the specified lifetime with some units
still unfailed, however, complications arise. A sample of this sort is called “censored” and
implies that some units were tested without generating failures as such. In this area, the
test termination time should not be used as the failure time for the unfailed units since,
clearly, they would probably have lasted longer, and consequently to do so would bias the
results. It is easily seen, from actuarial work, that mean lifetime would be grossly under-
stated if, in a sample of 100 people, the first death was used as average lifetime. Yet this is
exactly the result if a test is stopped at the first failure and the termination time used as the
failure time of the remaining elements in the sample.
Statistical methods have been developed for use with censored samples. Life tests may be
deliberately terminated after a given number of failures or a specified period and analyzed
using these procedures. This is usually done to speed up the test or for economic consider-
ations. The methods for dealing with censored data can also be used in situations in which
some of the units have not been tested to failure because of difficulties with the test equip-
ment, units failed for causes other than those being tested, broken, or stolen units.

Censored Sampling
Analysis of censored data can be made with varying degrees of sophistication. One of
the most useful tools in this regard is a properly constructed probability plot. The actual
failures observed are plotted against plotting positions that have been adjusted for the
amount and type of censoring in the sample. This method is based on an empirical deter-
mination of the hazard rate h(x), or instantaneous failure rate, associated with each of the
observed failures. The cumulative hazard rate H(x) may then be transformed into prob-
ability plotting positions. The development of the approach has been aptly described by
Nelson (1969) in an American Society for Quality Control Brumbaugh Award winning
paper. The method is as follows for a sample of n:

1. Order the data, including failure and censoring times. Distinguish the censored
observations by marking with an asterisk.
2. Calculate the hazard value h(x) associated with each failure time as the reciprocal
of the number k of units with failure and censoring times greater than or equal to
the failure observed. That is,
1
h=
k
3. Cumulate the hazard values to obtain the cumulative hazard value H for each
observed failure. For the ith failure,
i

Hi = åh
j =1
j

4. Convert the cumulative hazard value to a plotting position using the relation

( )
Pi = 1 - e - Hi 100

Reliability Sampling 553

Appendix Table T18.1 is a tabulation of values of Pi for associated 100Hi as

­calculated by Sheesley (1974).
5. Make a probability plot on appropriate paper to assess the shape of the life
­distribution and, if the points fail on a straight line, to estimate its parameters.

Special hazard probability plot papers are available for this purpose upon which cumula-
tive hazard may be plotted directly, avoiding steps 4 and 5. Plotting papers for the nor-
mal, lognormal, exponential, Weibull, and extreme value distributions are given by Nelson
(1969).
The technique is illustrated by sample data taken from a life test of class B insulation
on small motors or motorettes as given by Hahn and Nelson (1971). Ten motorettes were
tested at 170°C to obtain information on the distribution of insulation life at elevated tem-
peratures. The test was stopped at 5448 h with three motorettes still running. Results on
the other seven were, in the order of failure, 1764, 2772, 3444, 3542, 3780, 4860, and 5196.
Calculation of the probability positions is shown in Table 18.1. A normal probability plot
of these data is shown in Figure 18.1. Only the actual failures are plotted. The mean life of
this sample of motorettes appears to be about 4400 h.
Because of their frequently long duration, life test are often concluded before all units
have failed. Sometimes, this is not done by design. In such situations, the hazard plotting
procedure is an excellent technique for assessing the failure distribution models prepara-
tory to the initiation of a life test sampling plan. Plotting a substantial number of failures
on a variety of probability papers will give much insight into the probability distribu-
tion involved. More sophisticated methods are, of course, available and will be found in
­textbooks on reliability (Mann et al. 1974; Locks 1995; Nelson 2004).
The unique necessity for a time specification in reliability tests has led to the develop-
ment of a number of published sampling procedures. Two will be discussed here and are
based on the exponential and the Weibull distributions, respectively.

TABLE 18.1
Probability Positions for Motorette Data
Percent Cumulative
Hours Run Hazard Cumulative Hazard Probability
Ordered Unit k X h = 1/k H = Σh P = (1 – e−H)100

1 10 1764 0.100 0.100 9.5

2 9 2772 0.111 0.211 19.0
3 8 3444 0.125 0.336 28.5
4 7 3542 0.143 0.479 38.1
5 6 3780 0.167 0.646 47.6
6 5 4860 0.200 0.846 57.1
7 4 5196 0.250 1.096 66.6
8 3 5448a
9 2 5448a
10 1 5448a
a Still running, also known as right censored.
554 Acceptance Sampling in Quality Control

99

95
90
80

P
50

20
10
5

2000 3000 4000 5000

x

FIGURE 18.1
Probability plot of motorette data.

Variables Plans for Life Testing and Reliability (Juran 1999)

Variables sampling plans for life and reliability testing are similar in concept and oper-
ation to the plans previously described. They differ to the extent that when units are
not all run to failure, the length of the test becomes an important parameter determin-
ing the characteristics of the procedure. Further, time to failure tends to conform natu-
rally to skewed distributions such as the exponential or as approximated by Weibull.
Accordingly, many life test plans are based on these distributions. When time to failure is
normally distributed and all units tested are run to failure, the variables plans assuming
normality, discussed earlier, apply; attributes plans such as those in MIL-STD-105E may
also be used.
Life tests, terminated before all the units have failed, may be as follows:

Failure terminated: A given sample size, n, is tested until the rth failure occurs. The
1.
test is then terminated.
Time terminated: A given sample size, n, is tested until a preassigned termination
2.
time, T, is reached. The test is then terminated.

Furthermore, these tests may be based upon specifications written in terms of one of the
following characteristics:

1.
Mean life: The expected life of the product
2.
Hazard rate: Instantaneous failure rate at some specified time, t
3.
Reliable life: Life beyond which some specified proportion of items in the lot or
population will survive
Reliability Sampling 555

TABLE 18.2
Life Characteristics for Two Failure Distributions
Exponential: f(t) = [(1/μ)e−(t/μ)]
β
Weibull: f(t) = [(βtβ − 1/ηβ)e−(t/η) ], where μ = ηΓ(1 + 1/β) and g = [Γ(1 + 1/β)]β
Life Characteristic Symbol Exponential Weibull
β
Proportion failing before time t F(t) F(t) = 1 − e−(t/μ) F(t) = 1 − e−g(t/μ)
β
Proportion (r) of population surviving to time ρ ρr = 1 − F(p) = r ρr = e−ρ/μ ρr = e−g(ρ/μ)
Mean life or mean time between failures μ μ μ
Hazard rate: instantaneous failure rate at time t Z(t) = h(t) Z(t) = 1/μ Z(t) = βgtβ −1/μβ
Cumulative hazard rate for period 0 to t H(t) H(t) = t/μ H(t) = gtβ/μβ

TABLE 18.3
Values of γ = [Γ(1 + (1/β))]β for the Weibull Distribution
β 0.0 1.0 2.0 3.0 β g
0.0 1.0000 0.7854 0.7121 0.33 1.8171
0.1 4.5287 0.9615 0.7750 0.7073 0.67 1.2090
0.2 2.6052 0.9292 0.7655 0.7028 1.33 0.8936
0.3 1.9498 0.9018 0.7568 0.6986 1.67 0.8289
0.4 1.6167 0.8782 0.7489 0.6947 3.33 0.6973
0.5 1.4142 0.8577 0.7415 0.6909 4.00 0.6750
0.6 1.2778 0.8397 0.7348 0.6874 5.00 0.6525
0.7 1.1794 0.8238 0.7285 0.6840
0.8 1.1051 0.8096 0.7226 0.6809
0.9 1.0468 0.7969 0.7172 0.6778

Several sets of plans are available for the testing of life and reliability. They are illustrated
by presenting two classic plans developed by the U.S. Department of Defense, Handbook
H-108 (H-108) and the technical report TR7 (TR7), which cover life data distributed expo-
nentially and according to the Weibull distribution. TR7 is also available as an American
Society for Testing and Materials (ASTM) standard (E2555-07 “Standard Practice for Factors
and Procedures for Applying the MIL-STD-105 Plans in Life and Reliability Inspection”).
Tables 18.2 and 18.3 will be found useful in converting life test characteristics. Formulas
for various characteristics are shown in terms of mean life μ. Thus, using the tables, it will be
found that a specification of mean life μ = 1000 h for a Rayleigh distribution (Weibull, β = 2)
is equivalent to a hazard rate of .00157 at 1000 h or to a reliable life of 99% surviving at 113 h.

Handbook H-108
Quality Control and Reliability Handbook H-108 is intended to be used with quality characteris-
tics that are exponentially distributed. Its title, “Sampling Procedures and Tables for Life and
Reliability Testing,” suggests an emphasis on life testing and the standard deals primarily
with a specification of mean life. While the procedures are quite general in application for
the exponential distribution, they will be presented here in terms of the life testing problem.
556 Acceptance Sampling in Quality Control

H-108 is intended to test mean life, θ. Two values are specified:

θ0 is the acceptable mean life
θ1 is the unacceptable mean life
with risks
α, the producer’s risk
β, the consumer’s risk
respectively.
A given specific requirement on the mean of the exponential distribution can always
be stated in terms of the proportion, p, of the population failing by a specified time, T. As
shown in Table 18.2, the relation is

p = F ( T ) = 1 - e-T /q

Analogous to the specifications for the mean, in the notation of H-108,

p0 is the acceptable proportion of the lot failing before specified time, T

p1 is the unacceptable proportion of the lot failing before specified time, T
with ­associated risks α and β. Special tables are presented for use with this type of
­specification in time-terminated tests.

A unique feature of the exponential distribution, constant failure rate, allows testing to be
conducted either.

With replacement: Units are replaced on test as they fail with the replacements also
contributing to total test time and the number of failures.
Without replacement: Units are not replaced as they fail.

Beyond a necessary minimum number of units, n, the sample size is not specified since
with constant failure rate, increasing the number of units on test only decreases the length,
or waiting time, of the test but does not affect the number of failures per unit time. Thus,
replacements are possible without biasing the test. Provision is made in the handbook for
both failure-terminated and time-terminated tests. Sequential tests are also provided.
The handbook contains a wealth of information on exponential life testing. Its structure
is shown in Figure 18.2, which shows the location of material on each of the three types of
plans, including

Chapter 1 Definition
Chapter 2
Section 2A General description of life test sampling plans
Section 2B Failure-terminated plans for mean
Section 2C Time-terminated plans for mean and proportion
failing before specified time
Section 2D Sequential plans for mean

Operation
The three different types of tests provided in H-108 are conducted as follows.
 H 108

Chapter 2
Chapter 1
Sampling procedures
Definitions
and tables

Section 2B Section 2C
Failure terminated Time terminated
plans plans
Reliability Sampling

2B1–4 2B6 2C1 2C4–5 2C6 2C7

Acceptance Plans for Life test With Plans for Plans for
procedures specified plans replacement specified fraction of lot
α, β, θ1/θ0 α, β, θ1/θ0 failing before
specified time
2C2–3 2C4 2C5
2B1 2C3 2B4 Table 2B5 Without Selecting Acceptability
2B2 Table 2C5
Life Procedures Procedures Plans for replacement plan procedures
Selecting Plans for
test without with specified
the plan specified
plans replacement replacement α, β, θ1/θ0 2C2 2C3 Table 2C3 Table 2C4
Selecting Acceptability α, β, θ1/θ0
Plans without Plans with
procedures without
Table 2B1 2B5 plan replacement replacement
Expected duration replacement
Master table
and cost
for failure Table 2C1 Table 2C2
considerations
terminated Master table Master table
plans time terminated time terminated
with with
replacement replacement
Table 2B2a Table 2B2b Table 2B3 Table 2B4
Expected Expected Relative Relative
waiting waiting saving by saving by
time by time by increasing testing with
r, n r, n = kr sample size replacement Section 2D
Sequential life
test plans
Section 2A
Procedures and
discription of plans
2D1 2D2 2D3 2D4 2D5
2D6 2D7
Life Selecting Procedure Procedure Times for
2A2 2A3 Truncation Graphical
2A1 test the without with determining
Sampling OC of test procedures
Scope plans plan replacement replacement the acceptability
risks curves

Table 2A1 2D8 2D9 Appendix 2D–1

Sampling Table 2A2 Expected Expected Comparison
plan code Operating failures to time to of single and
designation curves reach decision reach decision sequential plans

FIGURE 18.2
557

Structure of H-108.
558 Acceptance Sampling in Quality Control

Failure Terminated
Place a sample of n units on test. Stop the test at the rth failure. Record the successive
­failure times as Xi,n, i = 1, 2, …, r. Compute the estimated mean life q̂r,n as

1é ù
r

q̂ r , n = ê
rê
ë
åX
i =1
i,n + ( n - r ) X r, n ú
úû
( without replacement )

or
nX r , n
q̂ r, n =
r
( with replacement )

Compare q̂r,n to the acceptance constant C.
If qˆ r , n ³ C , accept.
If qˆ r , n < C , reject.

Time Terminated
Place a sample of n units on test. Stop the test at time T. If r or fewer failures have occurred,
accept. If more than r failures have occurred, reject.

Sequential
Place a sample of n units on test. Record the successive failure times as Xi,n, i = 1, 2, …, k.
Compute the total survival time V(t) for k failures at time t as
k

V (t ) = åX i =1
i, n + (n - k )t ( without replacement )

or
V ( t ) = nt ( with replacement )
Compare to sequential limits. After the kth failure,

Accept if V ( t ) ³ h0 + ks or V ( t ) ³ sr0 .
Reject if V ( t ) £ h1 + ks or k = r0 and V ( t ) < sr0 .

where r0 is a truncation criterion for the sequential test. That is, the test is terminated at the
r0th failure and the truncation criterion applied.

Proportion Failing by Specified Time

Conduct a time-terminated test (without replacement) as specified by the plan selected.
A summary of the operation of the H-108 test plans is shown in Table 18.4.

Selection
The selection of a plan begins with H-108 Table 2A.1 given in Appendix Table T18.2. To
find a plan, the operating ratio
q
R= 1
q0
TABLE 18.4
Operation of H-108
Step Section B Section C—Parts I, II Section D Section C—Part III
Reliability Sampling

Characteristic Mean Fraction Failing before Specified Time

Type Test Failure Terminated Time Terminated Sequential Time Terminated

Specified T = time interval

θ0 = acceptable mean life, α = producer’s risk p0 = satisfactory fraction failing in time T
θ1 = unacceptable mean life, β = consumer’s risk p1 = unsatisfactory fraction failing in time T
G = failure rate in time T where p = GT
Criteria n = sample size n = sample size n = sample size n = sample size
r = termination number r = termination h0 = acceptance intercept r = termination number
C = acceptability constant number h1 = rejection intercept T = termination time
T = termination s = slope
time r0 = truncation criterion
Observation xi,n = time of ith failure xr, n = time of rth xi, n = time of ith failure xr,n = time of rth failure
failure
Statistics r k
1é ù
Without q̂ r ,n = ê xi ,n + ( n - r ) x r ,n ú xr, n V (t ) = åx i ,n + (n - k )t xr , n
replacement i =1
r êë å úû i =1

= estimate of mean life = total survival time at time t

nx with k failuress
With q̂ r ,n = r ,n = estimate of mean life xr, n xr , n
r
replacement V ( t ) = nt
= total survival time at time t
with k failures

Acceptability qˆ r ,n ³ C ® accept xr ,n ³ T ® accept Accept if V(t) ≥ h0 + ks or V(t) ≥ sr0 xr ,n ³ T ® accept

criterion qˆ r ,n < C ® reject xr ,n < T ® reject Reject if V(t) ≤ h1 + ks or k = r0 or xr ,n < T ® reject
V(t) < sr0
559
560 Acceptance Sampling in Quality Control

is formed for specified θ0 and θ1. The sampling plan code designation is then located under
the risks α and β desired. For example, for α = .05, β = .10 if

q0 = 600 h q1 = 200 h

so that
q1 200 1
R= = =
q0 600 3

Code B-8 would be selected.

Master tables for each type of test give the factors necessary to define the test. Indexed
by sampling plan code, the factors are multiplied by θ0 to give the test parameters. The
master tables are

Table 2B1 Failure terminated with or without replacement (gives r, C/θ0)

Table 2C1 Time terminated without replacement (gives r, T/θ0)
Table 2C2 Time terminated with replacement (gives r, T/θ0)
Table 2D1 Sequential with or without replacement (gives r0, h0/θ0, h1/θ0, s/θ0)

Selected values from these tables for α = .05 are given in Appendix Tables T18.3 through
T18.6. The handbook also gives special tables in which additional plans are indexed by
α, β, and R = θ1/θ0 for the failure- and time-terminated plans.
Time-terminated plans for proportion failing before a specified time will be found in
H-108 Table 2C5 given in Appendix Table T18.7. It shows values of r and the factor D
indexed by α, β, and R = p1/p0 where
D
n=
p0

The question of sample size is directly incorporated in time-terminated tests, which require
n to be specified before a termination time can be determined. For failure-terminated plans,
this is not the case. It would be theoretically possible to test, with replacement, one unit
at a time until r failures were generated. This, of course, could take inordinately long. As
a result, a number of units, n, are tested simultaneously to speed up the test. Guidance
in selecting n for failure-terminated tests is given in H-108, Part II of Section B. Tables
­presented include the following:

Table 2B.2(a) Expected waiting time indexed by r and n showing ratio

Expected waiting time for r failures in a sample of n
Mean life of lot
Table 2B.2(b) Expected waiting time indexed by r and n = kr, k = 1, 2, …, 10, and 20 showing ratio
Expected waiting time for r failures in a sample of n = kr
Mean life of lot
Table 2B.3 Expected relative saving in time by increasing sample size indexed by r and n showing ratio
Expected waiting time for r failures in a sample of n
Expected waiting timefor r failures in a sample of r
Table 2B.4 Expected relative saving in time by testing with replacement indexed by n and r showing ratio
( Expected waiting time for r failures in a sample of n when testing with replacement )
( Expected waiting time for r failures in a sample of n when testing without replacement )
Reliability Sampling 561

Sequential plans may also be conducted with a variety of sample sizes. Only minimum sam-
ple size is specified for tests without replacement. As a result, the sample size may also be
chosen with regard to waiting time or economic considerations. Special factors are presented
with the sequential plan to allow the determination of the expected number of failures
required for a decision and for calculation of the expected waiting time with any sample size.
A check sequence for the utilization of all these tables in determining a sampling plan is
shown in Figure 18.3.

Example of H-108 Application

After the manner of Hahn and Shapiro (1967, p. 107), consider a life test of a large lot of
very expensive batteries. The test is to be conducted such that

q0 = 70 with a = .05
q1 = 7 with b = .10

Then

q1 7
R= = = .1
q0 70

and H-108 Table 2A.1 shows Code B-2 to be relevant. Using the appropriate tables, possible
test plans are as follows.

Failure Terminated
H-108 Table 2B.1 gives r = 2 and C/θ0 = .178. Hence,

C = .178 ( 70 ) = 12.5

Sample size would be determined from economic considerations. H-108 Table 2B.3
shows waiting time could be reduced by 61% by using a sample of 4 rather than 2. H-108
Table 2B.4 shows a further reduction in waiting time by 14% could be affected by testing
with ­replacement. For a lot with mean life 70 h, H-108 Table 2B.2a shows expected waiting
time to be 40.8 h for a sample of 4 tested without replacement. Suppose a sample of 4 is
tested and failures are observed at 10 and 40 h. The test is stopped with the second failure.
If the test was, in fact, conducted without replacement,

1
q̂ r ,4 = é 50 + 2 ( 40 ) ùû = 65
2ë

Since 65 > C = 12.5, the lot is accepted.

Time Terminated
H-108 Table 2C.1b shows r = 2 and T/θ0 = .104 for a test of n = 2r = 4 units without
­replacement. Here, T = .104(70) = 7.28 h. Suppose a sample of 4 is tested with no failures
by 7 h, 17 min. Since two failures would have been necessary to reject, the lot is accepted.
562 Acceptance Sampling in Quality Control

Start

Mean life θ specified Failure rate G Fraction p

in period T failure before time T
specified specified

Convert G to p
Specify p0 = G0T
α, β, θ1/θ0 p0 = G1T

Specify
α, β, p1, p0
Determine
code designation
(Table 2A.1) Determine
termination number r
and factor D (Table 2C.5)

Check OC curve Determine

(Table 2A.2) n= D
P0

Apply time terminated

test for time T
Select type plan without replacement

Failure Time Sequential

terminated terminated

Determine Determine replacement Determine

acceptability policy acceptance line
criterion, C/θ0 and constants
h0, h1, s
termination number,
and truncation
r, from code Without With criterion r0
(Table 2B.1) replacement replacement (Table 2D.1)
or
(Table 2B.5) Minimum sample
Determine Determine
termination size is
termination
Determine time, T; n = r0
time, T;
sample size, n, termination termination
and replacement number, r; number, r;
policy and sample Obtain expected
and sample
size n waiting time constants
size n
(Table 2C.1) (Table 2C.2) E0(t), Eθ (t), ES(t), Eθ (t),
1 2
or or (Table 2D.1)
(Table 2C.3) (Table 2C.4)

Determine replacement policy

Expected Relative
saving in Relative
waiting time
waiting time by saving in
no replacement Without replacement With replacement
increasing n by
(Table 2B.2a)
n replacement
or
(Table 2B.3) (Table 2B.4) Fix n from Fix n from
(Table 2B.2b)
expected waiting expected waiting
time formulas time formulas
Specify n 2D.9 2D.9
and replacement policy

Check total Check total

Without With survival time Vt survival time Vt
replacement replacement calculation calculation
2D.3.2 2D.4.2

Check mean life Check mean life

estimate ^θr,n estimate ^θr,n
calculation calculation
2B.3.2 2B.4.2

Apply plan

FIGURE 18.3
Check sequence for selecting plan from H-108.
Reliability Sampling 563

Sequential
H-108 Table 2D.1b shows for Code B-2,

r0 = 6

h0 = .2254 ( 70 ) = 15.8

h1 = -.2894 ( 70 ) = -20.3

s = .2400 ( 70 ) = 16.8

with expected number of failures to reach a decision

Mean Life Expected Failures to a Decision

0 E0(r) = 1.2
θ1 = 7 Eq1(r) = 1.6
s = 16.8 Es(r) = 1.1
θ0 = 70 Eq0 (r) = 0.3

The acceptance and rejection lines are

Acceptance: V(t) = 15.8 + 16.8k

Rejection: V(t) = −20.3 + 16.8k

The usual sequential diagram can be represented in tabular form by solving the equa-
tions as follows:

Failure Reject Accept

1 * 32.6
2 13.3 49.4

where * indicates no rejection can occur on the first failure. Suppose 4 units are placed on
test with replacement. The first failure occurs after 10 h so that

V ( t ) = nt = 4 ( 10 ) = 40

Since V(t) = 40 exceeds the acceptance line value of 32.6, the lot is accepted.

Proportion of Lot Failing by Specified Time

The specifications on mean life can be converted to proportion failing by a specified time
by use of the relationship given in Table 18.2. In this case, for T = 13,

p0 = 1 - e (
-13/70 )
= .169
( -13/7 )
p1 = 1 - e = .844
564 Acceptance Sampling in Quality Control

Here,

p1 .844
R= = = 4.99
p0 .169

H-108 Table 2C.5 gives r = 4 and D = 1.37 for R = 5 with α = .05, β = .10, so

1.37
n= = 8.1 ~ 8
.169

For this test 8 units are placed on test for a maximum of 13 h without replacement. If four
or fewer units have failed at 13 h, the lot is accepted.

Measures
Operating characteristic (OC) curves are provided in Table 2A.2 of H-108 indexed by
life test sampling plan code designation. The curve for Code B-2 is shown in Figure 18.4.
The OC curves are for failure-terminated plans; the curves for sequential plans and time-­
terminated tests are essentially equivalent.

Further Considerations
The theory and development of the plans contained in H-108 will be found in a compre-
hensive two-part paper published by Epstein (1960a,b) in Technometrics.
ASTM International Standard E2696
E2696 “Standard Practice for Life and Reliability Testing Based on the Exponential
Distribution” is an ASTM standard that presents standard sampling procedures and tables

1.00
0.90
0.80
Probability of acceptance

0.70
0.60
B–1 B–2 B–3 B–4 B–5
0.50
0.40
0.30
0.20
0.10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Actual mean life/acceptable mean life (θ/θ 0)

FIGURE 18.4
H-108 Table 2A.2: OC curves for life tests terminated upon occurrence of preassigned number of failures. (From
United States Department of Defense, Sampling procedures and tables for life and reliability testing, Quality
Control and Reliability (Interim) Handbook (H-108), Office of the Assistant Secretary of Defense (Supply and
Logistics), Washington, DC, 1960, p. 2.9.)
Reliability Sampling 565

for life and reliability testing in procurement, supply, and maintenance quality control
operations and research and development activities. This practice describes general proce-
dures and definitions of terms used in life test sampling and describes specific procedures
and applications of the life test sampling plans for determining conformance to estab-
lished reliability requirements. E2696 is an adaptation of the Quality Control and Reliability
Handbook H-108, “Sampling procedures and tables for life and reliability testing (based on
exponential distribution),” U.S. Government Printing Office, April 29, 1960.

Technical Report TR7

Defense Department Quality Control and Reliability Technical Report TR7 (1965) provides
procedures and factors for adapting MIL-STD-105E plans to life and reliability testing
when a Weibull distribution of failure times can be assumed. It allows appropriate test
truncation times for specific reliability criteria to be determined for use of the plans when
all units are not run to failure. The reliability criteria are

Mean life μ: The expected life of the product

1.
2.
Hazard rate Z(t): The instantaneous failure rate at some specified time t
Reliable life ρr: The life ρ beyond which some specified proportion r of the items in
3.
the population will survive

Naturally, as with almost all variables criteria, these characteristics require Type B sampling.
All plans in TR7 are based on an underlying Weibull distribution. Its cumulative prob-
ability distribution function may be written as

é æ t - g öb ù
F ( t0 ) = p¢ = P ( t £ t0 ) = 1 - exp ê - ç 0 ÷ ú , t0 ³ g
êë è h ø úû

with density function

b (t - g )
b -1
é æt-gö b
ù
f (t ) = exp ê - ç ÷ ú, t ³ g
hb êë è h ø úû

where
γ is the location (or threshold) parameter
β is the shape parameter
η is the scale parameter (characteristic life)
μ is the mean life

The mean μ of the Weibull distribution is at

æ 1ö
m = g + hG ç 1 + ÷
è b ø
566 Acceptance Sampling in Quality Control

with a hazard rate

b -1
æ b öæ t - g ö
Z (t ) = ç ÷ ç ÷
è h øè h ø

and a reliable life

rr = g + h ( - ln r )
1/b

The location parameter γ is often taken to be zero. This is the case in TR7. When it is not
zero, for example, when γ = γ0, then the observations, t, are simply adjusted to t′ = t − γ0 so
that μ′ = μ − γ0 and the analysis is performed in terms of t′ and μ′. Naturally, the final results
are reported in terms of t and μ by reversing the process, to obtain

t = t¢ + g 0 m = m¢ + g 0

for final results t′ and μ′. The procedures of TR7 are independent of the scale parameter η,
and so it need not be specified.
Probability plots and goodness of fit tests must be used to assure that individual mea-
surements are distributed according to the Weibull model. When this distribution is found
to be an appropriate approximation to the failure distribution, methods are available to
characterize the product or a process by estimating the three parameters (γ, β, and η) of the
Weibull distribution. These methods include estimates from the probability plots and also
point and interval estimates.
The plans are given in TR7 and are based on theoretical material and tables generated
in three previous Defense Department Quality Control and Reliability Technical Reports,
each concerned with a specific reliability criterion used in TR7. These are

Mean Life Criterion, TR3 (1961)

Hazard Rate Criterion, TR4 (1962)
Reliable Life Criterion, TR6 (1963)

These three technical reports, written to be used with MIL-STD-105C, abound in excellent
examples and detailed descriptions of the methods utilized in TR7.
Once specified, the reliability criteria may be converted from one to the other using the
relationships shown in Table 18.2. Mean life will be emphasized here because of its simplic-
ity and the popularity of that criterion in nondefense life testing.

Mean Life Criterion

Technical Report TR3 (1961) provided plans and procedures for developing and applying
the Weibull plans using mean life μ as the criterion for acceptance. The dimensionless ratio
t/μ is related to the cumulative probability p′. Values of t or μ can easily be determined
from the ratio t/μ once the other is specified. Since p′ is the proportion of product failing
before time t, it can be used in the role of “percent defective” in any attributes plan. The
relationship of p′ to t/μ, then, ties the percent defective to specified values of test time t
and mean life μ.
Reliability Sampling 567

The relationship is straightforward since

b
p¢ = F ( t ) = 1 - e ( )
- t/h

and

æ1 ö
m = hG ç + 1 ÷
è b ø

so that
b
æ t æ 1 öö
- çç G ç + 1 ÷ ÷÷
è m è b øø
p¢ = 1 - e

and solving for t/μ

( )
1/b
t - ln ( 1 - p¢ )
=
m æ1 ö
G ç + 1÷
è b ø

for proportion defective p′ and any associated Pa.

Appendix Table T18.8 taken from TR3 (1961) shows values of the ratio 100(t/μ) corre-
sponding to selected percent defective tabulated for a number of Weibull shape param-
eters. The table can be used to develop plans based on mean life or to convert measures of
attributes sampling plans, notably their OCs in terms of mean life.
For example, suppose the plan n = 20, c = 1 is used in life testing of product with a
Weibull shape parameter β = 2. A test termination time of 200 h is employed and the num-
ber of failures counted at that time. What are the OCs of this plan in terms of mean life?
Using the OC curve of the attributes plan and Appendix Table T18.8, we have

Pa p 100(t/μ) = k μ = 100(200)(1/k)

.983 .01 11.31 1768

.940 .02 16.03 1248
.810 .04 22.79 878
.517 .08 32.59 614
.289 .12 40.34 496
.176 .15 45.48 440
.069 .20 53.30 375
.024 .25 60.53 330

Further, suppose an attributes plan is to be derived having a producer’s quality level

(PQL) of 1250 h and a consumer’s quality level (CQL) of 400 h with risks α = .05 and β = .10,
respectively. Units are to be tested for 200 h and it is known that failures are distributed
Weibull with β = 2. At

ætö æ 200 ö
PQL: 100 ç ÷ = 100 ç ÷ = 16
m
è ø è 1250 ø
568 Acceptance Sampling in Quality Control

and at

ætö æ 200 ö
CQL: 100 ç ÷ = 100 ç ÷ = 50
m
è ø è 400 ø

Using Appendix Table T18.8, we have

PQL: p = .02
CQL: p  .179

giving an operating ratio of

.179
R= = 8.95
.02

and using the table of unity values, the plan required is n = 20, c = 1. TR3 contains other
tables that facilitate the development of plans in this way, as do TR4 and TR6 for the other
reliability criteria.

Hazard Rate Criterion

Technical Report TR4 (1962) was patterned after TR3 using the product tZ(t) × 100 in place
of the dimensionless ratio (t/μ) x 100. Note that the value of t given is the termination
time of the test. For hazard rates specified for other times, tables were provided to convert
the hazard rate specified into a corresponding hazard rate at termination time of the test.
The cumulative probability p′ is related to tZ(t) × 100 in a manner similar to TR3. Resulting
values and classifications useful in converting any attributes plan to a Weibull life test,
where hazard rate is specified, are presented in TR4.

Reliable Life Criterion

Technical Report TR6 was also patterned somewhat after its predecessors, TR3, and TR4.
It used the dimensionless quantity (t/ρ) × 100 in the manner of (t/μ) × 100 and tZ(t) × 100
in the previous reports. The cumulative probability p′ is related to (t/ρ) × 100 and result-
ing values and classifications useful in converting any attributes plan to a Weibull life test
where reliable life is specified were tabulated.

TR7 Tables
TR7 combines the results of the preceding three technical reports in a document specifi-
cally intended to relate MIL-STD-105E to reliability testing where a Weibull distribution of
failures can be assumed. Tables of the appropriate conversion factors are provided for the
following criteria:

Table Criterion Conversion Factor

1 Mean life (t/μ) × 100
2 Hazard rate tZ(t) × 100
3 Reliable life (r = .90) (t/ρ) × 100
4 Reliable life (r = .99) (t/ρ) × 100
Reliability Sampling 569

Each table is presented in three parts, each of which is indexed by 10 values of β (β = 1/3,
1/2, 2/3, 1, 1–1/3, 1–2/3, 2, 2–1/2, 3–1/3, 4). The ASTM standard, E2555, contains tables
with additional β values (1–1/2, 3, 5, 10). The three parts are as follows for each criterion.

Part Tabulation

A Values of the conversion factor corresponding to the AQL shown

in MIL-STD-105E indexed by code letter and AQL
B Values of the conversion factor corresponding to a consumer’s
risk of Pa = .10 indexed by code letter and AQL
C Values of the conversion factor corresponding to a consumer’s
risk of Pa = .05 indexed by code letter and AQL

An additional table, TR7 Table 2D, allows conversion of a specified hazard rate to the
corresponding hazard rate at test truncation time for use with the tables that are in terms
of hazard rate at test truncation time. TR7 Table 2D presents values of Z(t2)/Z(t1) indexed
by the ratio of times involved, t2/t1, and the various shape parameters.
The tables for the mean life criterion are reproduced here as follows.

Appendix Table T18.9 TR7 Table 1A, 100t/μ ratios at the AQL
Appendix Table T18.10 TR7 Table 1B, 100t/μ ratios at the LQ level,
consumer’s risk = .10
Appendix Table T18.11 TR7 Table 1C, 100t/μ ratios at the LQ level,
consumer ’s risk = .05

The structure of TR7 is shown in Figure 18.5.

Operation
The conversion factors are employed in a manner identical to those presented in TR3, TR4,
and TR6 and can be used to

1. Determine the sample size necessary in testing for a fixed period to a specified
value of the test criterion (mean life, hazard rate, and reliable life) on the basis of
desired AQL
2. Determine the OCs in terms of the test criterion for given test times if an MIL-STD-
105E plan has already been specified
3. Determine an LQ level plan in terms of test time and MIL-STD-105E criteria for a
specified value of the test criterion
4. Determine an MIL-STD-105E plan most nearly matching the AQL, 10% LQ, or
5% LQ

The report gives a detailed explanation with examples of how to reach these ends.
A ­summary of the operation of TR7 is given in Table 18.5.
 570

TR7

General Master
description tables

Table 1 Table 2 Table 3 Table 4

Mean life, μ, Hazard rate, Z(t), Reliable life, pr, Reliable life, pr,
conversion factors conversion factors conversion factors conversion factors
r = 0.90 r = 0.99

Table 1A Table 1B Table 1C Table 3B Table 3C

Table 3A
100 t/μ ratios 100 t/μ ratios 100 t/μ ratios 100 t/p ratios for 100 t/p ratios for
100 t/μ ratios
by AQL and shape for 10% consumer for 5% consumer 10% consumer risk 5% consumer risk
by AQL and
parameter risk by code risk by code by code letter, by code letter,
shape parameter
letter, AQL and letter, AQL and AQL and shape AQL and shape
r = 0.90
shape parameter shape parameter parameter parameter
r = 0.90 r = 0.90

Table 2A Table 2D
100 tZ(t) products Table 2B Table 2C Hazard rate ratios Table 4B Table 4C
Table 4A
by AQL and shape 100 tZ(t) products 100 tZ(t) products 100 t/p ratios for 100 t/p ratios for
Z(t2)/Z(t1) to find 100 t/p ratios by
parameter for 10% consumer for 5% consumer 10% consumer risk 5% consumer risk
hazard rate at t2 AQL and shape
risk by code letter, risk by code letter, given hazard rate at by code letter, AQL by code letter, AQL
parameter
AQL and shape AQL and shape and shape parameter and shape parameter
t1, by ratio t2/t1 r = 0.90
parameter parameter and shape parameter r = 0.99 r = 0.99

FIGURE 18.5
Structure of TR7.
Acceptance Sampling in Quality Control
TABLE 18.5
Operation of TR7
Reliability Sampling

Step\
Criterion Mean Life Hazard Rate Reliable Life MIL-STD-105E Scheme

Type Test Time Terminated

Specified μ0 is the acceptable mean life. Z0(t0) is the acceptable hazard rate ρ0 is the acceptable reliable life for Code letter from MIL-STD-105E
μ1 is the unacceptable mean life. at time t0 proportion r surviving. Table 1
t is the test time. Z1(t0) is the unacceptable hazard ρ1 is the unacceptable reliable life for AQL in terms of μ, Z(t), or ρr.
β is the shape parameter. rate at time t0 proportion r surviving. Shape parameter.
t is the test time. r is the proportion surviving beyond
β is the shape parameter. life ρ.
t is the test time.
β is the shape parameter.
Criteria TR7 Table 1.A gives AQL If t ≠ t0 convert hazard rates to TR7 Table 3.A or 4.A gives AQL Obtain conversion factor from
corresponding to 100t/μ0. time t using TR7 Table 2.D corresponding to 100t/ρ for r Table A of appropriate section
TR7 Table 1.B or 1.C gives code TR7 Table 2.A gives AQL specified. given AQL and shape parameter.
letter corresponding to AQL corresponding to 100tZ(t) and TR7 Table 3.B, C or 4.B, C gives code Determine the test time t
and 100t/μ1 for consumer risk TR7 Table 2.B or 2.C gives code letter corresponding to AQL and algebraically from conversion
desired. letter corresponding to AQL and 100t/ρ for r specified and factor and AQL.
100tZ(t) for consumer’s risk consumer’s risk desired. Check LQ of normal plan through
desired. Tables B and C of appropriate
section.
Procedure Single plan—Use criteria from MIL-STD-105E normal plan for code letter and AQL. determined. Use MIL-STD-105E scheme for
Scheme—Use the MIL-STD-105E system with AQL and code letter determined from the normal plan earlier. code letter and AQL.
Test for time t and apply plan to failures observed. Test for time t and apply MIL-STD-
105E to failures observed.
571
572 Acceptance Sampling in Quality Control

A comparison of the criteria for the MIL-STD-105E normal plan, Code F, 2.5% AQL
(n  =  20, c = 1) for the case in which β = 2 may be instructive. TR7 gives the following
­conversion factors:

Limiting Quality
Criterion Factor AQL Pa = .10 Pa = .05
Percent defective (MIL-STD-105E) p′ × 100 2.50 19.5 23.7
Mean life (t/μ) × 100 18.0 50 56
Hazard rate tZ(t) × 100 5.06 40 50
Reliable life (r = .90) (t/ρ) × 100 49 130 150
Reliable life (r = .99) (t/ρ) × 100 159 440 500

These criteria may be used to characterize a specific application. For example, if

125 units are to be tested for t = 10 h, the converted values in terms of the specified cri-
teria become

Limiting Quality

Criterion AQL Pa = .10 Pa = .05

Percent defective (MIL-STD-105E) 2.50 19.5 23.7

Mean life 55.6 20 17.9
Hazard rate (t = 10 h) .00506 .04 .05
Hazard rate at 20 h .01012 .08 .10
Reliable life (r = .90) 20.4 7.7 6.7
Reliable life (r = .99) 6.3 2.3 2.0

The hazard conversion factors always give the hazard rate at the time of termination of the
test. The hazard rate at 20 h was determined using Table 2D that shows

Z ( t2 )
= 2.00
Z ( t1 )

when

t2
=2
t1

for the case of a Rayleigh distribution, β = 2. This is the effect of the linear increasing failure
rate typical of a Rayleigh distribution.
The plans presented in TR7 are, of course, time terminated. The conversion factors
given in the technical report can be used to determine the test termination time directly.
Reliability Sampling 573

Conversely, when an MIL-STD-105E test is conducted for a specified termination time, t,

associated values of the desired reliability criterion can be found as

Mean life: 100

m= ´ termination time
mean factor
hazard factor
Hazard: Z (t ) =
100 ´ termination time
100
Reliable life: rr = ´ termination time
reliable life factor

Using the factors, test termination time can be determined from the specified reliability
criterion as

Mean life: mean factor

t= ´ mean life
100
reliable life factor
Reliable life: t = ´ reliable life
100

When hazard rate is specified at time t0, the hazard rate must be transformed into the
­ azard rate at the specified test termination time t. This may be done through the use of
h
Table 18.6, developed by Schilling, which shows values of

t2 Z ( t2 )
Q=
t1Z ( t1 )

TABLE 18.6
Value of Q = t2Z(t2)/t1Z(t1) Corresponding to the Ratio t2/t1
Shape Parameter (β)
t2/t1 1/3 1/2 2/3 1 1–1/3 1–2/3 2 2–1/2 3–1/3 4

1.25 1.08 1.12 1.16 1.25 1.35 1.45 1.56 1.75 2.10 2.44
1.50 1.14 1.22 1.31 1.50 1.72 1.97 2.25 2.76 3.86 5.06
1.75 1.21 1.32 1.45 1.75 2.11 2.54 3.06 4.05 6.46 9.38
2.00 1.26 1.41 1.59 2.00 2.52 3.17 4.00 5.66 10.08 16.00
2.25 1.31 1.50 1.72 2.25 2.95 3.86 5.06 7.59 14.93 25.63
2.50 1.36 1.58 1.84 2.50 3.39 4.61 6.25 9.88 21.21 39.06
2.75 1.40 1.66 1.96 2.75 3.85 5.40 7.56 12.54 29.14 57.19
3.00 1.44 1.73 2.08 3.00 4.33 6.24 9.00 15.59 38.94 81.00
3.25 1.48 1.80 2.19 3.25 4.81 7.13 10.56 19.04 50.85 111.57
3.50 1.52 1.87 2.31 3.50 5.31 8.07 12.25 22.92 65.10 150.06
3.75 1.55 1.94 2.41 3.75 5.83 9.05 14.06 27.23 81.93 197.75
4.00 1.59 2.00 2.52 4.00 6.35 10.08 16.00 32.00 101.59 256.00
4.25 1.62 2.06 2.62 4.25 6.88 11.15 18.06 37.24 124.35 326.25
4.50 1.65 2.12 2.73 4.50 7.43 12.27 20.25 42.96 150.44 410.06
4.75 1.68 2.18 2.83 4.75 7.98 13.42 22.56 49.17 180.15 509.07
5.00 1.71 2.24 2.92 5.00 8.55 14.62 25.00 55.90 213.75 625.00
574 Acceptance Sampling in Quality Control

To use Table 18.6, obtain the conversion factor from the appropriate table in TR7. Calculate
Q0 as

100 t0 Z ( t0 )
Q0 =
hazard factor

Locate Q0 in the column for the applicable shape parameter and read the corresponding
value of t2/t1. The required test termination time is

æ 1 ö
tC ç ÷
è t2 /t1 ø

Note that if the product t0Z(t0) is less than the factor, it may be necessary to convert the
hazard rate to a longer time sufficient to make Q0 larger than 1.
For example, suppose Code F, 2.5% AQL was used with a termination time 500 h. The
three reliability criteria corresponding to the AQL are to be estimated with a shape param-
eter 2. From Table A of each section, we have

Reliability Criterion Factor

Mean life (100t/μ) 18.0

Hazard rate (100tZ(t)) 5.06
Reliable life r = .90 (100t/ρ) 49.0
Reliable life r = .99 (100t/ρ) 159

For a 500 h test, these correspond to

100
Mean life ´ 500 = 2778
18
5.06
Hazard rate ( at 500 h ) = .00010
100 ´ 500
100
Reliable life ( r = .90 ) ´ 500 = 1020
49
100
Reliable life ( r = .99 ) ´ 500 = 314
159

Now, suppose Code F, 2.5% AQL is specified to have 10% LQ for the following reliability
criteria. What would be the test time for each?

10% Limiting
Reliability Criterion Quality

Mean life 1000 h

Reliable life r = .99 113 h
Hazard rate at 1000 0.00157
Reliability Sampling 575

The conversion factors from Table B of each section are

Mean life 50
Reliable life 440
Hazard rate 40

so that the test termination times are

50
Mean life: t= ´ 1000 = 500 h
100
440
Reliable life: t= ´ 113 = 497 h
100

To determine the time necessary for testing the hazard rate, the ratio Q0 is formed

100 ´ 1000 ´ .00157

Q0 = = 3.925
40

Table 18.6 indicates t2/t1 = 2. Hence, test time will be

æ 1 ö æ1ö
t0 ç ÷ = 1000 ç ÷ = 500 h
è t2 /t1 ø è2ø
Note that all three reliability criteria are equivalent as pointed out earlier in the chapter and
they all lead to the same test termination time (500 h) for use with MIL-STD-105E, Code F,
2.5% AQL.
To select a normal plan from TR7 for specified μ1 and μ2 given test time t and shape
parameter β, proceed as follows:

1. Compute the ratios 100t/μ1 and 100t/μ2.

2. Enter TR7 Table 1.A (Appendix Table T18.9) to find the AQL corresponding to the
value of 100t/μ1 shown under the shape parameter.
3. Enter TR7 Table 1.B (Appendix Table T18.10) if Pa = .10 or Table 1.C (Appendix
Table T18.11) if Pa = .05 for the CQL. Moving down the column for the shape
parameter, find the point where the AQL found in step 2 matches the ratio 100t/μ2
listed in the column. This is the code letter for the plan.

Use the code letter and AQL to determine the sample size and acceptance number from the
normal plan given in MIL-STD-105E.
Thus, to determine a plan having AQL = 2800 h and 10% LQ of 1000 h for a test of 500 h
for a life distribution with shape β = 2,

1. The ratios for μ1 and μ2 are 17.9 and 50, respectively

2. AQL = 2.5
3. Plan is Code F, 2.5 AQL

The MIL-STD-105E normal plan n = 20, c = 1 testing for 500 h will give the protection
desired.
576 Acceptance Sampling in Quality Control

TR7 with the MIL-STD-105E System

The procedure of TR7 and the preceding three technical reports was intended to facilitate
the development of a single reliability of life test plan using the criteria of an MIL-STD-105
normal plan. Utilization of TR7 in conjunction with the MIL-STD-105E system and its
switching rules is not described in the technical reports. However, adaptation to use of the
switching rules is straightforward.
Once the sample size code letter and AQL have been determined together with test time
under the normal plan, the corresponding reduced and tightened acceptance criteria may be
substituted for those of the normal plan appropriate to the switching rules. This leads to greater
protection for both the producer and the consumer. The OC curve of the resulting scheme can
be obtained by adapting the scheme OC curves given by Schilling and Sheesley (1978) to the
reliability criterion used through the conversion factors given in TR3, TR4, and TR6.
For example, in using the switching rules, the normal plan Code F, 2.5 AQL, t = 500 h
would be incorporated into a scheme as follows:

Tightened n = 32 Ac = 1 Re = 2 t = 500
Normal n = 20 Ac = 1 Re = 2 t = 500
Reduced n=8 Ac = 0 Re = 2 t = 500

Here, the MIL-STD-105E limit numbers would not be used in switching to reduced inspec-
tion. Using Appendix Table T18.8 and β = 2, the nominal AQL would be
100t
= 17.95 m = 2786 h
m

From the Schilling–Sheesley tables (Appendix Table T11.21), the LTPD for scheme perfor-
mance would be about 12%, which converts to a mean life of
100t
= 40.34 m = 1239 h
m

Note that use of the normal plan alone would result in an LQ of 1000 h, which shows the
increased protection afforded by using the switching rules.
The selection of a plan depends upon the use to which it will be put. If a single plan is
to be obtained, the procedure is simply that of determining a suitable match between the
reliability criterion selected and the normal inspection attributes plans of MIL-STD-105E.
If the MIL-STD-105E system is to be used with the switching rules, an appropriate AQL
and code letter combination must be found (see Table 18.5). Schilling and Sheesley (1978)
have shown that the use of the system can, as a minimum, result in lowering the sample
size code letter at least to the next lower category. This has been incorporated in the check
sequence for selecting a plan given in Figure 18.6. The procedure given is for matching
PQL and CQL with the corresponding risks in a two-point procedure for both a single plan
and the MIL-STD-105E scheme.

Further Considerations
The development of TR3, TR4, and TR6, which culminated in Technical Report TR7, was the
outgrowth of the work of H.P. Goode and J.H.K. Kao at Cornell University. Three papers
were published in the Proceedings of the National Symposium on Reliability and Quality Control.
 Start

Characterize the Weibull

distribution in terms of
shape parameter β

Choose the consumer risk for

limiting quality, LQ = 0.10 or 0.05

Determine the test time t desired

Reliability Sampling

Determine the reliability measure in

which to express test: µ1Z(t) or pr

Mean life µ Hazard rate Z(t) Reliable life pr

Specify Convert the hazard rate Choose the fraction

µ0 = acceptable to test time t if not surviving r = 0.90 or 0.99
mean life previously specified
µ1 = unacceptable at that time
mean life (Table 2.D) Specify
p0 = acceptable
reliable life
Specify
p1 = unacceptable
Z0(t) = acceptable
reliable life
hazard rate
Z1(t) = unacceptable
hazard rate
r = 0.90 r = 0.99

Obtain AQL from Obtain AQL, from Obtain AQL from Obtain AQL from
shape and 100t/µ0 shape and 100tZ(t) shape and 100t/µ shape and 100t/p
(Table 1.A) (Table 2.A) (Table 3.A) (Table 4.A)

Consumer’s risk 0.10 Consumer’s risk 0.05 Consumer’s risk 0.10 Consumer’s risk 0.05 Consumer’s risk 0.10 Consumer’s risk 0.05 Consumer’s risk 0.10 Consumer’s risk 0.05
at limiting quality at limiting quality at limiting quality at limiting quality at limiting quality at limiting quality at limiting quality at limiting quality

Obtain the code letter Obtain the code letter Obtain the code letter Obtain the code letter Obtain the code letter Obtain the code letter Obtain the code letter Obtain the code letter
from AQL, shape and from AQL, shape and from AQL, shape and from AQL, shape and from AQL, shape and from AQL, shape and from AQL, shape and from AQL, shape and
100t/µ1 100t/µ1 100tZ(t) 100tZ(t) 100t/µ 100t/µ 100t/µ 100t/µ
(Table 1.B) (Table 1.C) (Table 2.B) (Table 2.C) (Table 3.B) (Table 3.C) (Table 4.B) (Table 4.C)

Choice of
plan or scheme

MIL-STD-105E scheme Single plan

to be used with to be used
switching rules

Apply normal plan

Use next lower specified in
code letter MIL-STD–105E
with same AQL

Apply MIL-STD-105E

FIGURE 18.6
577

Check sequence for determining the procedure for TR7.

578 Acceptance Sampling in Quality Control

These  classic works were “Sampling Plans based on the Weibull Distribution (1961),”
“Sampling Plans and Tables for Life and Reliability Testing Based on the Weibull Distribution
(1962),” and “Weibull Tables for Bio-Assaying and Fatigue Testing (1963).”
These papers led directly to straightforward application of the Weibull distribution in
acceptance sampling as typified by TR7, which was prepared by Professor Henry P. Goode.
An analogous procedure for variables inspection based on MIL-STD-414 was subsequently
developed by Kao (1964) while at New York University.
ASTM International Standard E2555.
E2555 “Standard Practice for Factors and Procedures for Applying the MIL-STD-105 Plans in
Life and Reliability Inspection” is an ASTM standard that presents a procedure and related
tables of factors for adapting Practice E2234 (equivalent to MIL-STD-105) sampling plans to
acceptance sampling inspection when the item quality of interest is life length or reliability.
Factors are provided for three alternative criteria for lot ­evaluation: mean life, hazard rate,
and reliable life. Inspection of the sample is by attributes with testing truncated at the end
of some prearranged period of time. The Weibull distribution, together with the exponential
distribution as a special case, is used as the underlying ­statistical model.

Problems
1. If the eighth observation in the motorette data of Table 18.1 was 6000 h with the other
two units still running, how would the probability plotting positions be changed?
2. If mean life from a Weibull distribution with shape parameter β = 3 is μ = 200 h, what is
a. Proportion failing before 200 h
b. Proportion surviving to 200 h
c. Hazard rate at 200 h
d. Cumulative hazard rate at 200 h
3. H-108 is to be used in a life test where θ0 = 150 h and θ1 = 50 h with α = .05 and β = .10.
Find the appropriate failure-terminated plan. Suppose 16 units are placed on test
with replacement and the eighth failure occurs at 40 h, should the lot be accepted?
4. Find a time-terminated test appropriate to the specification of Problem 3. The test is made
with replacement and a sample twice as big as r is used. Should the lot be accepted?
5. Suppose, by mistake, the test of Problem 4 was performed without replacement.
Should the lot be accepted?
6. The government inspector prefers that a sequential plan be used instead of the
­failure-terminated plan of Problem 3. The data are tested against a sequential plot.
What are the parameters?
a. r0
b. h0
c. h1
d. s
Compute V(t) for the eighth failure. What would this value of V(t) indicate as to the
disposition of the lot?
Reliability Sampling 579

7. The specifications of Problem 3 are identical to a proportion failing before 30 h of

p0 = .18 and p1 = .45. Find a plan appropriate to these specifications with α = .05 and
β = .10 if testing is performed without replacement.
8. Suppose the plan n = 32, c = 2 is used in life testing with a termination time of 100 h.
Use p.95 = .025, p.10 = .15. For a Weibull distribution with β = 1.67, what are the values
of mean life having probability of acceptance .95 and .10?
9. MIL-STD-105E normal plan Code K, 6.5 AQL is being used in the life testing of a
Weibull distribution having a shape parameter of β = 2.5 using a termination time of
50 h. Find, in terms of mean life, the
a. AQL
b. LQ at Pa = .10
c. LQ at Pa = .05
10. Select a normal sampling plan from MIL-STD-105E having a high probability (AQL)
of passing units with mean life 1500 h and a low probability (LQ at Pa = .10) of passing
units with mean life 500 h. The shape parameter of the applicable Weibull distribu-
tion is β = 1. Testing is to be for 15 h.

References
AGREE, 1957, Reliability of military electronic equipment, AGREE Task Group Report, U.S.
Government Printing Office, Washington, DC.
American Society for Testing and Materials, Standard Practice for Factors and Procedures for Applying
the MIL-STD-105 Plans in Life and Reliability Inspection, ASTM Standards E2555, Vol. 14.02, West
Conshohocken, PA.
American Society for Testing and Materials, Standard Practice for Life and Reliability Testing Based on the
Exponential Distribution, ASTM Standards E2696, Vol. 14.02, West Conshohocken, PA.
Epstein, B., 1960a, Tests for the validity of the assumption that the underlying distribution of life is
exponential, Part I, Technometrics, 2(1): 83–101.
Epstein, B., 1960b, Tests for the validity of the assumption that the underlying distribution of life is
exponential, Part II, Technometrics, 2(2): 167–183.
Goode, H. P. and J. H. K. Kao, 1961, Sampling plans based on the Weibull distribution, in Proceedings
of the Seventh National Symposium on Reliability and Quality Control, Philadelphia, PA, pp. 24–40.
Goode, H. P. and J. H. K. Kao, 1962, Sampling procedures and tables for life and reliability testing
based on the Weibull distribution (hazard rate criterion), in Proceedings of the Eighth National
Symposium on Reliability and Quality Control, Washington, DC, pp. 37–58.
Goode, H. P. and J. H. K. Kao, 1963, Weibull tables for bio-assaying and fatigue testing, in Proceedings
of the Ninth National Symposium on Reliability and Quality Control, San Francisco, CA, pp. 270–286.
Hahn, G. J. and W. Nelson, 1971, Graphical analysis of incomplete accelerated life test data, Insulation/
Circuits, 17(10): 79–84.
Hahn, G. J. and S. S. Shapiro, 1967, Statistical Models in Engineering, John Wiley & Sons, New York.
Juran, J. M., Ed., 1999, Quality Control Handbook, 5th ed., McGraw-Hill, New York.
Kao, J. H. K., 1964, Sampling procedures and tables for inspection by variables for percent defectives
(based on the Weibull distribution), in Proceedings of the Tenth National Symposium on Reliability
and Quality Control, Washington, DC, pp. 41–56.
Lloyd, D. K. and M. Lipow, 1962, Reliability: Management, Methods, and Mathematics, Prentice-Hall,
Englewood Cliffs, NJ.
Locks, M. O., 1995, Reliability, Maintainability, and Availability Assessment, ASQ Quality Press,
Milwaukee, WI.
580 Acceptance Sampling in Quality Control

Mann, N. R., R. E. Schafer, and N. D. Singpurwalla, 1974, Methods for Statistical Analysis of Reliability
and Life Data, John Wiley & Sons, New York.
Nelson, W., 1969, Hazard plotting for incomplete failure data, Journal of Quality Technology, 1(1):
27–52.
Nelson, W., 2004, Applied Life Data Analysis, John Wiley & Sons, New York.
Schilling, E. G. and J. H. Sheesley, 1978, The performance of MIL-STD-105D under the switching
rules, Journal of Quality Technology, Part 1, 10(2): 76–83; Part 2, 10(3): 104–124.
Sheesley, J. H., 1974, Tables to convert hazard rates to probabilities, Report Number 1300-1119,
General Electric Company, Cleveland, OH.
United States Department of Defense, 1957, Military Standard, Sampling Procedures and Tables for
Inspection by Variables for Percent Defective (MIL-STD-414), U.S. Government Printing Office,
Washington, DC.
United States Department of Defense, 1960, Sampling procedures and tables for life and reliabil-
ity testing, Quality Control and Reliability (Interim) Handbook (H-108), Office of the Assistant
Secretary of Defense (Supply and Logistics), Washington, DC.
United States Department of Defense, 1961, Sampling procedures and tables for life and reliability
testing based on the Weibull distribution (mean life criterion), Quality Control and Reliability
Technical Report (TR 3), Office of the Assistant Secretary of Defense (Installations and Logistics),
U.S. Government Printing Office, Washington, DC.
United States Department of Defense, 1962, Sampling procedures and tables for life and reliability
testing based on the Weibull distribution (hazard rate criterion), Quality Control and Reliability
Technical Report (TR 4), Office of the Assistant Secretary of Defense (Installations and Logistics),
U.S. Govern- ment Printing Office, Washington, DC.
United States Department of Defense, 1963, Sampling procedures and tables for life and reliability
testing based on the Weibull distribution (reliable life criterion), Quality Control and Reliability
Technical Report (TR 6), Office of the Assistant Secretary of Defense (Installations and Logistics),
U.S. Government Printing Office, Washington, DC.
United States Department of Defense, 1965, Factors and procedures for applying MIL-STD-105D
sampling plans to life and reliability testing, Quality Control and Reliability Assurance
Technical Report (TR 7), Office of the Assistant Secretary of Defense (Installations and Logistics,
Washington, DC.
United States Department of Defense, 1989, Military standard, sampling procedures and tables for
inspection by Attributes (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
19
Administration of Acceptance Sampling

Effective acceptance sampling involves more than the selection and application of specific
rules for lot inspection. As an integral part of the quality system, the acceptance sampling
plan, applied on a lot-by-lot basis, becomes an element in the overall approach to maximiz-
ing quality at minimum cost. Acceptance sampling plans are, after all, action rules and as
such must be adapted in a rational way to the current results and the nature and history
of the inspection performed. This is what we have called acceptance control, involving
a continuing strategy of selection, application, and modification of acceptance sampling
procedures to a changing inspection environment.
While acceptance sampling is sometimes regarded as a passive procedure for adjudica-
tion of quality, the active role of inspection was recognized early by Dodge. In accepting
the Shewhart Medal from the American Society for Quality Control, Dodge (1950, p. 6)
pointed out that

Using the inspection results as a basis for action on the product at hand for deciding
whether to accept or reject individual articles or lots of the product as they come along
is, of course, an immediate chore that we always have with us. However, the inspection
results also provide a basis for action on the production process for the benefit of the
future product, for deciding whether the process should be left alone or action taken to
find and eliminate disturbing causes.

As such, inspection should involve

1. Good data
2. Quick information
3. Incentives for the producer to provide quality at satisfactory levels
4. Quantity of inspection in keeping with quality history

Indeed, according to Dodge (1950, p. 8)

A product with a history of consistently good quality requires less inspection than the
one with no history or a history of erratic quality. Accordingly, it is good practice to
include in inspection procedures provisions for reducing or increasing the amount of
inspection, depending on the character and quantity of evidence at hand regarding the
level of quality and the degree of control shown.

Figure 19.1 illustrates this principle in terms of the extent and nature of quality his-
tory. It shows roughly how representative sampling procedures could be changed as
quality ­history is developed. It assumes that quality levels have been appropriately set
and that other suppliers are available. The overriding principle in acceptance control is to
­continually adapt the acceptance procedures to existing conditions. A control chart on the
inspection results is an excellent means to monitor the progress of the inspection or as a
check inspection device if more formal procedures have been discontinued. It will indicate

581
582 Acceptance Sampling in Quality Control

Quality history

Past results Little Moderate Extensive Criterion

Demerit rating or Almost no (<1%)
Excellent AQL plan Cumulative results plan
remove inspection lots rejected

Cumulative results Few (<10%) lots

Average or AQL plan
plan rejected
LTPD plan

or Discontinue Many (≥10%) lots

Poor 100% inspection
LTPD plan with cumulative acceptance rejected
results criterion
Amount Less than 10 lots 10–50 lots More than 50 lots

FIGURE 19.1 
Progression of sampling plans: extent of quality history.

TABLE 19.1
Life Cycle of Acceptance Control Application
Stage Step Method

Preparatory Choose plan appropriate to purpose Analysis of quality system to define the exact need
for the procedure
Determine producer capability Process performance evaluation using control charts
Determine consumer needs Process capability study using control charts
Set quality levels and risks Economic analysis and negotiation
Determine plan Standard procedure if possible
Initiation Train inspector Include plan, procedure, records, and action
Apply plan properly Insure random sampling
Analyze results Keep records and control charts
Operational Assess protection Periodically check quality history and OC curves
Adjust plan When possible change severity to reflect quality
history and cost
Decrease sample size if warranted Modify to use appropriate sampling plans taking
advantage of credibility of supplier with
cumulative results
Phase out Eliminate inspection effort where Use demerit rating or check inspection procedures
possible when quality is consistently good
Keep control charts
Elimination Spot check only Remove all inspection when warranted by extensive
favorable history

when the results show a need for reassessment of inspection procedures. The stages in the
application of a sampling procedure are shown in Table 19.1.
The preparatory phase involves setting specifications for acceptance sampling and
selecting a plan. When the plan is initiated, care should be taken to train the inspector
and to analyze the results of initial applications so that any discrepancies or problems can
be worked out of the procedure.
Later, analysis of feedback information allows tightening up if necessary but should be
geared toward a reduction of inspection effort if justified by the history of the application.
Administration of Acceptance Sampling 583

This may lead to the use of skip-lotting, chain sampling, acceptance control charts, or other
special procedures in the later stages of the application. Finally, the inspection should be
phased out altogether and replaced by such procedures as a check inspection or a control
chart. Sampling plans should be regarded as stopgap measures, instituted to correct an
immediate problem or to give the assurance desired on the present product. The informa-
tion the plans generate should be used to lessen the need for future inspection as much as
possible. Sampling procedures should be designed to self-destruct at the appropriate time.
Too often a sampling plan is instituted, not to be changed for years. Too often no one
involved can tell when a plan was originated, why, or to what criteria. “We’ve always used
that plan.” “It was written on the back of an old envelope when I took over.” Or “Joe told
us to sample in this way before he retired—you remember Joe ….” These are clear indica-
tions of lack of acceptance control. Acceptance sampling is not being controlled in such
cases—rigor mortis has set in.
It should be evident that the feedback of quality information is essential for a rational
system of acceptance control. Ott (1975, pp. 181–182) has pointed out the following:

There are two standard procedures that, though often good in themselves, can serve to
postpone careful analysis of the production process:
1. Online inspection stations (100% screening). These can become a way of life.
2. Online acceptance sampling plans that prevent excessively defective lots from
proceeding on down the production line but have no feedback procedure
included.
These procedures become bad when they allow or encourage carelessness in production.
It gets easy for production to shrug off responsibility for quality and criticize inspection
for letting bad quality proceed.

Sampling plans cost money to design and implement. They can be used to perform more
than a police function. The information generated is invaluable; it is regrettable that these
results are often simply filed away or never recorded. The institution of a sampling plan
should have associated with it effective procedures for the feedback and utilization of the
data resulting from the plan.
Above all, to be effective, a sampling procedure needs to be enforced. There is no
clearer signal to a supplier to relax quality standards than the consistent acceptance by
the consumer of substandard material. A sampling plan that cannot be enforced should be
dropped, for such a plan is nothing more than a costly exercise in futility.

Selection and Implementation of a Sampling Procedure

Sampling plans are the basic tools of acceptance control. As in any field, much of the skill
of the artisan is reflected in the ability to select the tools appropriate for the job. The uses
of some sampling plans are quite broad, for example, single sampling by attributes. Others
are used to meet a very specific need, such as H-108.
Table 19.2 presents a list of possible plans to meet varying needs. The two-point plans
shown could be single, double, multiple, or sequential as determined by the requirements
of the specific application. Note that some plans, such as the two-point plans, would fit
almost every category. Table 19.2 is simply suggestive of the type and variety of plans that
could be employed for the purposes shown.
584 Acceptance Sampling in Quality Control

TABLE 19.2
Selection of Plan
Purpose Supply Attributes Variables

Simple guarantee of PQLs Unique lot Two-point plan (Type A) Two-point plan (Type B)
and CQLs at stated risks
Series of lots Dodge–Romig LTPD, Two-point plan (Type B)
two-point plan (Type B)
Maintain level of Series of lots MIL-STD-105E (ANSI/ MIL-STD-414 (ANSI/ASQC
submitted quality at ASQC Z1.4), QSS Plan Z1.9), No-Calc plan
AQL or better
Rectification guaranteeing Series of lots Dodge–Romig AOQL, Romig variables plans
AOQL to consumer Anscombe plan
Flow of CSP-1, 2, 3, multilevel Use measurements as go/no-go
individual units plan, MIL-STD-1235B
Flow of segments Wald-Wolfowitz, Use measurements as go/no-go
of production Girshick
Reduced inspection after Series of lots Skip-lot, chain, deferred Lot plot, mixed variables-
good history sentencing attributes, narrow limit gaging
Check inspection Series of lots Demerit rating Acceptance control chart
Compliance to mandatory Unique lot Lot sensitive plan Mixed variables-attributes with
standards c=0
Series of lots TNT plan Simon grand lot plan
Reliability sampling Unique lot Two-point plan (Type B) H-108, TR7
Series of lots LTPD plan, QSS system, TR7 using MIL-STD-105E
CRC Plan switching rules

The implementation of a specific application is shown by the check sequence given in

Figure 19.2. This follows through the stages shown in Table 19.1 but emphasizes the role
of feedback in the continuing application of the plan. Plans may be installed in the areas of
receiving inspection, process inspection, and final inspection or as a check inspection
of a small quantity of the finished product. The approach remains much the same in all
areas. Of prime importance is the distinction between two prime purposes for sampling as
spelled out in American War Standard Z1.3 (American Standards Association, 1942):

To provide a basis for action on the product as it comes to the inspector, accept, reject
(or rework).
To provide a basis for action on the process in the interests of the future product, leave
the process alone or correct the process.

This distinction will bear on the type of plan installed, how it is administered, and, of
course, the type of operating characteristic (OC) curve calculated to assess its performance.
When a two-point plan is to be employed, a comparison of the administrative aspects of
single, double, multiple, and sequential sampling is shown in Table 19.3. Experience has
shown single-sampling plans to be the most frequently employed while double sampling
incorporates most of the advantages of repeated samples without suffering many of the
associated administrative burdens.
Multiple and sequential plans seem less frequently employed because of variability in
inspection load and complexity of administration even though they are, in terms of the
amount of inspection, more efficient procedures.
Administration of Acceptance Sampling 585

Determine purpose
of inspection

Determine nature
of supply and lot size

Unique Series of Continuous

lots lots process run

Choose
appropriate
plan/scheme

Determine producer’s
capability and
consumer’s need

Set quality levels

and risks by

Feedback

Feedback
negotiation

Implement
plan

Analyze
results

Unique Series
lot of lots
or run or runs

Discontinue Modify severity of

inspection inspection with
quality history

Eventually
phase out plan as
rejections become rare

Install check inspection

or control chart

Discontinue inspection
after extensive favorable
experience with supplier

Spot-check
as reasonable

FIGURE 19.2 
Check sequence for implementation of sampling procedure.
586 Acceptance Sampling in Quality Control

TABLE 19.3
Comparison of Administrative Aspects of Single, Double, Multiple, and Sequential
Sampling
Single Double Multiple Sequential

ASN Most Less Still less Least

Number samples per lot One Two Many Most
Maximum number items inspected per lot Least More Even more Most
Variability in amount of inspection None More Even more Most

Source: Adapted from Statistical Research Group, Sampling Inspection, McGraw-Hill, New York,
1948, p. 35.

Determining Quality Levels

Before a sampling plan can be derived, one or more nominal quality levels must be set to
define the protection to be afforded by the plan. These include

Acceptable quality limit (AQL): Maximum fraction defective that, for purposes of
acceptance sampling, can be considered satisfactory as a process average
Average outgoing quality limit (AOQL): Maximum average outgoing quality to the con-
sumer under rectification
Indifference quality (IQ): Level of quality with equal chance of being accepted or
rejected
Lot tolerance percent defective (LTPD): Objectionable level of quality that should be
rejected at least 90% of the time (also 10% limiting quality [LQ])
Producer’s quality level (PQL): Level of quality that should be passed most of the time
Consumer’s quality level (CQL): Level of quality that should be rejected most of the
time

The AQL, AOQL, IQ, and LTPD are used to index many existing acceptance sampling
plans and schemes. Furthermore, the PQL and CQL with associated risks are used in the
derivation of two-point plans. Risks must also be associated with the AQL, IQ, and LTPD;
the latter two are fixed at .50 and .10, respectively, while the former sometimes varies over
a range in the order of 0.5%–13% as with MIL-STD-105E. These nominal quality levels are
necessarily fixed by the consumer to meet the consumer’s needs with due consideration
to protecting the reasonable PQLs from rejection. Sometimes, this is done unilaterally, but
more often by negotiation between the consumer and the producer. The consumer should
be as much interested as the producer in good lots being accepted from the point of view
of scheduling and price.
In determining quality levels, the consumer should attempt to minimize the total cost in
terms of purchase, inspection, assembly, and eventual service. The first two costs increase
as more and more perfection is demanded, while the latter two decrease. The consumer
should not expect levels of quality better than those prevalent in an industry without
­special arrangements with the producer. Since it is usually not practical to set quality levels
for each customer, the producer must choose a level acceptable to all the intended custom-
ers at prices they are willing to pay. The sales, manufacturing, quality, and engineering
Administration of Acceptance Sampling 587

organizations of both the producer and the consumer should participate in setting quality
levels jointly weighing cost, feasibility, and customer acceptance. Quality levels should be
understood by both parties and form part of the purchasing specification either directly or
by reference to recognized standards. It is the producer’s responsibility to perform suffi-
cient inspection to assure conformance. The consumer, however, should judge the produc-
er’s performance on the basis of process average where possible and not on the results of a
single lot since inspection of a single lot will seldom give a meaningful estimate of longer-
run performance. These considerations have been amplified by the Electronic Industries
Association (1949). Moreover, an excellent discussion of some of the considerations impor-
tant in setting quality levels has been given by Hamaker (1949).

Setting AQL
The AQL is usually used as an index of sampling schemes and hence is used with a series
of lots. While the AQL is by nature associated with producer’s risk, its magnitude must
be  established by the consumer. It represents the consumer’s estimate of the maximum
fraction defective that can be tolerated for sampling purposes. Higher values are not
acceptable. Lower values are desirable. Zero is seldom attainable at reasonable cost.
The state-of-the-art process average should be the starting point for determining an
AQL. This may be evaluated from past inspection results or by engineering estimate. In
this regard, Bowker and Goode (1952, pp. 41–42) state the following:

The selection of an AQL range depends almost invariably upon a compromise between
the quality that is likely to be submitted by the supplier and the quality that is ideal
from a use standpoint (0% defective). The engineering and production staff of the
receiver can estimate the percent defective that can be tolerated from an economic
or technical point of view. The quality level one can reasonably expect from the sup-
plier can best be determined from experience. In lieu of any experience with the item
for which the plan is selected or with like items, some information might possibly be
obtained from the supplier. An estimate of the quality currently obtainable should be
made as close as possible in terms of percent defective, and if this estimate represents
a satisfactory working quality.

The state-of-the-art process average is not process capability. It should include allow-
ances for differences between manufacturers and for variations in the level of quality
by a single manufacturer over time. It is the level at which quality can be expected to
be maintained on a long-term basis or at least for as long as the product in question is
to be produced. If possible, in setting the AQL, the consumer should perform a process
performance evaluation on the data from previous inspection results on the same or
similar material. Such procedures are outlined by Mentch (1980). In referring to the use
of a process capability estimate from past data to set specifications, Mentch (1980, p. 121)
points out the following:

This estimate of process capability is based on the assumption that is feasible to bring
the process into control in a technical and economic sense. Since this assumption is not
always true, the use of this estimate of process capability to set specifications is not
advisable. When it is necessary to set a specification to the process capability, this should
be done on demonstrated performance in terms of consistently attainable levels … and
not on a collection of historical data adjusted to be “in control.”
588 Acceptance Sampling in Quality Control

The way to do this is to analyze the process data over a sufficiently long period to charac-
terize the overall level of performance. This information is used to set the AQL. Bowker
and Goode (1952, p. 42) suggest the following:

If the estimate of incoming quality is better (lower percent defective) than the quality
one is willing to tolerate, and particularly if this estimate represents the best figure from
a number of possible suppliers, it would be wise to make the AQL somewhat higher
than this estimate, so that the acceptance criterion will be less exacting, fewer lots will
be rejected, and costs will be reduced for all concerned. On the other hand, should the
estimate for incoming quality be a higher percentage than the percent defective one can
reasonably accept and use, the AQL class should be set at a lower percentage than the
estimate, provided that the rejection of an excessive number of lots will not hamper
the receiver’s operations.

This latter consequence would probably demand economic concessions to the producer
since an AQL lower than the state-of-the-art process average would demand extensive
screening of product. H-53 (1954, p. 13) points out that

Selecting extremely tight quality levels (low numerical values) might result in ­prohibitive
inspection and end item cost, frequent rejection of products, or possible refusal by the
supplier to accept procurement orders or sign contracts. On the other hand, selecting
very liberal quality levels (high numerical values) might result in delivery of large
­quantities of unsatisfactory products into the supply system.

Special considerations will, of course, motivate the consumer to move the AQL away from
the state-of-the-art process average. As listed by the Statistical Research Group (1948,
p. 84), some of these are as follows:

1.
Reduction in value of product occasioned by defectives. Sometimes, the loss occasioned
by a defective is so large that if there is more than a small percentage of defectives,
the product will be worth less than it costs. In such cases, it may be desirable to
fix the AQL at or below the break-even percentage even if this should involve the
rejection of a large proportion of submitted inspection lots.
2.
Class of defects. Major defects ordinarily reduce the value of product more than
minor defects. Consequently, the AQL should ordinarily be lower for major defec-
tives than that for minor defectives.
3.
Effect of defective product on later processing and assembling. If defective product
results in market waste of material and time during later processing and assem-
bling, the AQL should be more exacting (lower). The number of items that are
assembled may also play a part.
4.
Suppliers’ average quality and urgency of demand for product. If the quality that sup-
pliers can furnish is poor and cannot readily be improved and if output is needed
badly, the AQL may have to be higher than otherwise desired; if it is not higher,
excessive rejection may occur. If suppliers’ average quality can be expected to
improve over a period, gradual lowering of the AQL may be desirable.
5.
Kind of defects included in the defects list. In order to permit consistent inspection and
to keep close control over the quality of product submitted, it will sometimes be
desirable to include in the defects list those whose effect on functioning are ques-
tionable or to define defects more stringently than is strictly necessary for the use
Administration of Acceptance Sampling 589

to which the product is to be put. When this is done, inspection subjects the item
to a severer test than the item will receive when it is used and the AQL should
accordingly be more liberal than if each item were subjected to a less severe test.

Further consideration is the number of different types of defects accumulated for test
against a single AQL. Wadsworth (1970) has shown that grouping of defects under a single
inspection class (such as majors) results in an associated decrease in the effective AQL.
For example, if N independent defect type were grouped together, each having fraction
defective p, the effective fraction defective P for the group would be

P = 1 - (1 - p )
N

As a consequence of this formula, the AQL for the group, AQLG, might be increased to
N
AQLG æ AQL I ö
= 1- ç1- ÷
100 è 100 ø

where AQLI represents the AQL desired on each individual defect type. This relationship
is represented by Figure 19.3 taken from Wadsworth (1970). The AQL desired for each

100
90
80 N = 100
70 N = 50
60
50 N = 20
40
N = 10
30

N=6
20
N=4
Parts defective (%)

N=3

10 N=2
9
8
7
N=1
5
4

1
0 1 2 3 4 5 6
Parts containing each defect type (%)

FIGURE 19.3 
Effect of grouping defects on percent defective. (Reprinted from Wadsworth, H.M., J. Qual. Technol., 2, 182, 1970.
With permission.)
590 Acceptance Sampling in Quality Control

individual defect type is entered on the x-axis. The AQL for the group (or class) is read
from the y-axis. Thus, if the two defect types are classed as majors, each of which is to have
a 4% AQL, the AQL for the class should be 8%. This follows since

2
AQL æ 4 ö
= 1- ç1- ÷
100 è 100 ø
AQL = 7.8%.

Setting AOQL
An AOQL should also reflect the consumer’s need. Of course, this measure of quality is
meaningful only for a series of lots when rejected lots are 100% inspected. Too high an
AOQL will result in an uneconomic level of defective material for the consumer. Too low
an AOQL may cause excessive screening and higher costs particularly if set below the
state-of-the-art process average. Dodge (1948) suggests setting the AOQL about one and
one-half times higher than the state-of-the-art process average to avoid excessive amounts
of screening, which results when the process average is equal to the AOQL. Dodge (1948)
has described the administration of an AOQL plan in some detail.
In general, it is a good practice to have the producer perform any screening of rejected
lots—or at least pay for it. Thus, in internal sampling, the receiving department may
perform the sampling, but the producing department should be responsible for 100%
inspection.

Setting IQ
IQ or the point of control can be used as an element in the economic determination of qual-
ity levels as shown by Enell (1954). It is, of course, the level of quality having 50% prob-
ability of acceptance, that is, the 50:50 break-even point between acceptance and rejection.
Fortunately, the break-even point between the producer and the consumer is relatively
easy to determine, one of the chief advantages of plans. Hamaker et al. (1950, p. 363) have
pointed out that

The point of control may conveniently be interpreted as the point dividing “good” and
“bad” lots. Experience has taught that the producer and the consumer readily agree as to
a suitable choice of this parameter.

Thus, the IQ level can be a useful measure in characterizing a plan.

Setting LTPD (or LQ)

An LTPD (or LQ) may be used as a quality level in inspecting a single lot or it may be
used with a series of lots. It should be borne in mind that LTPD constitutes an extremely
pessimistic view of the protection afforded by a sampling plan. After all, nine to one odds
are roughly akin to obtaining three heads in three flips of a coin. It is not too likely, but it
can happen. However, no producer could stay in business if 90% of the lots were rejected.
Hence, the LTPD should be set well beyond the AOQL or the AQL. In discussing an LTPD
plan, Schilling (1978, p. 49) points out that for c = 0 plans
Administration of Acceptance Sampling 591

The fraction defective to be protected against by the plan should be set at a level no
less than 22 times the fraction defective that represents the state of the art. When the
level to be protected against is necessarily closer to the state-of-the-art fraction defective,
sampling plans allowing one or more defects in the sample should be used. Thus, for
example, by accepting if the sample contains three or fewer defectives, a single sampling
plan can be derived for which the fraction defective protected against can reasonably be
set at five times the state-of-the-art fraction defective.

Setting the LTPD at least five times the state-of-the-art fraction defective (usually reflected
in AQL) is probably a good rule of thumb.
Dodge and Romig (1959, p. 6) recommend that the LTPD be chosen at a level that will
almost surely be met by every lot. In fact, they suggest that

In choosing a value of LTPD, consider and compare the cost of inspection with the eco-
nomic loss that would ensue if quality as bad as the LTPD were accepted often. Even
though the evaluation of economic loss may be difficult, relative values for different
levels of percent defective may often be determined.

Thus, the LTPD should be carefully chosen to be an extremely pessimistic quality level that
should be rejected most of the time. In a series of independent lots, the probability that two
successive lots would pass at the LTPD level of quality is just 1%.

Relation of Levels
The relation of AQL (when defined as having a probability of acceptance of .95), AOQL,
IQ, and LTPD varies among individual plans. Using the Poisson approximation, it is pos-
sible to portray this relation by a modification of the Thorndyke chart. This is shown in
Figure 19.4 and can be used in assessing the effect and interrelationships of these quanti-
ties. Here, we have drawn additional curves labeled L and M, respectively, for pL = AOQL
and pM (the fraction defective at which the AOQL occurs) on the chart. For example, for
higher values of acceptance number, the AOQL of a single-sampling plan approaches and
exceeds the AQL, their being equal at about c = 17. Furthermore, when c = 17

npL = 11.6

so that for a plan with sample size 500

11.6
AOQL = = 0.23
500

The AOQL occurs at

npM = 13.5

so that

13.5
pM = = .027
500
 592

x=6 7 8 9 10 15 20 30 40 x = 50
0.99999 0.99999
4 5
0.9999 3 0.9999
10 50
9
0.999 2 8 0.999
7
6 30
5
0.99 4 0.99
1 3 20
40
2 15
0.9 x=0 10 0.9
1
0.8 9 0.8
8
0.7 7 0.7
L 6
0.6 5 0.6
0.5 30 0.5
4
0.4 3 0.4
0.3 M 0.3
2
0.2 0.2
1

Probability of X or less
Probability of X or less

0.1 0.1
0

20
0.01 0.01

15
0.001 0.001

0.0001 0.0001

0.00001 10 0.00001
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.9 0.0 2 3 4 5 6 7 8 9 10 x = 0 20 x = 5 30
Value of np, np1, npM

FIGURE 19.4 
Modified Thorndyke chart. (From Statistical Research Group, Columbia University, Sampling Inspection, McGraw-Hill, New York, 1948, p. 35. With permission.)
Acceptance Sampling in Quality Control
Administration of Acceptance Sampling 593

Setting PQL and CQL

The PQL, p1, and the CQL, p2, are used in setting up two-point plans. In general, the PQL
should be set much as the AQL is set for sampling schemes, that is, to reflect the state-of-
the-art process average. The CQL, on the other hand, should be a conservative level of
unacceptable quality set much as the LTPD. The PQL and CQL are meaningless unless
associated producer’s and consumer’s risks, α and β, are also quoted. It is a common prac-
tice to set α = .05 and β = .10.
In discussing two-point plans, Peach (1947, p. 27, 333) suggests that

… Much as we would like to demand perfection in our purchased materials, such a

demand is not practical. No method of inspection known can enforce such a standard.
The common sense alternative is to decide in advance to tolerate some small proportion
of defective material. This proportion … is designated by the symbol p1 or AQL.
In general, p1 should be known to the supplier and should be part of the specification to
be legally binding …. This does not apply to p2; indeed, as a matter of policy, informa-
tion about p2 should never be given to an outsider. The vendor’s contract sets a quality
standard, which is, or should be, p1; his job is to meet that standard. The customer may,
for the sake of economy in inspection, set some high value of p2, thus taking a consider-
able risk of accepting substandard material; but it is not part of the vendor’s business
to inquire how great this risk is. Such information could only be used to cheat the cus-
tomer, by enabling the vendor to manufacture, not to the standards of the contract but
to the loopholes in the customer’s inspection. When a supplier asks a customer “How
much inspection do you do?,” the answer should be “That information is confidential”
or “100 percent.”
In safeguarding his own rights, however, the buyer must not prejudice those of the sup-
plier. If for his own convenience he changes from one inspection plan to another, he
should be careful to keep p1 and α the same; the supplier has a right to demand a fixed
quality standard.

Military Standard 105E (U.S. Department of Defense 1989) makes a similar point in para-
graph 4.4, where it states that “the selection or use of an AQL shall not imply that the con-
tractor has the right to supply any defective unit of product.”

Economic Considerations
Ultimately, the selection of quality levels must be resolved by economic considerations.
The consequences of various possible nominal levels must be weighed against costs, OCs,
and other factors. This may be done explicitly or implicitly.
Kavanagh (1946) has provided an in-depth discussion of the procedure for determining
the unit cost of acceptance (in terms of saving by removing a defective) and unit cost of
inspection. A simple model for balancing these costs has been presented by Enell (1954).
Suppose costs are quoted on a per unit basis and
A is the unit cost of acceptance (i.e., the cost of one defective unit being accepted)
I is the cost of inspection of one piece
C is the cost of repairing or replacing one defective
p is the fraction defective in the lot
594 Acceptance Sampling in Quality Control

If R is the unit cost of rejection, then

I
R= +C
p

which amounts to the cost of inspection to find one defective piece plus the cost of correct-
ing it when found. At the break-even point between the cost of acceptance and the cost of
rejection,

I
A=R A= +C
p

with an associated fraction defective

I
pB =
A - C

at the break-even point between the costs. But this is also the break-even point between
acceptance and rejection since when p < pB the cost of rejection must exceed the cost of
acceptance. Also, when

p > pB

the cost of acceptance exceeds the cost of rejection. This can be seen in Figure 19.5.
Therefore, pB may be regarded as the IQ since at pB the risk of acceptance and rejection
would reasonably be the same (i.e., 50:50). Single-sampling plans may be set up for this IQ
using the relation

c + 2/3
pB =
n

20

15
Unit cost

10
Accept
Reject

5
PB

0
0 0.01 0.02 0.03 0.04
p

FIGURE 19.5 
Costs of acceptance and rejection. (Reprinted from Enell, J.W., Qual. Control, 10, 98, 1954. With permission.)
Administration of Acceptance Sampling 595

from the Poisson distribution as discovered by Campbell (1923). This gives

c + 2/3
n=
pB

where either n or c must be fixed before the formula can be used.

Enell, however, suggests that using MIL-STD-105E, a plan be selected having the indi-
cated IQ for the sample size code letter associated with the lot size in question. This might
be done using the Schilling and Sheesley (1978) tables and the MIL-STD-105E scheme.

A = $1.00 ,  I = $0.026 ,  C = $0.50

then

.026
pB = = .052
1.00 - .50

Suppose lots of 100 are shipped so that Code F would be used. The Schilling–Sheesley
tables show that pB = .055 is associated with AQL = 2.5% for Code F. Hence, the inspector
would use Code F, 2.5 AQL for lots of 100.
Other more sophisticated models for economic determination of quality levels and
sampling plans have, of course, been presented (Smith 1965; Singh and Palanki 1976;
Liebesman 1979).

Mandatory Standards
Setting quality levels for sampling plans is sometimes regarded in terms of an adversary
relationship between the producer and the consumer. Such a relationship is more apparent
than real, however, and probably made better semantics than sense.
The approach presented here implies consideration by the consumer of both the pro-
ducer’s and consumer’s risks in setting up a sampling plan. The producer’s risk must be
given due consideration to protect the availability of supply to the consumer and forestall
price increases made necessary by a demand for unreasonable levels of quality.
The situation is analogous when sampling to mandatory standards for, in effect, the
­government represents the consumer. Where necessary, it is, after all, the state-of-the-art
fraction defective that must be improved (not legislated). A cost–benefit analysis is clearly
in order in setting quality levels for mandatory standards. Such an approach and its impli-
cations for sampling to mandatory standards have been discussed in an excellent seminal
paper by Muehlhause et al. (1975).
Quality levels should be set in terms of the state-of-the-art fraction defective to be of the
greatest impact in the marketplace. To do otherwise would be to restrict supply and raise
costs through rigid enforcement of unreasonable levels or to invite manufacturers to cheat
on the standard through nonenforcement of unrealistic demands.
It should be remembered that the quality levels used in sampling are set for cost-effective
inspection and not as targets for performance. They should be changed as appropriate to
reflect improvements in the state-of-the-art fraction defective.
596 Acceptance Sampling in Quality Control

Basic Principle of Administration

The need for simplicity and practicality in applications of acceptance sampling cannot be
overstated. In no sense should integrity be sacrificed to expedience or theory to ­intuition.
Nevertheless, it is often possible to devise simple, straightforward, theoretically correct
methods that belie the complexity of more elegant procedures. A basic principle that
applies to the administration of sampling plans is that the shop requires methods and pro-
cedures that are safe, sure, swift, and simple.
Unfortunately, straightforward mathematics seems often to lead to complicated proce-
dures (e.g., MIL-STD-414), while simple methods (e.g., No-Calc) are often found almost
intractable from the viewpoint of mathematical statistics. Nevertheless, the ultimate pur-
pose of acceptance sampling is lot inspection on the factory floor. The methods must be
understood and trusted by nonstatisticians—inspectors, operators, supervisors, and the
like—before they are used. With Tukey (1959), we would agree with Churchill Eisenhart’s
definition of the practical power of a procedure as the product of its mathematical power
and the probability that the procedure will be used.
Examples of the importance of simplicity are legion. The use of the range over the stan-
dard deviation in variables sampling is because it is easy to understand and compute.
This is also true with the X, R chart. MIL-STD-105 went from a complicated control chart
approach for switching in the A, B, and C versions to an easy counting rule in MIL-STD-
105D and MIL-STD-105E because of its simplicity. The intricacy of MIL-STD-414 or its
derivatives has done much to forestall a wider application. Multiple sampling suffers, to a
lesser extent, from the same malady.
Acceptance sampling and the accompanying forms, methods, procedures, and presen-
tations must be made as uncomplicated as possible for successful implementation in an
industry. Ease of application is not for the lazy but for the industrious strapped by the
tyranny of time. In the words of Dodge (1973), “If you want a method or system used, keep
it simple.”

Problems
1. A variables sampling plan was instituted on the thickness of germanium pellets used
in early transistors. Corrective action resulting from the feedback of information from
this inspection, as plotted on control charts, led to an extensive excellent quality his-
tory. What action should be taken?
2. About 100,000 components are manufactured each month for purchase by an origi-
nal equipment manufacturer. A certain defect in a component could pose a potential
safety problem. Accordingly, the customer has imposed a requirement that not more
than 1 in 100,000 of these components may have the defect with 90% probability.
What sampling procedure should be recommended?
3. Ten defect types are combined in the major defective category. Each has an AQL of
1%. What should be the combined AQL for the group?
4. If a 2% AOQL has been successfully used for machine screws, what might be a rea-
sonable AQL = p.95 if a sampling scheme is to be used? What might be a reasonable
LTPD?
Administration of Acceptance Sampling 597

5. Experience has shown c = 2 to be a very desirable acceptance number for both the
producer and the consumer. Using the IQ as a base, form the modified Thorndyke
chart. What are the relative values of AOQL, IQ, and LTPD?
6. A sampling plan is to be instituted on machine screws of a certain type. The unit
cost of acceptance is $60 while the cost of inspecting a piece is $1. The cost to repair a
defective unit is $6. What should be the IQ?
7. Using the results of Problem 5, if c = 2 were to be used, what would be reasonable
values of the AQL, AOQL, and LTPD in Problem 6?
8. At present, a sampling inspection plan is applied manually at a cost of $.05 per piece
using the plan n = 100, c = 2 on lots of 10,000 with a satisfactory level of outgoing
quality. A computer is available that will perform the inspection on 100% of the prod-
uct at a cost of $.001 per piece. Is it economical to purchase the computer? At the cost
of inspection, would the installation of the computer be worthwhile?
9. If replacement cost is negligible, we have pB = I/A. Using the results of Problem 5,
convert this to a formula for AQL and for LTPD.
10. It has been traditional in some industries to “take a 10% sample of the lot,” implying
c = 0. This procedure has often been impugned since protection varies with lot size,
and to defeat the plan, it is necessary only to supply smaller lot sizes. If lots rejected
under this procedure are 100% inspected, develop a formula for the AOQL from the
following formula:

yæ nö
AOQL = ç1- ÷
nè N ø

Does the result confirm or refute the criticism that protection varies with lot size?

References
American Standards Association, 1942, Control Chart Method of Controlling Quality During Production,
Z1.3, American War Standard, New York.
Bowker, A. H. and H. P. Goode, 1952, Sampling Inspection by Variables, McGraw-Hill, New York.
Campbell, G. A., 1923, Probability curves showing Poisson’s exponential summation, Bell System
Technical Journal, 2(1): 95–113.
Dodge, H. F., 1948, Administration of a sampling inspection plan, Industrial Quality Control, 5(3):
12–19.
Dodge, H. F., 1950, Inspection for quality assurance, Industrial Quality Control, 7(1): 6–10.
Dodge, H. F., 1973, Keep it simple, Quality Progress, 6(8): 11–12.
Dodge, H. F. and H. G. Romig, 1959, Sampling Inspection Tables, Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York.
Electronics Industries Association, 1949, Acceptable quality levels and how they are set up, Quality
Acceptance Bulletin No. 2, Engineering Department, Electronics Industries Association,
New York.
Enell, J. W., 1954, Which sampling plan should I choose?, Industrial Quality Control, 10(6): 96–100.
Hamaker, H. C., 1949, Lot inspection by sampling, Philips Technical Review, 11(6): 176–182.
Hamaker, H. C., J. J. M. Taudin Chabot, and F. G. Willemze, 1950, The practical application of sam-
pling inspection plans and tables, Philips Technical Review, 11(12): 362–370.
598 Acceptance Sampling in Quality Control

Kavanagh, A. J., 1946, On the selection of an inspection plan, Industrial Quality Control, 2(5): 10–11.
Liebesman, B. S., 1979, The use of MIL-STD-105D to control average outgoing quality, Journal of
Quality Technology, 11(1): 36–43.
Mentch, C. C., 1980, Manufacturing process quality optimization studies, Journal of Quality Technology,
12(3): 119–129.
Muehlhause, C. O., V. L. Broussalian, A. J. Farrar, J. W. Lyons, M. G. Natrella, J. R. Rosenblatt, R. D.
Stiehler, and J. H. Winger, 1975, Considerations in the use of sampling plans for effecting com-
pliance with mandatory safety standards, United States Department of Commerce, National
Bureau of Standards, Washington, D.C., Report 75-697.
Ott, E. R., 1975, Process Quality Control, McGraw-Hill, New York.
Peach, P., 1947, An Introduction to Industrial Statistics and Quality Control, 2nd ed., Edwards &
Broughton Company, Raleigh, NC.
Schilling, E. G., 1978, A lot sensitive sampling plan for compliance testing and acceptance sampling,
Journal of Quality Technology, 10(2): 47–51.
Schilling, E. G. and J. H. Sheesley, 1978, The performance of MIL-STD-105D under the switching
rules, Journal of Quality Technology, Part 2, 10(3): 104–124.
Singh, V. P. and H. R. Palanki, 1976, Quality levels in acceptance sampling, Journal of Quality
Technology, 8(1): 37–47.
Smith, B., 1965, The economics of sampling inspection, Industrial Quality Control, 21(9): 453–458.
Statistical Research Group, Columbia University, 1948, Sampling Inspection, McGraw-Hill, New York.
Tukey, J. W., 1959, A quick, compact, two sample test to Duckworth’s specifications, Technometrics,
1(1): 31–48.
United States Department of Defense, 1954, Guide for sampling inspection, in Quality and Reliability
Assurance Handbook (H-53), Office of the Assistant Secretary of Defense (Installations and
Logistics), Washington, DC.
United States Department of Defense, 1989, Military Standard, Sampling Procedures and Tables for
Inspection by Attributes (MIL-STD-105E), U.S. Government Printing Office, Washington, DC.
Wadsworth, H. M., 1970, The effects of class inspection under MIL-STD-105D, Journal of Quality
Technology, 2(4): 181–185 (also April 1971, p. 106).
Answers to Problems

Chapter 1
Answers to the problems of Chapter 1 are given directly in the text.

Chapter 2
1. Probability of acceptance is 49/50 = .98.
2. Yes, each tablet has an equal chance to be selected.
6!
3. P26 = = 30.
4!
6!
4. C26 = = 15; C22 = 1. Probability = 1/15.
2! 4!
5. Start

G G G G B B

2B 3G 2B 3G 2B 3G 2B 4G 1B
3G 4G 1B

6. The two draws are not independent. Should be (2/6) (1/5) = 1/15. Probability both
are good is (4/6) (3/5) = 12/30 = 6/15. Probability both are the same is 1/15 + 6/15 =
7/15. Can be added since they are mutually exclusive.
7. .95 + .95 − (.95) (.95) = .9975.
8. AQL = .017, IQ = .206, LTPD = .536 (binomial).
9. .10 + .90(.10) = .19.
10.
Probability all fail .55 = .031
Probability all pass .55 = .031
Probability at least one failure 1 − .031 = .969

599
600 Answers to Problems

Chapter 3
C598C02 98 ! 2 ! 95 ! 5 ! 98 × 97 × 96 × 95 × 94
1. f ( 0 ) = = × × = = .9020
C5100 5 ! 93 ! 0 ! 2 ! 100 ! 100 × 99 × 98 × 97 × 96
C498C12 98 ! 2 ! 5 ! 95 ! 98 ! 95 !
f ( 1) = = × × = 10 ×
C5100 4 ! 94 ! 1! 1! 100 ! 94 ! 100 !
98 × 97 × 96 × 95
= 10 × = .0960
100 × 99 × 98 × 97 × 96
F ( £ 1) = f ( 0 ) + f ( 1) = .9020 + .0960 = .9980
æ 2 ö
m = np = 5 ç ÷ = .1
è 100 ø
N -n 100 - 5
s = npq = 5 ( .02 )( .98 ) = .3067
N -1 100 - 1

2. f ( 0 ) = C05 ( .02 ) ( .98 ) = .9039

0 5

f ( 1) = C15 ( .02 ) ( .98 ) = .0922

1 4

F ( £ 1) = .9961
m = np = 5 ( .02 ) = .1
s = npq = 5 ( .02 )( .98 ) = .3130

20 e -2
3. f ( 0 ) = = .1353
0!
21 e -2
f ( 1) = = .2707
1!
22 e -2
f ( 2) = = .2707
2!
F ( £ 2 ) = .6767 , Poisson distribution not symmetric

s = 2 = 1.4142
4. b -1 ( 5|1,.05 ) = C04 .051.95 4 = .0407
2 ( .95 )
m= = 38
.05

5. F ( 20,000 ) = 1 - e -20 , 000/10 , 000 = 1 - .1353 = .8647

s = m = 10, 000 h.
6. g = 0 b = 1 h = 10, 000
F ( 10,000 ) = .6321
7. Z = -2.5
P ( £ 1.005 ) = .0062
3 1 2
C1 .0062 .9938 = .0184
Answers to Problems 601

.0003
8. sX = = .0001
9
3.001 - 3
z= = 10
.0001
Very unlikely
9. “ f -binomial”
0 2-0
æ 5 ö æ 5 ö
f ( 0 ) = C02 ç ÷ ç1- ÷ = .64
è 25 ø è 25 ø
1 2 -1
æ 5 ö æ 5 ö
f ( 1) = C12 ç ÷ ç1- ÷ = .32
è 25 ø è 25 ø
P ( £ 1) = .96
48 - 24
10. sˆ - = 12
2
48 + 24
mˆ - = 36
2

Chapter 4
1. Type B, binomial, Poisson

2.
Pa Pa
p Type A Type B
.125 .50 .5863
.25 .2143 .3164
.375 .0714 .1526
.50 .0143 .0625

3. .5500, .2720, .1154, .0385. For n > 16, they would be even closer to the Type B prob-
abilities of Problem 2.
4. μ = 2/100 units
.1353, .2707, .2707, .1804
Pa = .6767.
5. Inspector sees fraction defective .10, .20, .30, .40. Effective OC curve using binomial
distribution is

Actual p .125, .25, .375, .50

Apparent p .10, .20, .30, .40
Pa .9185, .7373, .5282, .3370
602 Answers to Problems

6. Type B
p .125 .25 .375 .50
AOQ .0366 .0396 .0286 .0156

7.
p .125 .25 .375 .50
ATI 5.65 6.73 7.39 7.75

8. Using the Poisson approximation

3
ASN C = 15F ( 2|15 ) +
.1
(1 - F ( 3|16 ) )
= 15 ( .8088 ) + 30 ( .0788 ) = 14.5

2
9. pˆ = = .20
11 - 1
10. PQL = .0166, CQL = .122.

Chapter 5

1.
Binomial Poisson
a. n = 30 c=3 n = 35 c=3
b. n = 59 c=4 n = 66 c=4
c. n = 193 c=7 n = 200 c=7

2.
a. Pa .95 .75 .50 .25 .10
p (binomial) .028 .074 .126 .194 .268
p (Poisson) .027 .074 .129 .207 .299
b. Pa .95 .75 .50 .25 .10
p (binomial) .002 .009 .021 .042 .069
p (Poisson) .002 .009 .022 .043 .071
c. Pa .95 .75 .50 .25 .10
p (binomial or Poisson) .006 .014 .021 .031 .043

3.
a. n = .2(200) = 40, c = 2
b. Pa .75 .50 .25
p .045 .065 .09

n = 33, c = 1, μ.50 = 5.1.

4.
Answers to Problems 603

5. a. n = 13, c = 1.
Pa .95 .50 .10

AOQ .027 .062 .027

ATI 62.35 506.5 901.3
AOQL = .064

b.
n = 32, c = 0.
Pa .95 .50 .10
AOQ .002 .010 .007
ATI 80.4 516 903.2
AOQL = .011

c.
n = 125, c = 2.
Pa .95 .50 .10

AOQ .005 .009 .004

ATI 168.8 562.5 912.5
AOQL = .0096

6.
n = 5, c = 0 n = 5, c = 1

Pa .95 .50 .10 Pa .95 .50 .10

p .010 .129 .369 p .076 .314 .584
AOQ .010 .062 .036 AOQ .070 .153 .057
ATI 14.75 102.5 180.5 ATI 14.75 102.5 180.5
AOQL = .072 AOQL = .164

7.
n = 5, c = 1 n = 10, c = 1

Pa .95 .50 .10 Pa .95 .50 .10

p .071 .336 .778 p .036 .168 .389
AOQ .066 .164 .076 AOQ .032 .080 .037
ATI 14.75 102.5 180.5 ATI 19.5 105.0 181.0
AOQL = .164 AOQL = .080

8.
Binomial n = 128, c = 7
Hypergeometric n = .2(500) = 100, c = 5

9.
n = 131, c = 7. The Poisson is a conservative approximation of the binomial.

æ c2 ö
10. P ( x < np ) = 1 - p ç < cn2 = 2c + 2 ÷
è 2 ø
604 Answers to Problems

Hence,

npcn2 = 2c + 2 for given probability of acceptance

Since

( ) (
F c82 < 2.73 = .05 F c82 < 15.5 = .95
)
we have
2.73 15.5
np.95 = = 1.36 np.05 = = 7.75
2 2
and
np.05 15.5
R= = = 5.68
np.95 2.73

Chapter 6
1.
a. ni = 17 Ac = 0, 3
Re = 3, 4
b. ni = 34 Ac = 1, 4
Re = 4, 5
c. ni = 120 Ac = 3, 8
Re = 7, 9

2.
a. ni = 7 Ac = #, 0, 0, 1, 2, 3, 4
Re = 2, 3, 3, 4, 4, 5, 5
b. ni = 18 Ac = #, 1, 2, 3, 5, 7, 9
Re = 4, 5, 6, 7, 8, 9, 10
c. ni = 50 Ac = 0, 1, 3, 5, 7, 10, 13
Re = 4, 6, 8, 10, 11, 12, 14

Plans a and b exceed desired R.

3.
Pa p
.95 .026
.50 .126
.10 .311

4.
Pa p
.95 .034
.50 .139
.10 .306
Answers to Problems 605

5.
Pa ASN AOQ ATI
.95 9.34 .025 58.87
.50 10.94 .063 505.47
.10 9.65 .031 900.96

6.
Pa ASN AOQ ATI
.95 10.50 .032 59.32
.50 10.92 .070 508.04
.10 7.80 .031 901.34

7.
n = 35, c = 3 has p2 = .19. Corresponding matched plans are as follows:

Double ni = 24, Ac = 1, 4 Re = 4, 5; ASN at p1 is 29.9

Multiple ni = 9, Ac = #, 0, 1, 2, 3, 4, 6 Re = 3, 3, 4, 5, 6, 6, 7; ASN at p1 is 25.4

Savings of 15% and 27% are possible.

8.
Double n = 12, c = 1.25 ~ 1
Multiple n = 12, c = 1.43 ~ 1

9. 9.21, hardly worthwhile.

10.
Sample
Defectives 1 2
2 .0176 .0297
1 .1637 .1340
0 .8187

Sample
1 2 Total

Accept Aj .8187 .1340 .9527

Reject Rj .0176 .0297 .0473
Terminate Tj .8363 .1637 1.0
Indecision Ij .1637 0 x
Pa = .9527 ASN = 9.31

Chapter 7
1.
Y2 = 1.6131 + .1018k
Y1 = −1.2565 + .1018k.
2.
Y2 = 1.6617 + .1633k
Y1 = −1.2943 + .1633k.
606 Answers to Problems

3. Y2 = 2.5348 + .0365k
Y1 = −1.9743 + .0365k.
4. Increasing slope raises probability of acceptance. Decreasing slope increases prob-
ability of rejection. Increasing h2 decreases probability of rejection. Increasing h1
decreases probability of acceptance.
5.

p Pa ASN

.05 .95 120

.0723 .56 174
.10 .10 116

6.

p ASN AOQ
.03 38 .028
.12 29 .012

7.
Y2 = 1.6284 + 2.7606k
Y1 = −1.2684 + 2.7606k.
8.
Defects/100 Defects/unit ASN/100 ASN
1 .01 .683 68.3
5.9 .059 .426 42.6

9. 459/100 units or 45.9 units.

10.
1 - b 1 - .10
A= = = 18 a = log A = 1.2552
a .05
1 - a 1 - .05
B= = = 9.5 b = log B = .99777
b .10
p .06
G1 = 2 = =6 g1 = log G1 = .7782
p1 .01
1 - p1 1 - .01
G2 = = = 1.053 g 2 = log G2 = .0225
1 - p2 1 - .06
G = g1 + g 2 = .7782 + .0225 = .8007
a 1.2552
h2 = = = 1.5678
G .8007
b .9777
h2 = = = 1.2211
G .8007
g .0225
s= 2 = = .0281
G .8007
Answers to Problems 607

Chapter 8
1.
d = 2, t-test with n = 5.
2
æ7ö
2. Chi-square. c 2 = 14 ç ÷ = 19.05 < 23.7 , accept.
è6ø
2
æ ( 1.64 + 1.28 )( 1.5 ) ö
3. n = ç ÷ = 2.1 ~ 3
è 90 - 87 ø

1.64
d= 90 - 87 = 1.68
1.64 + 1.28
Lower ACL at 90 − 1.68 = 88.32
Reject lot means 88 and 87.

n ( APL - NCL ) 3 ( 2.5 )

4. NCL = 89.5, D 2 = = = 2.887 not < .619 so n = 3.
s 1.5
5. No necessary meaning since acceptance control charts may be set up to allow a
drift in the mean.
6.
μ1 = 400 α = .025
μ2 = 408 β = .10
Y2 = 28.67 + 404k
Y1 = −18.22 + 404k

Accept on the third sample.

7. See Problem 6
Y2¢ = −28.67 − 4k Y2 = 28.67 + 4k
Y1¢ = 18.22 − 4k Y1 = −18.22 + 4k

8. Y2 = 1027.7 + 79.3k
Y1 = −800.5 + 79.3k.
9. D = 1.0, Σ(x − μ) = −1, −1, −1.17, −1.55, −1.31, −0.72, −0.20, −0.39, −0.62, −0.62.
Accept on the fourth sample.
10. tan θ = 404 so θ = 89°51′, clearly a rescaling is needed.

d = h2/s = 28.67/404 = .071

Chapter 9
1. One sample from each of the 10 compartments.
2. d = (7 − 5)/3 = .67, n = 20, need 10 more samples, 1 per compartment.
608 Answers to Problems

5.5 - 5
3. t = = 1.12 < 1.73 , accept the shipment.
2/ 20
.7
4. Testing: s32 = .7 ; Reduction: s42 = .45 - = .1
2
2 . 2
5. s22 = 2.2 − .7 = 1.5; s12 = 4.75 - = 3.65
2
6.
Source SS df MS
Between segments 228 24 9.5
Increments within segments 55 25 2.2

1.5
7. n2 = = 6.8 ~ 8 to be even.
æ ( 5 - 7 )2 3.65 .7 .1 ö
16 ç - - - ÷
ç 8.567 16 4 2÷
è ø
3.65 1.5 .7 .1
8. sX = + + + = .4648 = .68
16 128 4 2

5.9 - 5.0 .9
z= = = 1.32 < 1.645 accept the shipment
.68 .68

9.
σ2 = 3.65 + 1.5 + .7 + .1 5.9 ± 1.96 (.68)
σ2 = 5.95 5.9 ± 1.33
σ = 2.44 4.57–7.23

1.5
10. n2 = = .64 ~ 1
3.65
16 ( 1.5 + 1 ( 3.65 ) )
n1 = 2
æ 1 ö
16 ( 1) ç ÷ + ( 1) ( 3.65 )
è 1.96 ø

Chapter 10

1.
a. n = 10 k=2
b. n = 30 k=2

X = 6.83, σ = .08.
2.
Accept since 6.83 < 6.84.
3.
n = 10, M = .017, p̂ = .0125, accept.
Answers to Problems 609

M
4. = I.31 (14, 14) = .019; p̂U = I.22 (14, 14) = .0006; accept.
100
5. Take eight subgroups of 5, MAR = .92(U − L); use plan with sˆ = R/2.35; hence
X + 2s ≤ U becomes X + .85R ≤ U
6. n = 13, k = 1.63, MSD = (U − L)/3.9.
7. X = 65, s = 3.37 > 2.30 = MSD, reject.
8. For n = 13, k = 1.83 by interpolation

65 + 1.83 ( 3.37 ) = 71.2 > 69, reject :

65 - 1.83 ( 3.37 ) = 58.8 < 60, reject :

U - X 7.0 - 6.834
9. TU1 = = = 1.953 . Yes, resample.
s1 .085
10. p.50 = .0228 from k = 2.

Chapter 11
1. Code H, 1.0 AQL

a. Normal n = 50 Ac = 1 Re = 2
Tightened n = 80 Ac = 1 Re = 2
Reduced n = 20 Ac = 0 Re = 2
b. Normal ni = 32 Ac = 0, 1 Re = 2, 2
Tightened ni = 50 Ac = 0, 1 Re = 2, 2
Reduced ni = 13 Ac = 0, 0 Re = 2, 2
c. Normal ni = 13 Ac = #, #, 0, 0, 1, 1, 2 Re = 2, 2, 2, 3, 3, 3, 3
Tightened ni = 20 Ac = #, #, 0, 0, 1, 1, 2 Re = 2, 2, 2, 3, 3, 3, 3
Reduced ni = 5 Ac = #, #, 0, 0, 0, 0, 1 Re = 2, 2, 2, 3, 3, 3, 3

æ 50 ö
2. a. 1.7 ç 1 - ÷ = 1.48%
è 390 ø
æ 80 ö
b. 1.1 ç 1 - ÷ = 0.87%
è 390 ø
AOQL of tightened is about AQL.
AOQL of scheme is about AOQL tightened.
3. a. 7.6%.
b. 7.8 defects per 100 units.
4. Double.
5. a. No action.
b. No action, already back to normal.
c. Switch to normal.
610 Answers to Problems

6. a. 95%.
b. 18.3.
c. 4.0%.
d. 23.3.
7. a. .510 = .001.
b. .910 = .349.
8. C12 .11.91 = .18 .
9. 5000, 1, .015%.
10. n = 35, c = 3; Code D, 10.0 AQL.

Chapter 12
1. Code I, 1.0 AQL:
a. Normal n = 25 k = 1.85
b. Tightened n = 25 k = 1.98
c. Reduced n = 10 k = 1.58

2.
a. 2 > 1.85 Accept
b. 2 > 1.98 Accept
c. 2 > 1.58 Accept

3. Code I, 1.0 AQL

a. Normal n = 25 M = 2.86
b. Tightened n = 25 M = 2.00
c. Reduced n = 10 M = 4.77

4.
a. Normal 3.82 > 2.86 Reject
b. Tightened 3.82 > 2.00 Reject
c. Reduced 2.34 < 4.77 Accept

5. MSD = 9.52, s = 10, does not pass.

6. LTPD = 11.2%, IQ = 2.8%.
7. No action. Must have 8 lots of 10 to switch to tightened.
7 ( 4.0 ) + 3 ( 0 )
Minimum process average is = 2.8%.
10
Cannot switch to reduced, all lots must have pˆ < 3.94% . But seven lots have
pˆ > 4.0%.
8. Code M, 0.65 AQL.
9. Reject since 1.88 + 1.88 = 3.76 > M = 2.81.
10. Reject since QU = QL = 2 giving pU = pL = 2.275% and p̂ = 4.55 > 2.59.
Answers to Problems 611

Chapter 13
1.
P(i) = 97.9, 2.063 < 3.23, accept.
2. z = 2, p̂ = .023, reject.
3. NLG = 110 − 1.5(6) = 101, reject.
4.
p Zp Zg pg npg Pa (Poisson) Pa (Binomial)

.0025 2.81 .89 .19 .95 .93 .95

.034 1.82 −.10 .54 2.7 .49 .43
.109 1.23 −.69 .75 3.75 .28 .10
Note: Poisson only an approximation and at p = .109, n not large, p not small.

5.
n ≃ 25 ,  c ≃ 9 ,  t ≃ 1.15
6.
Tightened n = 14 Ac = 7 Re = 8 t = 2.17
Normal n = 13 Ac = 8 Re = 9 t = 2.27
Reduced n=9 Ac = 2 Re = 5 t = 1.43

7. a. .98
b. 12
c. .039
8. a. .42
b. 18.2
c. .021
c + 2/3
9. P0 =
n
c Formula Table
0 .0067 .0069
1 .0167 .0168
2 .0267 .0267
3 .0367 .0367
4 .0467 .0467
5 .0567 .0567

10. From Table 13.3, NLG: n = 22, c = 11, t = 1.95; single sampling plan: n = 109, c = 3.

Chapter 14
1. a. 1.7%
b. 1.1%
c. 1.8%
612 Answers to Problems

2. n = 28, c = 2.
3. Using tables in text
a. n = 19 c=1 n1 = 15 n2 = 17 c1 = 0 c2 = 2
b. n = 34 c=2 n1 = 21 n2 = 44 c1 = 0 c2 = 4
c. n = 210 c = 13 n1 = 22 n2 = 58 c1 = 0 c2 = 5

4. Using tables in text

a. n = 50 c=0 n1 = 55 n2 = 30 c1 = 0 c2 = 1
b. n = 195 c=4 n1 = 110 n2 = 195 c1 = 1 c2 = 7
c. n = 575 c = 17 n1 = 320 n2 = 585 c1 = 7 c2 = 27

5. n = 195, c = 4, ATI = 237.5.

6. n = 34, c = 2, pM = .067.
7. f1 = .3690, f2 = .1900, ASN = 54.9, AOQL = 1.1%, pM = 2.4%.
8. n = 50, c = 0, AOQL = 0.55%, Imin = 70.
9. ASN/N = f1, AOQ = 0, NAOQL = 4.69, NpM = 10.3.
10. N = 50, a = 0.02, k = 4(50) = 200, n = 9, c = 0.

Chapter 15
1. f = .10, i = 27, UAOQL = 24.3%.
2. f = .10, i = 36, UAOQL = 33.3%.
3. a. .51.
b. .53.
4. r = 2.99 ≈ 3.
5. i = 12, f = .23.
6. i = 58, f = .1, AOQL = .039.
7. N0 = 576, k = 24, f = .042, M* = 2.
8. N = 24, m = 2.
9. f = 1/7, i = 14, S = 59.
10. No.

Chapter 16
1.
n = 40, c = 2, i = 14, f = .20.
2.
Pa = .972, F = .55, ASN = 90.8, AOQL1 = .013, AOQL2 = .06.
3.
n = 160, c = 7; n1 = n2 = 98, Ac = 3, 7, Re = 8, 9;
ni = 41, Ac = 0, 1, 3, 5, 7, 10, 13, Re = 4, 6, 8, 10, 11, 12, 14.
ASN (single) = 160, ASN (double) = 119, ASN (multiple) = 106, ASNsk = 92.8.
Answers to Problems 613

4. Pa = .99, .95, .50, .10, .05, .01 have p = .003, .007, .038, .115, .150, .230, AOQL = .02,
Pa (chain) =.122.
5. n = 7, i = 2.
6. n = 8, i = 5, p.95 = .014, AOQL = .046, pM = .122.
7. n = 5, D = 47, IQ = .07.
8. n = 142.
9. US = .74, CS = 39.54; LIMITS .74 ± .597, D = 114, D/n = .114 Yes, out of control low.
10. Discontinue the criterion since sample result meets CRC2.

Chapter 17
1. n = 3150, c = 0, AOQL = .01%.
2. Pa: .95, .75, .50, .25, .10; p = .0015%, .008%, .02%, .04%, .065%.
3. n = 3900, c = 0.
4. t = 5, s = 4, n1 = 29, n2 = 6.
5. Pa = .228, ASN = 28.9, AOQ = .011.
6. CN = 1, CT = 0, n = 32, IQ = .036.
7. Sample 460 from each lot. If three lots form grand lot, then use plan n = 1380,
c = 3 on grand lots to demonstrate the compliance.
8.
s: E 5.9, 84.1, M 14.9, 75.1; X: E 444.6, 555.4, M 457.4, 542.6; accept.
9. n = 5, s: E 1.0, 89.0, M 17.1, 72.9; X: E 437.8, 562.2, M 460.6, 539.4.
10. Code letters, A, A, C, E, B, D, A, C, D, E, E; sample sizes, 192, 192, 320, 512, 256, 384,
192, 320, 384, 512, 512; lot disposition, A, R, A, A, A, A, A, R, R, A, A; stages, N, N,
N, N, N, N, N, N, T, T, T.

Chapter 18
1. The eighth-ordered unit would show h = .33, H = 1.426, P = 76.0.
2. a. .509
b. .491
c. .01
d. .7121
3. Code B-8, r = 8, c = 74.7; q̂ = 80 ≥ 74.7, accept.
4. Code B-8, r = 8, T = 37; <8 failures at 37 h, accept.
5. Code B-8, r = 8, T = 50; >8 failures at 50 h, reject.
6. Code B-8, r0 = 24, h0 = 172, h1 = −221, s = 83, V(t) = 640, continue testing.
7. p1/p0 = 2.5, r = 11, n = 35, T = 30 h.
614 Answers to Problems

8. μ.95 = 811, μ.10 = 266.

9. AQL = 130.6, LQ(10) = 89.3, LQ(5) = 86.2.
10. Code N, 1.0 AQL.

Chapter 19
1. Institute the demerit rating and consider discontinuing the inspection.
2. LSP indicates 90% of a lot of 100,000 must be sampled. To reduce the sampling
frequency to .2, D = Npt = 10.3 which implies N = 1,030,000. Use grand lot scheme
to combine 10 months production with acceptance number of zero.
3. 9.6 ~ 10%.
4. AQL = 2/1.5 = 1.3%, LTPD = 5(1.3) = 6.5%.
5. AQL: AOQL: IQ: LTPD = .3: .5: 1: 2.
6. pB = .019.
7. AQL = .57%, AOQL = .95%, LTPD = 3.8%.
8. No. Manual costs $5.00 per lot. Computer costs $10.00 per lot. Breakeven at $.10
per piece.
9. AQL = .3 I/A, LTPD = 2 I/A.
10. Use n = .1N to obtain AOQL = 3.311/N.
Appendix

615
TABLE T1.1
616

Control Chart Limits for Samples of n

Sample Sample Samplea
Standard Standard Samplea Proportions p̂
Sample Mean X Sample Mean X Deviation Sample Range Deviation s or Proportions p̂ or Defects per
against against Past against against Range R against or Defects per Unit û against
Standard μ with Data Using X Standard Standard Past Data Using Unit û against Past Data Using
Plot σ Known and s or R (Known) σ (Known) σ s or R Standard p or u p or u
Upper control m + 3s/ n B6σ D2σ B4 s
X + A3 s p (1 - p ) p (1 - p )
limit = m + As D4 R p+3 p+3
X + A2 R n n
u u
u+3 u+3
n n
Centerline μ c4σ d2σ s p p
X
R u u
Lower control m - 3s/ n B5σ D1σ B3 s
X - A3 s p (1 - p ) p (1 - p )
limit = m - As D3 R p-3 p-3
X - A2 R n n
u u
u-3 u-3
n n

n A A2 A3 B3 B4 B5 B6 C4 d2 D1 D2 D3 D4
2 2.121 1.880 2.659 0.000 3.267 0.000 2.606 0.7979 1.128 0.000 3.686 0.000 3.267
3 1.732 1.023 1.954 0.000 2.568 0.000 2.276 0.8862 1.693 0.000 4.358 0.000 2.575
4 1.500 0.729 1.628 0.000 2.266 0.000 2.088 0.9213 2.059 0.000 4.698 0.000 2.282
5 1.342 0.577 1.427 0.000 2.089 0.000 1.964 0.9400 2.326 0.000 4.918 0.000 2.115
6 1.225 0.483 1.287 0.030 1.970 0.029 1.874 0.9515 2.534 0.000 5.078 0.000 2.004
7 1.134 0.419 1.182 0.118 1.882 0.113 1.806 0.9594 2.704 0.205 5.203 0.076 1.924
8 1.061 0.373 1.099 0.185 1.815 0.179 1.751 0.9650 2.847 0.387 5.307 0.136 1.864
9 1.000 0.337 1.032 0.239 1.761 0.232 1.707 0.9693 2.970 0.546 5.394 0.184 1.816
10 0.949 0.308 0.975 0.284 1.716 0.276 1.669 0.9727 3.078 0.687 5.469 0.223 1.777
Source: Adapted from ASQC Standard A1, Definitions, Symbols, Formulas, and Tables for Control Charts, American Society for Quality Control, Milwaukee, WI, 1970.
With permission.
a For defects chart use u with n = 1.
Appendix
Appendix 617

TABLE T2.1
Random Numbers
1368 9621 9151 2066 1208 2664 9822 6599 6911 5112
5953 5936 2541 4011 0408 3593 3679 1378 5936 2651
7226 9466 9553 7671 8599 2119 5337 5953 6355 6889
8883 3454 6773 8207 5576 6386 7487 0190 0867 1298
7022 5281 1168 4099 8069 8721 8353 9952 8006 9045
4576 1853 7884 2451 3488 1286 4842 7719 5795 3953
8715 1416 7028 4616 3470 9938 5703 0196 3465 0034
4011 0408 2224 7626 0643 1149 8834 6429 8691 0143
1400 3694 4482 3608 1238 8221 5129 6105 5314 8385
6370 1884 0820 4854 9161 6509 7123 4070 6759 6113
4522 5749 8084 3932 7678 3549 0051 6761 6952 7041
7195 6234 6426 7148 9945 0358 3242 0519 6550 1327
0054 0810 2937 2040 2299 4198 0846 3937 3986 1019
5166 5433 0381 9686 5670 5129 2103 1125 3404 8785
1247 3793 7415 7819 1783 0506 4878 7673 9840 6629
8529 7842 7203 1844 8619 7404 4215 9969 6948 5643
8973 3440 4366 9242 2151 0244 0922 5887 4883 1177
9307 2959 5904 9012 4951 3695 4529 7197 7179 3239
2923 4276 9467 9868 2257 1925 3382 7244 1781 8037
6372 2808 1238 8098 5509 4617 4099 6705 2386 2830
6922 1807 4900 5306 0411 1828 8634 2331 7247 3230
9862 8336 6453 0545 6127 2741 5967 8447 3017 5709
3371 1530 5104 3076 5506 3101 4143 5845 2095 6127
6712 9402 9588 7019 9248 9192 4223 6555 7947 2474
3071 8782 7157 5941 8830 8563 2252 8109 5880 9912
4022 9734 7852 9096 0051 7387 7056 9331 1317 7833
9682 8892 3577 0326 5306 0050 8517 4376 0788 5443
6705 2175 9904 3743 1902 5393 3032 8432 0612 7972
1872 8292 2366 8603 4288 6809 4357 1072 6822 5611
2559 7534 2281 7351 2064 0611 9613 2000 0327 6145
4399 3751 9783 5399 5175 8894 0296 9483 0400 2272
6074 8827 2195 2532 7680 4288 6807 3101 6850 6410
5155 7186 4722 6721 0838 3632 5355 9369 2006 7681
3193 2800 6184 7891 9838 6123 9397 4019 8389 9508
8610 1880 7423 3384 4625 6653 2900 6290 9286 2396
4778 8818 2992 6300 4239 9595 4384 0611 7687 2088
3987 1619 4164 2542 4042 7799 9084 0278 8422 4330
2977 0248 2793 3351 4922 8878 5703 7421 2054 4391
1312 2919 8220 7285 5902 7882 1403 5354 9913 7109
3890 7193 7799 9190 3275 7840 1872 6232 5295 3148
6605 6380 4599 3333 0713 8401 7146 8940 2629 2006
8399 8175 3525 1646 4019 8390 4344 8975 4489 3423
8053 3046 9102 4515 2944 9763 3003 3408 1199 2791
9837 9378 3237 7016 7593 5958 0068 3114 0456 6840
2557 6395 9496 1884 0612 8102 4402 5498 0422 3335
(Continued)
618 Appendix

TABLE T2.1 (Continued)

Random Numbers
2671 4690 1550 2262 2597 8034 0785 2978 4409 0237
9111 0250 3275 7519 9740 4577 2064 0286 3398 1348
0391 6035 9230 4999 3332 0608 6113 0391 5789 9926
2475 2144 1886 2079 3004 9686 5669 4367 9306 2595
5336 5845 2095 6446 5694 3641 1085 8705 5416 9066
6808 0423 0155 1652 7897 4335 3567 7109 9690 3739
8525 0577 8940 9451 6726 0876 3818 7607 8854 3566
0398 0741 8787 3043 5063 0617 1770 5048 7721 7032
3623 9636 3638 1406 5731 3978 8068 7238 9715 3363
0739 2644 4917 8866 3632 5399 5175 7422 2476 2607
6713 3041 8133 8749 8835 6745 3597 3476 3816 3455
7775 9315 0432 8327 0861 1515 2297 3375 3713 9174
8599 2122 6842 9202 0810 2936 1514 2090 3067 3574
7955 3759 5254 1126 5553 4713 9605 7909 1658 5490
4766 0070 7260 6033 7997 0109 5993 7592 5436 1727
5165 1670 2534 8811 8231 3721 7947 5719 2640 1394
9111 0513 2751 8256 2931 7783 1281 6531 7259 6993
1667 1084 7889 8963 7018 8617 6381 0723 4926 4551
2145 4587 8585 2412 5431 4667 1942 7238 9613 2212
2739 5528 1481 7528 9368 1823 6979 2547 7268 2467
8769 5480 9160 5354 9700 1362 2774 7980 9157 8788
6531 9435 3422 2474 1475 0159 3414 5224 8399 5820
2937 4134 7120 2206 5084 9473 3958 7320 9878 8609
1581 3285 3727 8924 6204 0797 0882 5945 9375 9153
6268 1045 7076 1436 4165 0143 0293 4190 7171 7932
4293 0523 8625 1961 1039 2856 4889 4358 1492 3804
6936 4213 3212 7229 1230 0019 5998 9206 6753 3762
5334 7641 3258 3769 1362 2771 6124 9813 7915 8960
9373 1158 4418 8826 5665 5896 0358 4717 8232 4859
6968 9428 8950 5346 1741 2348 8143 5377 7695 0685
4229 0587 8794 4009 9691 4579 3302 7673 9629 5246
3807 7785 7097 5701 6639 0723 4819 0900 2713 7650
4891 8829 1642 2155 0796 0466 2946 2970 9143 6590
1055 2968 7911 7479 8199 9735 8271 5339 7058 2964
2983 2345 0568 4125 0894 8302 0506 6761 7706 4310
4026 3129 2968 8053 2797 4022 9838 9611 0975 2437
4075 0260 4256 0337 2355 9371 2954 6021 5783 2827
8488 5450 1327 7358 2034 8060 1788 6913 6123 9405
1976 1749 5742 4098 5887 4567 6064 2777 7830 5668
2793 4701 9466 9554 8294 2160 7486 1557 4769 2781
0916 6272 6825 7188 9611 1181 2301 5516 5451 6832
5961 1149 7946 1950 2010 0600 5655 0796 0569 4365
3222 4189 1891 8172 8731 4769 2782 1325 4238 9279
1176 7834 4600 9992 9449 5824 5344 1008 6678 1921
2369 8971 2314 4806 5071 8908 8274 4936 3357 4441
(Continued)
Appendix 619

TABLE T2.1 (Continued)

Random Numbers
0041 4329 9265 0352 4764 9070 7527 7791 1094 2008
0803 8302 6814 2422 6351 0637 0514 0246 1845 8594
9965 7804 3930 8803 0268 1426 3130 3613 3947 8086
0011 2387 3148 7559 4216 2946 2865 6333 1916 2259
1767 9871 3914 5790 5287 7915 8959 1346 5482 9251

Source: Owen, D.B., Handbook of Statistical Tables, Addison-Wesley, Reading, MA, 1962, pp. 520–521. With
permission.
620 Appendix

TABLE T3.1
Values of e−x
Units Place

0 1 2 3 4 5 6 7 8 9

Tens/ 0.00 1.0000 0.3679 0.1353 0.0498 0.0183 0.0067 0.0025 0.0009 0.0003 0.0001
hundredths 0.05 0.9512 0.3499 0.1287 0.0474 0.0174 0.0064 0.0024 0.0009 0.0003 0.0001
place 0.10 0.9048 0.3329 0.1225 0.0450 0.0166 0.0061 0.0022 0.0008 0.0003 0.0001
0.15 0.8607 0.3166 0.1165 0.0429 0.0158 0.0058 0.0021 0.0008 0.0003 0.0001
0.20 0.8187 0.3012 0.1108 0.0408 0.0150 0.0055 0.0020 0.0007 0.0003 0.0001
0.25 0.7788 0.2865 0.1054 0.0388 0.0143 0.0052 0.0019 0.0007 0.0003 0.0001
0.30 0.7408 0.2725 0.1003 0.0369 0.0136 0.0050 0.0018 0.0007 0.0002 0.0001
0.35 0.7047 0.2592 0.0954 0.0351 0.0129 0.0047 0.0017 0.0006 0.0002 0.0001
0.40 0.6703 0.2466 0.0907 0.0334 0.0123 0.0045 0.0017 0.0006 0.0002 0.0001
0.45 0.6376 0.2346 0.0863 0.0317 0.0117 0.0043 0.0016 0.0006 0.0002 0.0001
0.50 0.6065 0.2231 0.0821 0.0302 0.0111 0.0041 0.0015 0.0006 0.0002 0.0001
0.55 0.5769 0.2122 0.0781 0.0287 0.0106 0.0039 0.0014 0.0005 0.0002 0.0001
0.60 0.5488 0.2019 0.0743 0.0273 0.0101 0.0037 0.0014 0.0005 0.0002 0.0001
0.65 0.5220 0.1920 0.0707 0.0260 0.0096 0.0035 0.0013 0.0005 0.0002 0.0001
0.70 0.4966 0.1827 0.0672 0.0247 0.0091 0.0033 0.0012 0.0005 0.0002 0.0001
0.75 0.4724 0.1738 0.0639 0.0235 0.0087 0.0032 0.0012 0.0004 0.0002 0.0001
0.80 0.4493 0.1653 0.0608 0.0224 0.0082 0.0030 0.0011 0.0004 0.0002 0.0001
0.85 0.4274 0.1572 0.0578 0.0213 0.0078 0.0029 0.0011 0.0004 0.0001 0.0001
0.90 0.4066 0.1496 0.0550 0.0202 0.0074 0.0027 0.0010 0.0004 0.0001 0.0001
0.95 0.3867 0.1423 0.0523 0.0193 0.0071 0.0026 0.0010 0.0004 0.0001 0.0000
Appendix 621

TABLE T3.2
Cumulative Normal Probability, F(z)
z .00 .01 .02 .03 .04 .05 .06 .07 .08 .09

−3.5 .0002 .0002 .0002 .0002 .0002 .0002 .0002 .0002 .0002 .0002
−3.4 .0003 .0003 .0003 .0003 .0003 .0003 .0003 .0003 .0003 .0002
−3.3 .0005 .0005 .0005 .0004 .0004 .0004 .0004 .0004 .0004 .0003
−3.2 .0007 .0007 .0006 .0006 .0006 .0006 .0006 .0005 .0005 .0005
−3.1 .0010 .0009 .0009 .0009 .0008 .0008 .0008 .0008 .0007 .0007
−3.0 .0013 .0013 .0013 .0012 .0012 .0011 .0011 .0011 .0010 .0010
−2.9 .0019 .0018 .0018 .0017 .0016 .0016 .0015 .0015 .0014 .0014
−2.8 .0026 .0025 .0024 .0023 .0023 .0022 .0021 .0021 .0020 .0019
−2.7 .0035 .0034 .0033 .0032 .0031 .0030 .0029 .0028 .0027 .0026
−2.6 .0047 .0045 .0044 .0043 .0041 .0040 .0039 .0038 .0037 .0036
−2.5 .0062 .0060 .0059 .0057 .0055 .0054 .0052 .0051 .0049 .0048
−2.4 .0082 .0080 .0078 .0075 .0073 .0071 .0069 .0068 .0066 .0064
−2.3 .0107 .0104 .0102 .0099 .0096 .0094 .0091 .0089 .0087 .0084
−2.2 .0139 .0136 .0132 .0129 .0125 .0122 .0119 .0116 .0113 .0110
−2.1 .0179 .0174 .0170 .0166 .0162 .0158 .0154 .0150 .0146 .0143
−2.0 .0228 .0222 .0217 .0212 .0207 .0202 .0197 .0192 .0188 .0183
−1.9 .0287 .0281 .0274 .0268 .0262 .0256 .0250 .0244 .0239 .0233
−1.8 .0359 .0351 .0344 .0336 .0329 .0322 .0314 .0307 .0301 .0294
−1.7 .0446 .0436 .0427 .0418 .0409 .0401 .0392 .0384 .0375 .0367
−1.6 .0548 .0537 .0526 .0516 .0505 .0495 .0485 .0475 .0465 .0455
−1.5 .0668 .0655 .0643 .0630 .0618 .0606 .0594 .0582 .0571 .0559
−1.4 .0808 .0793 .0778 .0764 .0749 .0735 .0721 .0708 .0694 .0681
−1.3 .0968 .0951 .0934 .0918 .0901 .0885 .0869 .0853 .0838 .0823
−1.2 .1151 .1131 .1112 .1093 .1075 .1056 .1038 .1020 .1003 .0985
−1.1 .1357 .1335 .1314 .1292 .1271 .1251 .1230 .1210 .1190 .1170
−1.0 .1587 .1562 .1539 .1515 .1492 .1469 .1446 .1423 .1401 .1379
−0.9 .1841 .1814 .1788 .1762 .1736 .1711 .1685 .1660 .1635 .1611
−0.8 .2119 .2090 .2061 .2033 .2005 .1977 .1949 .1922 .1894 .1867
−0.7 .2420 .2389 .2358 .2327 .2296 .2266 .2236 .2206 .2177 .2148
−0.6 .2743 .2709 .2676 .2643 .2611 .2578 .2546 .2514 .2483 .2451
−0.5 .3085 .3050 .3015 .2981 .2946 .2912 .2877 .2843 .2810 .2776
−0.4 .3446 .3409 .3372 .3336 .3300 .3264 .3228 .3192 .3156 .3121
−0.3 .3821 .3783 .3745 .3707 .3669 .3632 .3594 .3557 .3520 .3483
−0.2 .4207 .4168 .4129 .4090 .4052 .4013 .3974 .3936 .3897 .3859
−0.1 .4602 .4562 .4522 .4483 .4443 .4404 .4364 .4325 .4286 .4247
−0.0 .5000 .4960 .4920 .4880 .4840 .4801 .4761 .4721 .4681 .4641
+0.0 .5000 .5040 .5080 .5121 .5160 .5199 .5239 .5279 .5319 .5359
+0.1 .5398 .5438 .5478 .5517 .5557 .5596 .5636 .5675 .5714 .5753
+0.2 .5793 .5832 .5871 .5910 .5948 .5987 .6026 .6064 .6103 .6141
+0.3 .6179 .6217 .6255 .6293 .6331 .6368 .6406 .6443 .6480 .6517
+0.4 .6554 .6591 .6628 .6664 .6700 .6736 .6772 .6808 .6844 .6879
+0.5 .6915 .6950 .6985 .7019 .7054 .7088 .7123 .7157 .7190 .7224
+0.6 .7257 .7291 .7324 .7357 .7389 .7422 .7454 .7486 .7517 .7549
(Continued)
622 Appendix

TABLE T3.2 (Continued)

Cumulative Normal Probability, F(z)
z .00 .01 .02 .03 .04 .05 .06 .07 .08 .09
+0.7 .7580 .7611 .7642 .7673 .7704 .7734 .7764 .7794 .7823 .7852
+0.8 .7881 .7910 .7939 .7967 .7995 .8023 .8051 .8078 .8106 .8133
+0.9 .8159 .8186 .8212 .8238 .8264 .8289 .8315 .8340 .8365 .8389
+1.0 .8413 .8438 .8461 .8485 .8508 .8531 .8554 .8577 .8599 .8621
+1.1 .8643 .8665 .8686 .8708 .8729 .8749 .8770 .8790 .8810 .8830
+1.2 .8849 .8869 .8888 .8907 .8925 .8944 .8962 .8980 .8997 .9015
+1.3 .9032 .9049 .9066 .9082 .9099 .9115 .9131 .9147 .9162 .9177
+1.4 .9192 .9207 .9222 .9236 .9251 .9265 .9279 .9292 .9306 .9319
+1.5 .9332 .9345 .9357 .9370 .9382 .9394 .9406 .9418 .9429 .9441
+1.6 .9452 .9463 .9474 .9484 .9495 .9505 .9515 .9525 .9535 .9545
+1.7 .9554 .9564 .9573 .9582 .9591 .9599 .9608 .9616 .9625 .9633
+1.8 .9641 .9649 .9656 .9664 .9671 .9678 .9686 .9693 .9699 .9706
+1.9 .9713 .9719 .9726 .9732 .9738 .9744 .9750 .9756 .9761 .9767
+2.0 .9772 .9778 .9783 .9788 .9793 .9798 .9803 .9808 .9812 .9817
+2.1 .9821 .9826 .9830 .9834 .9838 .9842 .9846 .9850 .9854 .9857
+2.2 .9861 .9864 .9868 .9871 .9875 .9878 .9881 .9884 .9887 .9890
+2.3 .9893 .9896 .9898 .9901 .9904 .9906 .9909 .9911 .9913 .9916
+2.4 .9918 .9920 .9922 .9925 .9927 .9929 .9931 .9932 .9934 .9936
+2.5 .9938 .9940 .9941 .9943 .9945 .9946 .9948 .9949 .9951 .9952
+2.6 .9953 .9955 .9956 .9957 .9959 .9960 .9961 .9962 .9963 .9964
+2.7 .9965 .9966 .9967 .9968 .9969 .9970 .9971 .9972 .9973 .9974
+2.8 .9974 .9975 .9976 .9977 .9977 .9978 .9979 .9979 .9980 .9981
+2.9 .9981 .9982 .9982 .9983 .9984 .9984 .9985 .9985 .9986 .9986
+3.0 .9987 .9987 .9987 .9988 .9988 .9989 .9989 .9989 .9990 .9990
+3.1 .9990 .9991 .9991 .9991 .9992 .9992 .9992 .9992 .9993 .9993
+3.2 .9993 .9993 .9994 .9994 .9994 .9994 .9994 .9995 .9995 .9995
+3.3 .9995 .9995 .9995 .9996 .9996 .9996 .9996 .9996 .9996 .9997
+3.4 .9997 .9997 .9997 .9997 .9997 .9997 .9997 .9997 .9997 .9998
+3.5 .9998 .9998 .9998 .9998 .9998 .9998 .9998 .9998 .9998 .9998
Source: Burr, I.W., Engineering Statistics and Quality Control, McGraw-Hill, New York,
1953. With permission.
TABLE T3.3
Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)
Appendix

N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

2 1 1 0 0.500000 0.500000 6 2 2 2 1.000000 0.066667 7 4 2 1 0.714286 0.571429 8 3 3 2 0.982143 0.267857

2 1 1 1 1.000000 0.500000 6 3 1 0 0.500000 0.500000 7 4 2 2 1.000000 0.285714 8 3 3 3 1.000000 0.017857
3 1 1 0 0.666667 0.666667 6 3 1 1 1.000000 0.500000 7 4 3 0 0.028571 0.028571 8 4 1 0 0.500000 0.500000
3 1 1 1 1.000000 0.333333 6 3 2 0 0.200000 0.200000 7 4 3 1 0.371429 0.342857 8 4 1 1 1.000000 0.500000
3 2 1 0 0.333333 0.333333 6 3 2 1 0.800000 0.600000 7 4 3 2 0.885714 0.514286 8 4 2 0 0.214286 0.214286
3 2 2 1 1.000000 0.666667 6 3 2 2 1.000000 0.200000 7 4 3 3 1.000000 0.114286 8 4 2 1 0.785714 0.571429
3 2 2 1 0.666667 0.666667 6 3 3 0 0.050000 0.050000 7 4 4 1 0.114286 0.114286 8 4 2 2 1.000000 0.214286
3 2 2 2 1.000000 0.333333 6 3 3 1 0.500000 0.450000 7 4 4 2 0.628571 0.514286 8 4 3 0 0.071429 0.071429
4 1 1 0 0.750000 0.750000 6 3 3 2 0.950000 0.450000 7 4 4 3 0.971428 0.342857 8 4 3 1 0.500000 0.428571
4 1 1 1 1.000000 0.250000 6 3 3 3 1.000000 0.050000 7 4 4 4 1.000000 0.028571 8 4 3 2 0.928571 0.428571
4 2 1 0 0.500000 0.500000 6 4 1 0 0.333333 0.333333 7 5 1 0 0.285714 0.285714 8 4 3 3 1.000000 0.071429
4 2 1 1 1.000000 0.500000 6 4 1 1 1.000000 0.666667 7 5 1 1 1.000000 0.714286 8 4 4 0 0.014286 0.014286
4 2 2 0 0.166667 0.166667 6 4 2 0 0.066667 0.066667 7 5 2 0 0.047619 0.047619 8 4 4 1 0.242857 0.228571
4 2 2 1 0.833333 0.666667 6 4 2 1 0.600000 0.533333 7 5 2 1 0.523809 0.476190 8 4 4 2 0.757143 0.514286
4 2 2 2 1.000000 0.166667 6 4 2 2 1.000000 0.400000 7 5 2 2 1.000000 0.476190 8 4 4 3 0.985714 0.228571
4 3 1 0 0.250000 0.250000 6 4 3 1 0.200000 0.200000 7 5 3 1 0.142857 0.142857 8 4 4 4 1.000000 0.014286
4 3 1 1 1.000000 0.750000 6 4 3 2 0.800000 0.600000 7 5 3 2 0.714286 0.571429 8 5 1 0 0.375000 0.375000
4 3 2 1 0.500000 0.500000 6 4 3 3 1.000000 0.200000 7 5 3 3 1.000000 0.285714 8 5 1 1 1.000000 0.625000
4 3 2 2 1.000000 0.500000 6 4 4 2 0.400000 0.400000 7 5 4 2 0.285714 0.285714 8 5 2 0 0.107143 0.107143
4 3 3 2 0.750000 0.750000 6 4 4 3 0.933333 0.533333 7 5 4 3 0.857143 0.571429 8 5 2 1 0.642857 0.535714
4 3 3 3 1.000000 0.250000 6 4 4 4 1.000000 0.066667 7 5 4 4 1.000000 0.142857 8 5 2 2 1.000000 0.357143
5 1 1 0 0.800000 0.800000 6 5 1 0 0.166667 0.166667 7 5 5 3 0.476190 0.476190 8 5 3 0 0.017857 0.017857
5 1 1 1 1.000000 0.200000 6 5 1 1 1.000000 0.833333 7 5 5 4 0.952381 0.476190 8 5 3 1 0.285714 0.267857
5 2 1 0 0.600000 0.600000 6 5 2 1 0.333333 0.333333 7 5 5 5 1.000000 0.047619 8 5 3 2 0.821429 0.535714
5 2 1 1 1.000000 0.400000 6 5 2 2 1.000000 0.666667 7 6 1 0 0.142857 0.142857 8 5 3 3 1.000000 0.178571
5 2 2 0 0.300000 0.300000 6 5 3 2 0.500000 0.500000 7 6 1 1 1.000000 0.857143 8 5 4 1 0.071429 0.071429
5 2 2 1 0.900000 0.600000 6 5 3 3 1.000000 0.500000 7 6 2 1 0.285714 0.285714 8 5 4 2 0.500000 0.428571
5 2 2 2 1.000000 0.100000 6 5 4 3 0.666667 0.666667 7 6 2 2 1.000000 0.714286 8 5 4 3 0.928571 0.428571
(Continued)
623
TABLE T3.3 (Continued)
624

Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)

N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

5 3 2 2 1.000000 0.300000 7 1 1 1 1.000000 0.142857 7 6 5 4 0.714286 0.714286 8 5 5 5 1.000000 0.017857

5 3 3 1 0.300000 0.300000 7 2 1 0 0.714286 0.714286 7 6 5 5 1.000000 0.285714 8 6 1 0 0.250000 0.250000
5 3 3 2 0.900000 0.600000 7 2 1 1 1.000000 0.285714 7 6 6 5 0.857143 0.857143 8 6 1 1 1.000000 0.750000
5 3 3 3 1.000000 0.100000 7 2 2 0 0.476190 0.476190 7 6 6 6 1.000000 0.142857 8 6 2 0 0.035714 0.035714
5 4 1 0 0.200000 0.200000 7 2 2 1 0.952381 0.476190 8 1 1 0 0.875000 0.875000 8 6 2 1 0.464286 0.428571
5 4 1 1 1.000000 0.800000 7 2 2 2 1.000000 0.047619 8 1 1 1 1.000000 0.125000 8 6 2 2 1.000000 0.535714
5 4 2 1 0.400000 0.400000 7 3 1 0 0.571429 0.571429 8 2 1 0 0.750000 0.750000 8 6 3 1 0.107143 0.107143
5 4 2 2 1.000000 0.600000 7 3 1 1 1.000000 0.428571 8 2 1 1 1.000000 0.250000 8 6 3 2 0.642857 0.535714
5 4 3 2 0.600000 0.600000 7 3 2 0 0.285714 0.285714 8 2 2 0 0.535714 0.535714 8 6 3 3 1.000000 0.357143
5 4 3 3 1.000000 0.400000 7 3 2 1 0.857143 0.571429 8 2 2 1 0.964286 0.428571 8 6 4 2 0.214286 0.214286
5 4 4 3 0.800000 0.800000 7 3 2 2 1.000000 0.142857 8 2 2 2 1.000000 0.035714 8 6 4 3 0.785714 0.571429
5 4 4 4 1.000000 0.200000 7 3 3 0 0.114286 0.114286 8 3 1 0 0.625000 0.625000 8 6 4 4 1.000000 0.214286
6 1 1 0 0.833333 0.833333 7 3 3 1 0.628571 0.514286 8 3 1 1 1.000000 0.375000 8 6 5 3 0.357143 0.357143
6 1 1 1 1.000000 0.166667 7 3 3 2 0.971428 0.342857 8 3 2 0 0.357143 0.357143 8 6 5 4 0.892857 0.535714
6 2 1 0 0.666667 0.666667 7 3 3 3 1.000000 0.028571 8 3 2 1 0.892857 0.535714 8 6 5 5 1.000000 0.107143
6 2 1 1 1.000000 0.333333 7 4 1 0 0.428571 0.428571 8 3 2 2 1.000000 0.107143 8 6 6 4 0.535714 0.535714
6 2 2 0 0.400000 0.400000 7 4 1 1 1.000000 0.571429 8 3 3 0 0.178571 0.178571 8 6 6 5 0.964286 0.428571
6 2 2 1 0.933333 0.533333 7 4 2 0 0.142857 0.142857 8 3 3 1 0.714286 0.535714 8 6 6 6 1.000000 0.035714
N = 2–8
8 7 1 0 0.125000 0.125000 9 5 3 1 0.404762 0.357143 9 7 6 6 1.000000 0.083333 10 5 1 0 0.500000 0.500000
8 7 1 1 1.000000 0.875000 9 5 3 2 0.880952 0.476190 9 7 7 5 0.583333 0.583333 10 5 1 1 1.000000 0.500000
8 7 2 1 0.250000 0.250000 9 5 3 3 1.000000 0.119048 9 7 7 6 0.972222 0.388889 10 5 2 0 0.222222 0.222222
8 7 2 2 1.000000 0.750000 9 5 4 0 0.007936 0.007936 9 7 7 7 1.000000 0.027778 10 5 2 1 0.777778 0.555556
8 7 3 2 0.375000 0.375000 9 5 4 1 0.166667 0.158730 9 8 1 0 0.111111 0.111111 10 5 2 2 1.000000 0.222222
8 7 3 3 1.000000 0.625000 9 5 4 2 0.642857 0.476190 9 8 1 1 1.000000 0.888889 10 5 3 0 0.083333 0.083333
8 7 4 3 0.500000 0.500000 9 5 4 3 0.960317 0.317460 9 8 2 1 0.222222 0.222222 10 5 3 1 0.500000 0.416667
8 7 4 4 1.000000 0.500000 9 5 4 4 1.000000 0.039683 9 8 2 2 1.000000 0.777778 10 5 3 2 0.916667 0.416667
8 7 5 4 0.625000 0.625000 9 5 5 1 0.039683 0.039683 9 8 3 2 0.333333 0.333333 10 5 3 3 1.000000 0.083333
(Continued)
Appendix
TABLE T3.3 (Continued)
Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)
Appendix

N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

8 7 5 5 1.000000 0.375000 9 5 5 2 0.357143 0.317460 9 8 3 3 1.000000 0.666667 10 5 4 0 0.023810 0.023810

8 7 6 5 0.750000 0.750000 9 5 5 3 0.833333 0.476190 9 8 4 3 0.444444 0.444444 10 5 4 1 0.261905 0.238095
8 7 6 6 1.000000 0.250000 9 5 5 4 0.992063 0.158730 9 8 4 4 1.000000 0.555556 10 5 4 2 0.738095 0.476190
8 7 7 6 0.875000 0.875000 9 5 5 5 1.000000 0.007936 9 8 5 4 0.555556 0.555556 10 5 4 3 0.976190 0.238095
8 7 7 7 1.000000 0.125000 9 6 1 0 0.333333 0.333333 9 8 5 5 1.000000 0.444444 10 5 4 4 1.000000 0.023810
9 1 1 0 0.888889 0.888889 9 6 1 1 1.000000 0.666667 9 8 6 5 0.666667 0.666667 10 5 5 0 0.003968 0.003968
9 1 1 1 1.000000 0.111111 9 6 2 0 0.083333 0.083333 9 8 6 6 1.000000 0.333333 10 5 5 1 0.103175 0.099206
9 2 1 0 0.777778 0.777778 9 6 2 1 0.583333 0.500000 9 8 7 6 0.777778 0.777778 10 5 5 2 0.500000 0.396825
9 2 1 1 1.000000 0.222222 9 6 2 2 1.000000 0.416667 9 8 7 7 1.000000 0.222222 10 5 5 3 0.896825 0.396825
9 2 2 0 0.583333 0.583333 9 6 3 0 0.011905 0.011905 9 8 8 7 0.888889 0.888889 10 5 5 4 0.996032 0.099206
9 2 2 1 0.972222 0.388889 9 6 3 1 0.226190 0.214286 9 8 8 8 1.000000 0.111111 10 5 5 5 1.000000 0.003968
9 2 2 2 1.000000 0.027778 9 6 3 2 0.761905 0.535714 10 1 1 0 0.900000 0.900000 10 6 1 0 0.400000 0.400000
9 3 1 0 0.666667 0.666667 9 6 3 3 1.000000 0.238095 10 1 1 1 1.000000 0.100000 10 6 1 1 1.000000 0.600000
9 3 1 1 1.000000 0.333333 9 6 4 1 0.047619 0.047619 10 2 1 0 0.800000 0.800000 10 6 2 0 0.133333 0.133333
9 3 2 0 0.416667 0.416667 9 6 4 2 0.404762 0.357143 10 2 1 1 1.000000 0.200000 10 6 2 1 0.666667 0.533333
9 3 2 1 0.916667 0.500000 9 6 4 3 0.880952 0.476190 10 2 2 0 0.622222 0.622222 10 6 2 2 1.000000 0.333333
9 3 2 2 1.000000 0.083333 9 6 4 4 1.000000 0.119048 10 2 2 1 0.977778 0.355556 10 6 3 0 0.033333 0.033333
9 3 3 0 0.238095 0.238095 9 6 5 2 0.119048 0.119048 10 2 2 2 1.000000 0.022222 10 6 3 1 0.333333 0.300000
9 3 3 1 0.773809 0.535714 9 6 5 3 0.595238 0.476190 10 3 1 0 0.700000 0.700000 10 6 3 2 0.833333 0.500000
9 3 3 2 0.988095 0.214286 9 6 5 4 0.952381 0.357143 10 3 1 1 1.000000 0.300000 10 6 3 3 1.000000 0.166667
9 3 3 3 1.000000 0.011905 9 6 5 5 1.000000 0.047619 10 3 2 0 0.466667 0.466667 10 6 4 0 0.004762 0.004762
9 4 1 0 0.555556 0.555556 9 6 6 3 0.238095 0.238095 10 3 2 1 0.933333 0.466667 10 6 4 1 0.119048 0.114286
9 4 1 1 1.000000 0.444444 9 6 6 4 0.773809 0.535714 10 3 2 2 1.000000 0.066667 10 6 4 2 0.547619 0.428571
9 4 2 0 0.277778 0.277778 9 6 6 5 0.988095 0.214286 10 3 3 0 0.291667 0.291667 10 6 4 3 0.928571 0.380952
9 4 2 1 0.833333 0.555556 9 6 6 6 1.000000 0.011905 10 3 3 1 0.816667 0.525000 10 6 4 4 1.000000 0.071429
9 4 2 2 1.000000 0.166667 9 7 1 0 0.222222 0.222222 10 3 3 2 0.991667 0.175000 10 6 5 1 0.023810 0.023810
9 4 3 0 0.119048 0.119048 9 7 1 1 1.000000 0.777778 10 3 3 3 1.000000 0.008333 10 6 5 2 0.261905 0.238095
(Continued)
625
 626
TABLE T3.3 (Continued)
Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)
N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

9 4 3 1 0.595238 0.476190 9 7 2 0 0.027778 0.027778 10 4 1 0 0.600000 0.600000 10 6 5 3 0.738095 0.476190

9 4 3 2 0.952381 0.357143 9 7 2 1 0.416667 0.388889 10 4 1 1 1.000000 0.400000 10 6 5 4 0.976190 0.238095
9 4 3 3 1.000000 0.047619 9 7 2 2 1.000000 0.583333 10 4 2 0 0.333333 0.333333 10 6 5 5 1.000000 0.023810
9 4 4 0 0.039683 0.039683 9 7 3 1 0.083333 0.083333 10 4 2 1 0.866667 0.533333 10 6 6 2 0.071429 0.071429
9 4 4 1 0.357143 0.317460 9 7 3 2 0.583333 0.500000 10 4 2 2 1.000000 0.133333 10 6 6 3 0.452381 0.380952
9 4 4 2 0.833333 0.476190 9 7 3 3 1.000000 0.416667 10 4 3 0 0.166667 0.166667 10 6 6 4 0.880952 0.428571
9 4 4 3 0.992063 0.158730 9 7 4 2 0.166667 0.166667 10 4 3 1 0.666667 0.500000 10 6 6 5 0.995238 0.114286
9 4 4 4 1.000000 0.007936 9 7 4 3 0.722222 0.555556 10 4 3 2 0.966667 0.300000 10 6 6 6 1.000000 0.004762
9 5 1 0 0.444444 0.444444 9 7 4 4 1.000000 0.277778 10 4 3 3 1.000000 0.033333 10 7 1 0 0.300000 0.300000
9 5 1 1 1.000000 0.555556 9 7 5 3 0.277778 0.277778 10 4 4 0 0.071429 0.071429 10 7 1 1 1.000000 0.700000
9 5 2 0 0.166667 0.166667 9 7 5 4 0.833333 0.555556 10 4 4 1 0.452381 0.380952 10 7 2 0 0.066667 0.066667
9 5 2 1 0.722222 0.555556 9 7 5 5 1.000000 0.166667 10 4 4 2 0.880952 0.428571 10 7 2 1 0.533333 0.466667
9 5 2 2 1.000000 0.277778 9 7 6 4 0.416667 0.416667 10 4 4 3 0.995238 0.114286 10 7 2 2 1.000000 0.466667
9 5 3 0 0.047619 0.047619 9 7 6 5 0.916667 0.500000 10 4 4 4 1.000000 0.004762 10 7 3 0 0.008333 0.008333
N = 8–10
10 7 3 1 0.183333 0.175000 10 9 5 4 0.500000 0.500000 11 5 4 1 0.348485 0.303030 11 7 5 1 0.015152 0.015152
10 7 3 2 0.708333 0.525000 10 9 5 5 1.000000 0.500000 11 5 4 2 0.803030 0.454545 11 7 5 2 0.196970 0.181818
10 7 3 3 1.000000 0.291667 10 9 6 5 0.600000 0.600000 11 5 4 3 0.984848 0.181818 11 7 5 3 0.651515 0.454545
10 7 4 1 0.033333 0.033333 10 9 6 6 1.000000 0.400000 11 5 4 4 1.000000 0.015152 11 7 5 4 0.954545 0.303030
10 7 4 2 0.333333 0.300000 10 9 7 6 0.700000 0.700000 11 5 5 0 0.012987 0.012987 11 7 5 5 1.000000 0.045455
10 7 4 3 0.833333 0.500000 10 9 7 7 1.000000 0.300000 11 5 5 1 0.175325 0.162338 11 7 6 2 0.045455 0.045455
10 7 4 4 1.000000 0.166667 10 9 8 7 0.800000 0.800000 11 5 5 2 0.608225 0.432900 11 7 6 3 0.348485 0.303030
10 7 5 2 0.083333 0.083333 10 9 8 8 1.000000 0.200000 11 5 5 3 0.932900 0.324675 11 7 6 4 0.803030 0.454545
10 7 5 3 0.500000 0.416667 10 9 9 8 0.900000 0.900000 11 5 5 4 0.997835 0.064935 11 7 6 5 0.984848 0.181818
10 7 5 4 0.916667 0.416667 10 9 9 9 1.000000 0.100000 11 5 5 5 1.000000 0.002164 11 7 6 6 1.000000 0.015152
10 7 5 5 1.000000 0.083333 11 1 1 0 0.909091 0.909091 11 6 1 0 0.454545 0.454545 11 7 7 3 0.106061 0.106061
10 7 6 3 0.166667 0.166667 11 1 1 1 1.000000 0.090909 11 6 1 1 1.000000 0.545455 11 7 7 4 0.530303 0.424242
10 7 6 4 0.666667 0.500000 11 2 1 0 0.818182 0.818182 11 6 2 0 0.181818 0.181818 11 7 7 5 0.912121 0.381818
(Continued)
Appendix
TABLE T3.3 (Continued)
Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)
Appendix

N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

10 7 6 5 0.966667 0.300000 11 2 1 1 1.000000 0.181818 11 6 2 1 0.727273 0.545455 11 7 7 6 0.996970 0.084848

10 7 6 6 1.000000 0.033333 11 2 2 0 0.654545 0.654545 11 6 2 2 1.000000 0.272727 11 7 7 7 1.000000 0.003030
10 7 7 4 0.291667 0.291667 11 2 2 1 0.981818 0.327273 11 6 3 0 0.060606 0.060606 11 8 1 0 0.272727 0.272727
10 7 7 5 0.816667 0.525000 11 2 2 2 1.000000 0.018182 11 6 3 1 0.424242 0.363636 11 8 1 1 1.000000 0.727273
10 7 7 6 0.991667 0.175000 11 3 1 0 0.727273 0.727273 11 6 3 2 0.878788 0.454545 11 8 2 0 0.054545 0.054545
10 7 7 7 1.000000 0.008333 11 3 1 1 1.000000 0.272727 11 6 3 3 1.000000 0.121212 11 8 2 1 0.490909 0.436364
10 8 1 0 0.200000 0.200000 11 3 2 0 0.509091 0.509091 11 6 4 0 0.015152 0.015152 11 8 2 2 1.000000 0.509091
10 8 1 1 1.000000 0.800000 11 3 2 1 0.945455 0.436364 11 6 4 1 0.196970 0.181818 11 8 3 0 0.006061 0.006061
10 8 2 0 0.022222 0.022222 11 3 2 2 1.000000 0.054545 11 6 4 2 0.651515 0.454545 11 8 3 1 0.151515 0.145455
10 8 2 1 0.377778 0.355556 11 3 3 0 0.339394 0.339394 11 6 4 3 0.954545 0.303030 11 8 3 2 0.660606 0.509091
10 8 2 2 1.000000 0.622222 11 3 3 1 0.848485 0.509091 11 6 4 4 1.000000 0.045455 11 8 3 3 1.000000 0.339394
10 8 3 1 0.066667 0.066667 11 3 3 2 0.993939 0.145455 11 6 5 0 0.002164 0.002164 11 8 4 1 0.024242 0.024242
10 8 3 2 0.533333 0.466667 11 3 3 3 1.000000 0.006061 11 6 5 1 0.067100 0.064935 11 8 4 2 0.278788 0.254545
10 8 3 3 1.000000 0.466667 11 4 1 0 0.636364 0.636364 11 6 5 2 0.391775 0.324675 11 8 4 3 0.787879 0.509091
10 8 4 2 0.133333 0.133333 11 4 1 1 1.000000 0.363636 11 6 5 3 0.824675 0.432900 11 8 4 4 1.000000 0.212121
10 8 4 3 0.666667 0.533333 11 4 2 0 0.381818 0.381818 11 6 5 4 0.987013 0.162338 11 8 5 2 0.060606 0.060606
10 8 4 4 1.000000 0.333333 11 4 2 1 0.890909 0.509091 11 6 5 5 1.000000 0.012987 11 8 5 3 0.424242 0.363636
10 8 5 3 0.222222 0.222222 11 4 2 2 1.000000 0.109091 11 6 6 1 0.012987 0.012987 11 8 5 4 0.878788 0.454545
10 8 5 4 0.777778 0.555556 11 4 3 0 0.212121 0.212121 11 6 6 2 0.175325 0.162338 11 8 5 5 1.000000 0.121212
10 8 5 5 1.000000 0.222222 11 4 3 1 0.721212 0.509091 11 6 6 3 0.608225 0.432900 11 8 6 3 0.121212 0.121212
10 8 6 4 0.333333 0.333333 11 4 3 2 0.975758 0.254545 11 6 6 4 0.932900 0.324675 11 8 6 4 0.575758 0.454545
10 8 6 5 0.866667 0.533333 11 4 3 3 1.000000 0.024242 11 6 6 5 0.997835 0.064935 11 8 6 5 0.939394 0.363636
10 8 6 6 1.000000 0.133333 11 4 4 0 0.106061 0.106061 11 6 6 6 1.000000 0.002164 11 8 6 6 1.000000 0.060606
10 8 7 5 0.466667 0.466667 11 4 4 1 0.530303 0.424242 11 7 1 0 0.363636 0.363636 11 8 7 4 0.212121 0.212121
10 8 7 6 0.933333 0.466667 11 4 4 2 0.912121 0.381818 11 7 1 1 1.00000 0.636364 11 8 7 5 0.721212 0.509091
10 8 7 7 1.000000 0.066667 11 4 4 3 0.996970 0.084848 11 7 2 0 0.109091 0.109091 11 8 7 6 0.975758 0.254545
10 8 8 6 0.622222 0.622222 11 4 4 4 1.000000 0.003030 11 7 2 1 0.618182 0.509091 11 8 7 7 1.000000 0.024242
10 8 8 7 0.977778 0.355556 11 5 1 0 0.545455 0.545455 11 7 2 2 1.000000 0.381818 11 8 8 5 0.339394 0.339394
(Continued)
627
 628

TABLE T3.3 (Continued)

Lieberman–Owen Table of Hypergeometric Distribution, F(x) (N = 2, n = 1 through N = 11, n = 9)
N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x) N n k x P(x) p(x)

10 8 8 8 1.000000 0.022222 11 5 1 1 1.000000 0.454545 11 7 3 0 0.024242 0.024242 11 8 8 6 0.848485 0.509091

10 9 1 0 0.100000 0.100000 11 5 2 0 0.272727 0.272727 11 7 3 1 0.278788 0.254545 11 8 8 7 0.993939 0.145455
10 9 1 1 1.000000 0.900000 11 5 2 1 0.818182 0.545455 11 7 3 2 0.787879 0.509091 11 8 8 8 1.000000 0.006061
10 9 2 1 0.200000 0.200000 11 5 2 2 1.000000 0.181818 11 7 3 3 1.000000 0.212121 11 9 1 0 0.181818 0.181818
10 9 2 2 1.000000 0.800000 11 5 3 0 0.121212 0.121212 11 7 4 0 0.003030 0.003030 11 9 1 1 1.000000 0.818182
10 9 3 2 0.300000 0.300000 11 5 3 1 0.575758 0.454545 11 7 4 1 0.087879 0.084848 11 9 2 0 0.018182 0.018182
10 9 3 3 1.000000 0.700000 11 5 3 2 0.939394 0.363636 11 7 4 2 0.469697 0.381818 11 9 2 1 0.345454 0.327273
10 9 4 3 0.400000 0.400000 11 5 3 3 1.000000 0.060606 11 7 4 3 0.893939 0.424242 11 9 2 2 1.000000 0.654545
10 9 4 4 1.000000 0.600000 11 5 4 0 0.045455 0.045455 11 7 4 4 1.000000 0.106061 11 9 3 1 0.054545 0.054545
N = 10–11

Source: Reprinted from Lieberman, G.J. and Owen, D.B., Tables of the Hypergeometric Probability Distribution, Stanford University Press, Stanford, CA, pp. 33–35.
With permission Copyright 1961 by the Board of Trustees of the Leland Stanford Junior University.
Appendix
Appendix 629

TABLE T3.4
Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

1 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.01000 0.02000 0.03000 0.04000 0.05000 0.06000 0.06250 0.07000 0.08000 0.08333
2 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.01990 0.03960 0.05910 0.07840 0.09750 0.11640 0.12109 0.13510 0.15360 0.15972
2 0.00010 0.00040 0.00090 0.00160 0.00250 0.00360 0.00391 0.00490 0.00640 0.00694
3 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.02970 0.05881 0.08733 0.11526 0.14263 0.16942 0.17603 0.19564 0.22131 0.22975
2 0.00030 0.00118 0.00265 0.00467 0.00725 0.01037 0.01123 0.01401 0.01818 0.01968
3 0.00000 0.00001 0.00003 0.00006 0.00013 0.00022 0.00024 0.00034 0.00051 0.00058
4 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.03940 0.07763 0.11471 0.15065 0.18549 0.21925 0.22752 0.25195 0.28361 0.29393
2 0.00059 0.00234 0.00519 0.00910 0.01402 0.01991 0.02153 0.02673 0.03443 0.03718
3 0.00000 0.00003 0.00011 0.00025 0.00048 0.00083 0.00093 0.00130 0.00193 0.00217
4 0.00000 0.00000 0.00000 0.00001 0.00001 0.00002 0.00002 0.00004 0.00005

5 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.04901 0.09608 0.14127 0.18463 0.22622 0.26610 0.27580 0.30431 0.34092 0.35277
2 0.00098 0.00384 0.00847 0.01476 0.02259 0.03187 0.03440 0.04249 0.05436 0.05858
3 0.00001 0.00008 0.00026 0.00060 0.00116 0.00197 0.00222 0.00308 0.00453 0.00509
4 0.00000 0.00000 0.00000 0.00001 0.00003 0.00006 0.00007 0.00011 0.00019 0.00023
5 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000

6 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.05852 0.11416 0.16703 0.21724 0.26491 0.31013 0.32107 0.35301 0.39364 0.40671
2 0.00146 0.00569 0.01246 0.02155 0.03277 0.04592 0.04949 0.06082 0.07729 0.08309
3 0.00002 0.00015 0.00050 0.00117 0.00223 0.00376 0.00423 0.00584 0.00851 0.00955
4 0.00000 0.00000 0.00001 0.00004 0.00009 0.00018 0.00021 0.00032 0.00054 0.00063
5 0.00000 0.00000 0.00000 0.00000 0.00001 0.00001 0.00002 0.00002
6 0.00000 0.00000 0.00000 0.00000

7 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.06793 0.13187 0.19202 0.24855 0.30166 0.35152 0.36350 0.39830 0.44215 0.45615
2 0.00203 0.00786 0.01709 0.02938 0.04438 0.06178 0.06647 0.08127 0.10259 0.11006
3 0.00003 0.00026 0.00086 0.00198 0.00376 0.00629 0.00706 0.00969 0.01401 0.01567
4 0.00000 0.00001 0.00003 0.00008 0.00019 0.00039 0.00046 0.00071 0.00118 0.00137
5 0.00000 0.00000 0.00000 0.00001 0.00001 0.00002 0.00003 0.00006 0.00007
6 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000

8 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.07726 0.14924 0.21626 0.27861 0.33658 0.39043 0.40328 0.44042 0.48678 0.50147
2 0.00269 0.01034 0.02234 0.03815 0.05724 0.07916 0.08503 0.10347 0.12976 0.13890
3 0.00005 0.00042 0.00135 0.00308 0.00579 0.00962 0.01077 0.01470 0.02110 0.02354
4 0.00000 0.00001 0.00005 0.00016 0.00037 0.00075 0.00087 0.00134 0.00220 0.00256
5 0.00000 0.00000 0.00001 0.00002 0.00004 0.00005 0.00008 0.00015 0.00018
(Continued)
630 Appendix

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

6 0.00000 0.00000 0.00000 0.00000 0.00000 0.00001 0.00001

7 0.00000 0.00000

9 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.08648 0.16625 0.23977 0.30747 0.36975 0.42701 0.44058 0.47959 0.52784 0.54301
2 0.00344 0.01311 0.02816 0.04777 0.07121 0.09784 0.10492 0.12705 0.15832 0.16912
3 0.00008 0.00061 0.00198 0.00448 0.00836 0.01380 0.01541 0.02091 0.02979 0.03315
4 0.00000 0.00002 0.00009 0.00027 0.00064 0.00128 0.00149 0.00227 0.00372 0.00431
5 0.00000 0.00000 0.00001 0.00003 0.00008 0.00010 0.00017 0.00031 0.00038
6 0.00000 0.00000 0.00000 0.00000 0.00001 0.00002 0.00002
7 0.00000 0.00000 0.00000

10 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.09562 0.18293 0.26258 0.33517 0.40126 0.46138 0.47554 0.51602 0.56561 0.58110
2 0.00427 0.01618 0.03451 0.05815 0.08614 0.11759 0.12590 0.15173 0.18788 0.20027
3 0.00011 0.00086 0.00276 0.00621 0.01150 0.01884 0.02101 0.02834 0.04008 0.04448
4 0.00000 0.00003 0.00015 0.00044 0.00103 0.00203 0.00236 0.00358 0.00580 0.00672
5 0.00000 0.00001 0.00002 0.00006 0.00015 0.00018 0.00031 0.00059 0.00071
6 0.00000 0.00000 0.00000 0.00001 0.00001 0.00002 0.00004 0.00005
7 0.00000 0.00000 0.00000 0.00000 0.00000

11 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.10466 0.19927 0.28470 0.36176 0.43120 0.49370 0.50832 0.54990 0.60036 0.61600
2 0.00518 0.01951 0.04135 0.06923 0.10189 0.13822 0.14775 0.17723 0.21810 0.23201
3 0.00016 0.00117 0.00372 0.00829 0.01524 0.02476 0.02756 0.03698 0.05190 0.05747
4 0.00000 0.00005 0.00023 0.00067 0.00155 0.00304 0.00353 0.00531 0.00854 0.00986
5 0.00000 0.00001 0.00004 0.00011 0.00026 0.00032 0.00054 0.00100 0.00121
6 0.00000 0.00000 0.00001 0.00002 0.00002 0.00004 0.00009 0.00011
7 0.00000 0.00000 0.00000 0.00000 0.00001 0.00001
8 0.00000 0.00000

12 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.11362 0.21528 0.30616 0.38729 0.45964 0.52408 0.53905 0.58140 0.63233 0.64800
2 0.00617 0.02311 0.04865 0.08094 0.11836 0.15954 0.17029 0.20332 0.24868 0.26401
3 0.00021 0.00154 0.00485 0.01073 0.01957 0.03157 0.03507 0.04680 0.06520 0.07201
4 0.00000 0.00007 0.00033 0.00098 0.00224 0.00434 0.00503 0.00753 0.01201 0.01383
5 0.00000 0.00002 0.00006 0.00018 0.00043 0.00052 0.00088 0.00161 0.00193
6 0.00000 0.00000 0.00001 0.00003 0.00004 0.00008 0.00016 0.00020
7 0.00000 0.00000 0.00000 0.00000 0.00001 0.00002
8 0.00000 0.00000

13 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.12248 0.23098 0.32697 0.41180 0.48666 0.55263 0.56786 0.61071 0.66175 0.67734
2 0.00725 0.02695 0.05637 0.09319 0.13542 0.18142 0.19333 0.22978 0.27937 0.29601
3 0.00027 0.00197 0.00616 0.01354 0.02451 0.03925 0.04353 0.05775 0.07987 0.08801
(Continued)
Appendix 631

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

4 0.00001 0.00010 0.00047 0.00137 0.00310 0.00598 0.00691 0.01028 0.01627 0.01868
5 0.00000 0.00000 0.00003 0.00010 0.00029 0.00067 0.00080 0.00134 0.00244 0.00292
6 0.00000 0.00001 0.00002 0.00006 0.00007 0.00013 0.00027 0.00034
7 0.00000 0.00000 0.00000 0.00000 0.00001 0.00002 0.00003
8 0.00000 0.00000 0.00000

14 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.13125 0.24636 0.34716 0.42533 0.51233 0.57948 0.59487 0.63796 0.68881 0.70423
2 0.00840 0.03103 0.06449 0.10593 0.15299 0.20369 0.21674 0.25645 0.30996 0.32779
3 0.00034 0.00247 0.00767 0.01672 0.03005 0.04778 0.05289 0.06980 0.09583 0.10534
4 0.00001 0.00014 0.00064 0.00185 0.00417 0.00797 0.00919 0.01360 0.02136 0.02446
5 0.00000 0.00001 0.00004 0.00015 0.00043 0.00098 0.00118 0.00197 0.00354 0.00423
6 0.00000 0.00000 0.00001 0.00003 0.00009 0.00012 0.00022 0.00045 0.00056
7 0.00000 0.00000 0.00001 0.00001 0.00002 0.00004 0.00006
8 0.00000 0.00000 0.00000 0.00000 0.00000

15 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.13994 0.26143 0.36675 0.45791 0.53671 0.60471 0.62019 0.66330 0.71370 0.72887
2 0.00963 0.03534 0.07297 0.11911 0.17095 0.22624 0.24038 0.28315 0.34027 0.35916
3 0.00042 0.00304 0.00937 0.02029 0.03620 0.05713 0.06313 0.08286 0.11297 0.12388
4 0.00001 0.00018 0.00085 0.00245 0.00547 0.01036 0.01193 0.1753 0.02731 0.03120
5 0.00000 0.00001 0.00006 0.00022 0.00061 0.00140 0.00168 0.00278 0.00497 0.00592
6 0.00000 0.00000 0.00001 0.00005 0.00015 0.00018 0.00034 0.00070 0.00086
7 0.00000 0.00000 0.00001 0.00002 0.00003 0.00008 0.00010
8 0.00000 0.00000 0.00000 0.00001 0.00001
9 0.00000 0.00000

16 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.14854 0.27620 0.38575 0.47960 0.55987 0.62843 0.64393 0.68687 0.73661 0.75147
2 0.01093 0.03986 0.08179 0.13266 0.18924 0.24895 0.26411 0.30976 0.37015 0.38997
3 0.00051 0.00369 0.01128 0.02424 0.04294 0.06728 0.07421 0.09688 0.13115 0.14349
4 0.00002 0.00024 0.00110 0.00316 0.00700 0.01317 0.01513 0.02211 0.03417 0.03892
5 0.00000 0.00001 0.00008 0.00031 0.00086 0.00194 0.00232 0.00381 0.00676 0.00803
6 0.00000 0.00000 0.00002 0.00008 0.00022 0.00028 0.00051 0.00104 0.00129
7 0.00000 0.00001 0.00002 0.00003 0.00005 0.00013 0.00016
8 0.00000 0.00000 0.00000 0.00000 0.00001 0.00002
9 0.00000 0.00000

17 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.15706 0.29068 0.40417 0.50041 0.58188 0.65072 0.66618 0.70879 0.75768 0.77218
2 0.01231 0.04459 0.09090 0.14654 0.20777 0.27171 0.28785 0.33616 0.39946 0.42009
3 0.00061 0.00441 0.01339 0.02858 0.05025 0.07818 0.08608 0.11178 0.15027 0.16403
4 0.00002 0.00031 0.00141 0.00401 0.00880 0.01641 0.01882 0.02734 0.04192 0.04763
5 0.00000 0.00002 0.00011 0.00042 0.00116 0.00261 0.00312 0.00509 0.00895 0.01060
(Continued)
632 Appendix

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

6 0.00000 0.00001 0.00003 0.00012 0.00032 0.00040 0.00074 0.00149 0.00185

7 0.00000 0.00000 0.00001 0.00003 0.00004 0.0009 0.00020 0.00026
8 0.00000 0.00000 0.00000 0.00001 0.00002 0.00003
9 0.00000 0.00000 0.00000

18 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.16549 0.30486 0.42205 0.52040 0.60279 0.67168 0.68704 0.72917 0.77706 0.79116
2 0.01376 0.04951 0.10030 0.16069 0.22648 0.29445 0.31150 0.36224 0.42812 0.44943
3 0.00073 0.00521 0.01572 0.03330 0.05813 0.08979 0.09869 0.12749 0.17020 0.18537
4 0.00003 0.00039 0.00177 0.00499 0.01087 0.02012 0.02302 0.03325 0.05059 0.05733
5 0.00000 0.00002 0.000150 0.00057 0.00155 0.00344 0.00411 0.00665 0.01159 0.01369
6 0.00000 0.00001 0.00005 0.00017 0.00046 0.00057 0.00105 0.00209 0.00258
7 0.00000 0.00000 0.00002 0.00005 0.00006 0.00013 0.00030 0.00039
8 0.00000 0.00000 0.00001 0.00001 0.000046 0.00005
9 0.00000 0.00000 0.00000 0.00000

19 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.17383 0.31877 0.43939 0.53958 0.62265 0.69138 0.70660 0.74813 0.79490 0.80857
2 0.01527 0.05462 0.10996 0.17508 0.24529 0.31709 0.33497 0.38793 0.45604 0.47791
3 0.00086 0.00610 0.01826 0.03840 0.06655 0.10207 0.11199 0.14392 0.19084 0.20737
4 0.00003 0.00049 0.00219 0.00612 0.01324 0.02430 0.02775 0.03985 0.06016 0.06800
5 0.00000 0.00003 0.00020 0.00074 0.00201 0.00444 0.00529 0.00851 0.01471 0.01732
6 0.00000 0.00001 0.00007 0.00024 0.00064 0.00079 0.00144 0.00285 0.00350
7 0.00000 0.00001 0.00002 0.00007 0.00010 0.00020 0.00045 0.00057
8 0.00000 0.00000 0.00001 0.00001 0.00002 0.00006 0.00008
9 0.00000 0.00000 0.00000 0.00001 0.00001
10 0.00000 0.00000

20 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.18209 0.33239 0.45621 0.55800 0.64151 0.70989 0.72494 0.76576 0.81131 0.82452
2 0.01686 0.05990 0.11984 0.18966 0.26416 0.33955 0.35820 0.41314 0.48314 0.50546
3 0.00100 0.00707 0.02101 0.04386 0.07548 0.11497 0.12592 0.16100 0.21205 0.22992
4 0.00004 0.00060 0.00267 0.00741 0.01590 0.02897 0.03302 0.04713 0.07062 0.07962
5 0.00000 0.00004 0.00026 0.00096 0.00257 0.00563 0.00669 0.01071 0.01834 0.02155
6 0.00000 0.00002 0.00010 0.00033 0.00087 0.00108 0.00193 0.00380 0.00465
7 0.00000 0.00001 0.00003 0.00011 0.00014 0.00028 0.00064 0.00082
8 0.00000 0.00000 0.00001 0.00001 0.00003 0.00009 0.00012
9 0.00000 0.00000 0.00000 0.00001 0.00001
10 0.00000 0.00000

21 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.19027 0.34574 0.47252 0.57568 0.65944 0.72730 0.74213 0.78216 0.82640 0.83914
2 0.01851 0.06535 0.12993 0.20440 0.28303 0.36177 0.38112 0.43783 0.50940 0.53205
3 0.00116 0.00813 0.02397 0.04969 0.08492 0.12845 0.14044 0.17865 0.23374 0.25288
(Continued)
Appendix 633

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

4 0.00005 0.00073 0.00322 0.00887 0.01888 0.03413 0.03882 0.05510 0.08193 0.09214
5 0.00000 0.00005 0.00033 0.00122 0.00324 0.00703 0.00834 0.01326 0.02253 0.02639
6 0.00000 0.00003 0.00013 0.00044 0.00115 0.00143 0.00255 0.00496 0.00606
7 0.00000 0.00001 0.00005 0.00015 0.00020 0.00040 0.00089 0.00113
8 0.00000 0.00000 0.00002 0.00002 0.00005 0.00013 0.00018
9 0.00000 0.00000 0.00001 0.00002 0.00002
10 0.00000 0.00000 0.00000

22 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.19837 0.35883 0.48834 0.59265 0.67647 0.74366 0.75825 0.79741 0.84029 0.85255
2 0.02023 0.07096 0.14021 0.21925 0.30185 0.38370 0.40368 0.46193 0.53476 0.55764
3 0.00134 0.00927 0.02715 0.05588 0.09482 0.14245 0.15548 0.19679 0.25579 0.27614
4 0.00006 0.00088 0.00384 0.01050 0.02218 0.03979 0.04517 0.06375 0.09408 0.10554
5 0.00000 0.00006 0.00042 0.00152 0.00402 0.00866 0.01024 0.01619 0.02728 0.03187
6 0.00000 0.00004 0.00018 0.00058 0.00151 0.00186 0.00330 0.00637 0.00776
7 0.00000 0.00002 0.00007 0.00021 0.00027 0.00055 0.00122 0.00155
8 0.00000 0.00001 0.00003 0.00003 0.00008 0.00019 0.00026
9 0.00000 0.00000 0.00000 0.00001 0.00003 0.00004
10 0.00000 0.00000 0.00000

23 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.20639 0.37165 0.50369 0.60894 0.69264 0.75904 0.77336 0.81159 0.85307 0.86484
2 0.02201 0.07671 0.15065 0.23418 0.32058 0.40530 0.42584 0.48541 0.55920 0.58222
3 0.00152 0.01050 0.03054 0.06242 0.10517 0.15692 0.17100 0.21535 0.27811 0.29960
4 0.00008 0.00104 0.00454 0.01232 0.02581 0.04595 0.05207 0.07307 0.10701 0.11975
5 0.00000 0.00008 0.00052 0.00188 0.00493 0.01053 0.01243 0.01952 0.03262 0.03801
6 0.00000 0.00005 0.00023 0.00075 0.00194 0.00238 0.00420 0.00804 0.00977
7 0.00000 0.00002 0.00009 0.00029 0.00037 0.00074 0.00163 0.00206
8 0.00000 0.00001 0.00004 0.00005 0.00011 0.00027 0.00036
9 0.00000 0.00000 0.00001 0.00001 0.00004 0.00005
10 0.00000 0.00000 0.00000 0.00001
11 0.00000

24 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.21432 0.38422 0.51858 0.62459 0.70801 0.77350 0.78752 0.82478 0.86482 0.87610
2 0.02385 0.08261 0.16124 0.24917 0.33918 0.42652 0.44756 0.50825 0.58271 0.60577
3 0.00173 0.01183 0.03415 0.06929 0.11594 0.17182 0.18692 0.23426 0.30060 0.32315
4 0.00009 0.00123 0.00532 0.01432 0.02978 0.05260 0.05950 0.08303 0.12070 0.13474
5 0.00000 0.00010 0.00064 0.00230 0.00597 0.01265 0.01490 0.02326 0.03857 0.04482
6 0.00001 0.00006 0.00030 0.00096 0.00245 0.00301 0.00527 0.01001 0.01212
7 0.00000 0.00000 0.00003 0.00013 0.00039 0.00050 0.00098 0.00214 0.00270
8 0.00000 0.00001 0.00005 0.00007 0.00015 0.00038 0.00050
9 0.00000 0.00001 0.00001 0.00002 0.00006 0.00008
(Continued)
634 Appendix

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

10 0.00000 0.00000 0.00000 0.00001 0.00001

11 0.00000 0.00000

25 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 022218 039654 0.53303 0.63960 0.72261 0.78709 0.80080 0.83704 0.87564 0.88642
2 0.02576 0.08865 0.17196 0.26419 0.35762 0.44734 0.46881 0.53040 0.60528 0.62830
3 0.00195 0.01324 0.03796 0.07648 0.12711 0.18711 0.20321 0.25344 0.32317 0.34670
4 0.00011 0.00145 0.00619 0.01652 0.03409 0.05976 0.06746 0.09361 0.13509 0.15044
5 0.00000 0.00012 0.00078 0.00278 0.00716 0.01505 0.01769 0.02745 0.04514 0.05231
6 0.00001 0.00008 0.00038 0.00121 0.00306 0.00375 0.00653 0.01229 0.01484
7 0.00000 0.00001 0.00004 0.00017 0.00051 0.00066 0.00128 0.00277 0.00349
8 0.00000 0.00000 0.00002 0.00007 0.00010 0.00021 0.00052 0.00069
9 0.00000 0.00001 0.00001 0.00003 0.00008 0.00011
10 0.00000 0.00000 0.00000 0.00001 0.00002
11 0.00000 0.00000

26 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.22996 0.40860 0.54703 0.65402 0.73648 0.79986 0.81325 0.84845 0.88558 0.89589
2 0.02772 0.09480 0.18279 0.27921 0.37587 0.46772 0.48956 0.55187 0.62691 0.64981
3 0.00219 0.01475 0.04198 0.08399 0.13863 0.20272 0.21981 0.27283 0.34574 0.37017
4 0.00013 0.00168 0.00714 0.01892 0.03874 0.06740 0.07595 0.10480 0.15014 0.16680
5 0.00001 0.00015 0.00094 0.00333 0.00851 0.01773 0.02080 0.03208 0.05234 0.06049
6 0.00000 0.00001 0.00010 0.00047 0.00151 0.00378 0.00462 0.00800 0.01492 0.01797
7 0.00000 0.00001 0.00005 0.00022 0.00067 0.00085 0.00165 0.00353 0.00444
8 0.00000 0.00001 0.00003 0.00010 0.00013 0.00029 0.00070 0.00092
9 0.00000 0.00000 0.00001 0.00002 0.00004 0.00012 0.00016
10 0.00000 0.00000 0.00001 0.00002 0.00002
11 0.00000 0.00000 0.00000

27 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.23766 0.42043 0.56062 0.66786 0.74966 0.81187 0.82492 0.85906 0.89474 0.90456
2 0.02975 0.10108 0.19372 0.29420 0.39390 0.48765 0.50979 0.57263 0.64760 0.67031
3 0.00244 0.01635 0.04620 0.09180 0.15049 0.21862 0.23667 0.29236 0.36823 0.39347
4 0.00015 0.00194 0.00818 0.02152 0.04374 0.07552 0.08494 0.11656 0.16579 0.18375
5 0.00001 0.00018 0.00113 0.00395 0.01002 0.02071 0.02425 0.03717 0.06016 0.06935
6 0.0000 0.00001 0.00013 0.00059 0.00186 0.00462 0.00564 0.00968 0.01791 0.02151
7 0.00000 0.00001 0.00007 0.00029 0.00085 0.00109 0.00209 0.00444 0.00556
8 0.00000 0.00001 0.00004 0.00013 0.00018 0.00038 0.00093 0.00121
9 0.00000 0.00000 0.00002 0.00002 0.00006 0.00017 0.00023
10 0.00000 0.00000 0.00001 0.00003 0.00004
11 0.00000 0.00000 0.00000

28 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.24528 0.43202 0.57380 0.68114 0.76217 0.82316 0.83587 0.86892 0.90316 0.91252
(Continued)
Appendix 635

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

2 0.03182 0.10747 0.20473 0.30915 0.41169 0.50711 0.52949 0.59268 0.66737 0.68984
3 0.00272 0.01805 0.05063 0.09990 0.16266 0.23476 0.25374 0.31198 0.39058 0.41654
4 0.00017 0.00223 0.00932 0.02433 0.04907 0.08410 0.09442 0.12887 0.18198 0.20122
5 0.00001 0.00021 0.00134 0.00466 0.01171 0.02400 0.02804 0.04273 0.06861 0.07888
6 0.00000 0.00002 0.00016 0.00072 0.00227 0.00559 0.00680 0.01161 0.2129 0.02550
7 0.00000 0.00001 0.00009 0.00036 0.00108 0.00137 0.00263 0.00552 0.00689
8 0.00000 0.00001 0.00005 0.00018 0.00023 0.00050 0.00121 0.00158
9 0.00000 0.00001 0.00002 0.00003 0.00008 0.00023 0.00031
10 0.00000 0.00000 0.00000 0.00001 0.00004 0.00005
11 0.00000 0.00001 0.00001
12 0.00000 0.00000

29 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.25283 0.44338 0.58659 0.69390 0.77406 0.83377 0.84613 0.87810 0.91091 0.91981
2 0.03396 0.11396 0.21580 0.32403 0.42922 0.52607 0.54863 0.61202 0.68624 0.70839
3 0.00301 0.01984 0.05525 0.10827 0.17512 0.25110 0.27098 0.33163 0.41272 0.43932
4 0.00019 0.00255 0.01056 0.02736 0.05475 0.09314 0.10438 0.14168 0.19867 0.21917
5 0.00001 0.00025 0.00158 0.00544 0.01358 0.02761 0.03219 0.04876 0.07768 0.08908
6 0.00000 0.00002 0.00019 0.00088 0.00274 0.00669 0.00813 0.01378 0.02508 0.02994
7 0.00000 0.00002 0.00012 0.00046 0.00135 0.00171 0.00325 0.00678 0.00844
8 0.00000 0.00001 0.00006 0.00023 0.00030 0.00065 0.00156 0.00202
9 0.00000 0.00001 0.00003 0.00005 0.00011 0.00031 0.00041
10 0.00000 0.00000 0.00001 0.00002 0.00005 0.00007
11 0.00000 0.00000 0.00001 0.00001
12 0.00000 0.00000

30 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.26030 0.45452 0.59899 0.70614 0.78536 0.84374 0.85574 0.88663 0.91803 0.92649
2 0.03615 0.12055 0.22692 0.33882 0.44646 0.54453 0.56723 0.63064 0.70421 0.72601
3 0.00332 0.02172 0.06007 0.11690 0.18782 0.26760 0.28833 0.35125 0.43460 0.46174
4 0.00022 0.00289 0.01190 0.03059 0.06077 0.10262 0.11479 0.15498 0.21579 0.23751
5 0.00001 0.00030 0.00185 0.00632 0.01564 0.03154 0.03670 0.05526 0.08736 0.09992
6 0.00000 0.00003 0.00023 0.00106 0.00328 0.00795 0.00963 0.01623 0.02929 0.03487
7 0.00000 0.00002 0.00015 0.00057 0.00167 0.00211 0.00399 0.00825 0.01023
8 0.00000 0.00002 0.00008 0.00030 0.00039 0.00083 0.00197 0.00255
9 0.00000 0.00001 0.00005 0.00006 0.00015 0.00041 0.00055
10 0.00000 0.00001 0.00001 0.00002 0.00007 0.00010
11 0.00000 0.00000 0.00000 0.00001 0.00002
12 0.00000 0.00000

31 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.26770 0.46543 0.61102 0.71790 0.79609 0.85312 0.86476 0.89457 0.92459 0.93262
2 0.03839 0.12723 0.23809 0.35351 0.46340 0.56248 0.58526 0.64856 0.72131 0.74272
(Continued)
636 Appendix

TABLE T3.4 (Continued)

Harvard Table of the Binomial Distribution, 1 − F(r − 1), Pr(x ≥ r|n, p)
n r p = 0.01 p = 0.02 p = 0.03 p = 0.04 p = 0.05 p =0.06 p = 1/16 p = 0.07 p = 0.08 p = 1/12

3 0.00365 0.02369 0.06507 0.12577 0.20075 0.28422 0.30576 0.37081 0.45617 0.48376
4 0.00025 0.00327 0.01335 0.03405 0.06712 0.11252 0.12564 0.16872 0.23330 0.25620
5 0.00001 0.00035 0.00215 0.00729 0.01789 0.03580 0.04158 0.06224 0.09764 0.11138
6 0.00000 0.00003 0.00028 0.00127 0.00390 0.00936 0.01132 0.01896 0.03393 0.04029
7 0.00000 0.00003 0.00018 0.00071 0.00205 0.00258 0.00485 0.00993 0.01229
8 0.00000 0.00002 0.00011 0.00038 0.00050 0.00105 0.00248 0.00319
9 0.00000 0.00001 0.00006 0.00008 0.00020 0.00053 0.00071
10 0.00000 0.00001 0.00001 0.00003 0.00010 0.00014
11 0.00000 0.00000 0.00000 0.00002 0.00002
12 0.00000 0.00000

32 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.27502 0.47612 0.62269 0.72918 0.80629 0.86193 0.87321 0.90195 0.93062 0.93823
2 0.04068 0.13399 0.24927 0.36809 0.48004 0.57992 0.60273 0.66578 0.73758 0.75854
3 0.00399 0.02577 0.07027 0.13488 0.21389 0.30091 0.32323 0.39025 0.47738 0.50534
4 0.00029 0.00368 0.01490 0.03771 0.07381 0.12282 0.13690 0.18287 0.25113 0.27516
5 0.00002 0.00041 0.00249 0.00836 0.02035 0.04041 0.04684 0.06970 0.10849 0.12345
6 0.00000 0.00004 0.00034 0.00151 0.00460 0.01095 0.01321 0.02199 0.03903 0.04622
7 0.00000 0.00004 0.00023 0.00087 0.00249 0.00313 0.00584 0.01185 0.01462
8 0.00000 0.00003 0.00014 0.00048 0.00063 0.00132 0.00307 0.00395
9 0.00000 0.00002 0.00008 0.00011 0.00026 0.00069 0.00092
10 0.00000 0.00001 0.00002 0.00004 0.00013 0.00019
11 0.00000 0.00000 0.00001 0.00002 0.00003
12 0.00000 0.00000 0.00001
13 0.00000

33 0 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000 1.00000
1 0.28227 0.48659 0.63401 0.74001 0.81597 0.87022 0.88114 0.90881 0.93617 0.94338
2 0.04303 0.14083 0.26048 0.38253 0.49635 0.59684 0.61963 0.68231 0.75302 0.77352
3 0.00436 0.02793 0.07564 0.14421 0.22719 0.31765 0.34070 0.40954 0.49820 0.52644
4 0.00032 0.00412 0.01656 0.04160 0.08081 0.13351 0.14854 0.19738 0.26923 0.29434
5 0.00002 0.00048 0.00286 0.00954 0.02303 0.04535 0.05247 0.07762 0.11990 0.13609
6 0.00000 0.00004 0.00040 0.00179 0.00539 0.01271 0.01532 0.02533 0.04459 0.05265
7 0.00000 0.00005 0.00028 0.00106 0.00299 0.00376 0.00697 0.01402 0.01725
8 0.00000 0.00004 0.00018 0.00060 0.00079 0.00164 0.00377 0.00484
9 0.00000 0.00003 0.00010 0.00014 0.00033 0.00088 0.00117
10 0.00000 0.00002 0.00002 0.00006 0.00018 0.00025
11 0.00000 0.00000 0.00001 0.00003 0.00005
12 0.00000 0.00000 0.00001
13 0.00000

Source: Reprinted from Harvard University Computing Laboratory, Tables of Cumulative Binomial Probability
Distribution, Harvard University Press, Cambridge, MA. With permission.
Appendix 637

TABLE T3.5
ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c – 1), Pr ( x ³ c|a ) =
x =c x !

c a = .001 a = .002 a = .003 a = .004

0 1.0000000 1.0000000 1.0000000 1.0000000
1 .0009995 .0019980 .0029955 .0039920
2 .0000005 .0000020 .0000045 .0000080
c a = .005 a = .006 a = .007 a = .008
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .0049875 .0059820 .0069756 .0079681
2 .0000125 .0000179 .0000244 .0000318
3 .0000001 .0000001

c a = .009 a = .010. a = .02 a = .03

0 1.0000000 1.0000000 1.0000000 1.0000000
1 .0089596 .0099502 .0198013 .0295545
2 .0000403 .0000497 .0001973 .0004411
3 .0000001 .0000002 .0000013 .0000044
c a = .04 a = .05 a = .06 a =.07
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .0392106 .0487706 .0582355 .0676062
2 .0007790 .0012091 .0017296 .0023386
3 .0000104 .0000201 .0000344 .0000542
4 .0000001 .0000003 .0000005 .0000009
c a = .08 a = .09 a = .10 a =.11
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .0768837 .0860688 .0951626 .1041659
2 .0030343 .0038150 .0046788 .0056241
3 .0000804 .0001136 .0001547 .0002043
4 .0000016 .0000025 .0000038 .0000056
5 .0000001
c a = .12 a = .13 a = .14 a = .15
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .1130796 .1219046 .1306418 .1392920
2 .0066491 .0077522 .0089316 .0101858
3 .0002633 .0003323 .0004119 .0005029
4 .0000079 .0000107 .0000143 .0000187
5 .0000002 .0000003 .0000004 .0000006
c a = .16 a = .17 a = .18 a = .19
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .1478562 .1563352 .1647298 .1730409
2 .0115132 .0129122 .0143812 .0159187
3 .0006058 .0007212 .0008498 .0009920
4 .0000240 .0000304 .0000379 .0000467
5 .0000008 .0000010 .0000014 .0000018
6 .0000001
c a = .20 a = .21 a = .22 a = .23
0 1.0000000 1.0000000 1.0000000 1.0000000
(Continued)
638 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

1 .1812692 .1894158 .1974812 .2054664

2 .0175231 .0191931 .0209271 .0227237
3 .0011485 .0013197 .0015060 .0017083
4 .0000568 .0000685 .0000819 .0000971
5 .0000023 .0000029 .0000036 .0000044
6 .0000001 .0000001 .0000001 .0000002
c a = .24 a = .25 a = .26 a = .27
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .2133721 .2211992 .2289484 .2366205
2 .0245815 .0264990 .0284750 .0305080
3 .0019266 .0021615 .0024135 .0026829
4 .0001142 .0001334 .0001548 .0001786
5 .0000054 .0000066 .0000080 .0000096
6 .0000002 .0000003 .0000003 .0000004
c a = .28 a = .29 a = .30 a = .4
0 1.0000000 1.0000000 1.0000000 1.0000000
1 .2442163 .2517364 .2591818 .3296800
2 .0325968 .0347400 .0369363 .0615519
3 .0029701 .0032755 .0035995 .0079263
4 .0002049 .0002339 .0002658 .0007763
5 .0000113 .0000134 .0000158 .0000612
6 .0000005 .0000006 .0000008 .0000040
7 .0000002
c a = .5 a = .6 a = .7 a = .8
0 1.000000 1.000000 1.000000 1.000000
1 .393469 .451188 .503415 .550671
2 .090204 .121901 .155805 .191208
3 .014388 .023115 .034142 .047423
4 .001752 .003358 .005753 .009080
5 .000172 .000394 .000786 .001411
6 .000014 .000039 .000090 .000184
7 .000001 .000003 .000009 .000021
8 .000001 .000002
c a = .9 a = 1.0 a = 1.1 a = 1.2
0 1.000000 1.000000 1.000000 1.000000
1 .593430 .632121 .667129 .698806
2 .227518 .264241 .300971 .337373
3 .062857 .080301 .099584 .120513
4 .013459 .018988 .025742 .033769
5 .002344 .003660 .005435 .007746
6 .000343 .000594 .000968 .001500
7 .000043 .000083 .000149 .000251
8 .000005 .000010 .000020 .000037
9 .000001 .000002 .000005
10 .000001
(Continued)
Appendix 639

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

c a = 1.3 a = 1.4 a = 1.5 a = 1.6

0 1.000000 1.000000 1.000000 1.000000
1 .727468 .753403 .776870 .798103
2 .373177 .408167 .442175 .475069
3 .142888 .166502 .191153 .216642
4 .043095 .053725 .065642 .078813
5 .010663 .014253 .018576 .023682
6 .002231 .003201 .004456 .006040
7 .000404 .000622 .000926 .001336
8 .000064 .000107 .000170 .000260
9 .000009 .000016 .000028 .000045
10 .000001 .000002 .000004 .000007
11 .000001 .000001
c a = 1.7 a = 1.8 a = 1.9 a = 2.0
0 1.000000 1.000000 1.000000 1.000000
1 .817316 .834701 .850431 .864665
2 .506754 .537163 .566251 .593994
3 .242777 .269379 .296280 .323324
4 .093189 .108708 .125298 .142877
5 .029615 .036407 .044081 .052653
6 .007999 .010378 .013219 .016564
7 .001875 .002569 .003446 .004534
8 .000388 .000562 .000793 .001097
9 .000072 .000110 .000163 .000237
10 .000012 .000019 .000030 .000046
11 .000002 .000003 .000005 .000008
12 .000001 .000001
c a = 2.1 a = 2.2 a = 2.3 a = 2.4 a = 2.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .877544 .889197 .899741 .909282 .917915
2 .620385 .645430 .669146 .691559 .712703
3 .350369 .377286 .403961 .430291 .456187
4 .161357 .180648 .200653 .221277 .242424
5 .062126 .072496 .083751 .095869 .108822
6 .020449 .024910 .029976 .035673 .042021
7 .005862 .007461 .009362 .011594 .014187
8 .001486 .001978 .002589 .003339 .004247
9 .000337 .000470 .000642 .000862 .001140
10 .000069 .000101 .000144 .000202 .000277
11 .000013 .000020 .000029 .000043 .000062
12 .000002 .000004 .000006 .000008 .000013
13 .000001 .000001 .000002 .000002
(Continued)
640 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

c a = 2.6 a = 2.7 a = 2.8 a = 2.9 a = 3.0

0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .925726 .932794 .939190 .944977 .950213
2 .732615 .751340 .768922 .785409 .800852
3 .481570 .506375 .530546 .554037 .576810
4 .263998 .285908 .308063 .330377 .352768
5 .122577 .137092 .152324 .168223 .184737
6 .049037 .056732 .065110 .074174 .083918
7 .017170 .020569 .024411 .028717 .033509
8 .005334 .006621 .008131 .009885 .011905
9 .001487 .001914 .002433 .003058 .003803
10 .000376 .000501 .000660 .000858 .001102
11 .000087 .000120 .000164 .000220 .000292
12 .000018 .000026 .000037 .000052 .000071
13 .000004 .000005 .000008 .000011 .000016
14 .000001 .000001 .000002 .000002 .000003
15 .000001
c a = 3.1 a = 3.2 a = 3.3 a = 3.4 a = 3.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .954951 .959238 .963117 .966627 .969803
2 .815298 .828799 .841402 .853158 .864112
3 .598837 .620096 .640574 .660260 .679153
4 .375160 .397480 .419662 .441643 .463367
5 .201811 .219387 .237410 .255818 .274555
6 .094334 .105408 .117123 .129458 .142386
7 .038804 .044619 .050966 .057853 .065288
8 .014213 .016830 .019777 .023074 .026739
9 .004683 .005714 .006912 .008293 .009874
10 .001401 .001762 .002195 .002709 .003315
11 .000383 .000497 .000638 .000810 .001019
12 .000097 .000129 .000171 .000223 .000289
13 .000023 .000031 .000042 .000057 .000076
14 .000005 .000007 .000010 .000014 .000019
15 .000001 .000001 .000002 .000003 .000004
16 .000001 .000001
c a = 3.6 a = 3.7 a = 3.8 a = 3.9 a = 4.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .972676 .975276 .977629 .979758 .981684
2 .874311 .883799 .892620 .900815 .908422
3 .697253 .714567 .731103 .746875 .761897
4 .484784 .505847 .526515 .546753 .566530
5 .293562 .312781 .332156 .351635 .371163
6 .155881 .169912 .184444 .199442 .214870
(Continued)
Appendix 641

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

7 .073273 .081809 .090892 .100517 .110674

8 .030789 .035241 .040107 .045402 .051134
9 .011671 .013703 .015984 .018533 .021363
10 .004024 .004848 .005799 .006890 .008132
11 .001271 .001572 .001929 .002349 .002840
12 .000370 .000470 .000592 .000739 .000915
13 .000100 .000130 .000168 .000216 .000274
14 .000025 .000034 .000045 .000059 .000076
15 .000006 .000008 .000011 .000015 .000020
16 .000001 .000002 .000003 .000004 .000005
17 .000001 .000001 .000001
c a = 4.1 a = 4.2 a = 4.3 a = 4.4 a = 4.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .983427 .985004 .986431 .987723 .988891
2 .915479 .922023 .928087 .933702 .938901
3 .776186 .789762 .802645 .814858 .826422
4 .585818 .604597 .622846 .640552 .657704
5 .390692 .410173 .429562 .448816 .467896
6 .230688 .246857 .263338 .280088 .297070
7 .181352 .132536 .144210 .156355 .168949
8 .057312 .063943 .071032 .078579 .086586
9 .024492 .027932 .031698 .035803 .040257
10 .009540 .011127 .012906 .014890 .017093
11 .003410 .004069 .004825 .005688 .006669
12 .001125 .001374 .001666 .002008 .002404
13 .000345 .000431 .000534 .000658 .000805
14 .000098 .000126 .000160 .000201 .000252
15 .000026 .000034 .000045 .000058 .000074
16 .000007 .000009 .000012 .000016 .000020
17 .000002 .000002 .000003 .000004 .000005
18 .000001 .000001 .000001
c a = 4.6 a = 4.7 a = 4.8 a = 4.9 a = 5.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .989948 .990905 .991770 .992553 .993262
2 .943710 .948157 .952267 .956065 .959572
3 .837361 .847700 .857461 .866669 .875348
4 .674294 .690316 .705770 .720655 .734974
5 .486766 .505391 .523741 .541788 .559507
6 .314240 .331562 .348994 .366499 .384039
7 .181971 .195395 .209195 .223345 .237817
8 .095051 .103969 .113334 .123138 .133372
9 .045072 .050256 .055817 .061761 .068094
10 .019527 .022206 .025141 .028345 .031828
11 .007777 .009022 .010417 .011971 .013695
(Continued)
642 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

12 .002863 .003389 .003992 .004677 .005453

13 .000979 .001183 .001422 .001699 .002019
14 .000312 .000385 .000473 .000576 .000698
15 .000093 .000118 .000147 .000183 .000226
16 .000026 .000034 .000043 .000055 .000069
17 .000007 .000009 .000012 .000015 .000020
18 .000002 .000002 .000003 .000004 .000005
19 .000001 .000001 .000001 .000001
c a = 5.1 a = 5.2 a = 5.3 a = 5.4 a = 5.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .993903 .994483 .995008 .995483 .995913
2 .962810 .965797 .968553 .971094 .973436
3 .883522 .891213 .898446 .905242 .911624
4 .748732 .761935 .774590 .786709 .798301
5 .576875 .593872 .610482 .626689 .642482
6 .401580 .419087 .436527 .453868 .471081
7 .252580 .267607 .282866 .298329 .313964
8 .144023 .155078 .166523 .178341 .190515
9 .074818 .081935 .089446 .097350 .105643
10 .035601 .039674 .044056 .048755 .053777
11 .015601 .017699 .020000 .022514 .025251
12 .006328 .007310 .008409 .009632 .010988
13 .002387 .002809 .003289 .003835 .004451
14 .000841 .001008 .001202 .001427 .001685
15 .000278 .000339 .000412 .000498 .000599
16 .000086 .000108 .000133 .000164 .000200
17 .000025 .000032 .000041 .000051 .000063
18 .000007 .000009 .000012 .000015 .000019
19 .000002 .000002 .000003 .000004 .000005
20 .000001 .000001 .000001 .000001
c a = 5.6 a = 5.7 a = 5.8 a = 5.9 a = 6.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .996302 .996654 .996972 .997261 .997521
2 .975594 .977582 .979413 .981098 .982649
3 .917612 .923227 .928489 .933418 .938031
4 .809378 .819952 .830037 .839647 .848796
5 .657850 .672785 .687282 .701335 .714943
6 .488139 .505015 .521685 .538127 .554320
7 .329742 .345634 .361609 .377639 .393697
8 .203025 .215851 .228974 .242371 .256020
9 .114322 .123382 .132814 .142611 .152763
10 .059130 .064817 .070844 .077212 .083924
11 .028222 .031436 .034901 .038627 .042621
12 .012487 .014138 .015950 .017931 .020092
(Continued)
Appendix 643

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

13 .005144 .005922 .006790 .007756 .008827

14 .001981 .002319 .002703 .003138 .003628
15 .000716 .000852 .001010 .001192 .001400
16 .000244 .000295 .000356 .000426 .000509
17 .000078 .000096 .000118 .000144 .000175
18 .000024 .000030 .000037 .000046 .000057
19 .000007 .000009 .000011 .000014 .000018
20 .000002 .000002 .000003 .000004 .000005
21 .000001 .000001 .000001 .000001
c a = 6.1 a = 6.2 a = 6.3 a = 6.4 a = 6.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .997757 .997971 .998164 .998338 .998497
2 .984076 .985388 .986595 .987704 .988724
3 .942347 .946382 .950154 .953676 .956964
4 .857499 .865771 .873626 .881081 .888150
5 .728106 .740823 .753096 .764930 .776328
6 .570246 .585887 .601228 .616256 .630959
7 .409755 .425787 .441767 .457671 .473476
8 .269899 .283984 .298252 .312679 .327242
9 .163258 .174086 .185233 .196685 .208427
10 .090980 .098379 .106121 .114201 .122616
11 .046890 .051441 .056280 .061411 .066839
12 .022440 .024985 .027734 .030697 .033880
13 .010012 .011316 .012748 .014316 .016027
14 .004180 .004797 .005485 .006251 .007100
15 .001639 .001910 .002217 .002565 .002956
16 .000605 .000716 .000844 .000992 .001160
17 .000211 .000254 .000304 .000362 .000430
18 .000070 .000085 .000104 .000126 .000151
19 .000022 .000027 .000034 .000041 .000051
20 .000007 .000008 .000010 .000013 .000016
21 .000002 .000002 .000003 .000004 .000005
22 .000001 .000001 .000001 .000001 .000001
c a = 6.6 a = 6.7 a = 6.8 a = 6.9 a = 7.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .998640 .998769 .998886 .998992 .999088
2 .989661 .990522 .991313 .992038 .992705
3 .960032 .962894 .965562 .968048 .970364
4 .894849 .901192 .907194 .912870 .918235
5 .787296 .797841 .807969 .817689 .827008
6 .645327 .659351 .673023 .686338 .699292
7 .489161 .504703 .520084 .535285 .550289
8 .341918 .356683 .371514 .386389 .401286
9 .220443 .232716 .245230 .257967 .270909
(Continued)
644 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

10 .131361 .140430 .149816 .159510 .169504

11 .072567 .078598 .084934 .091575 .098521
12 .037291 .040937 .044825 .048961 .053350
13 .017889 .019910 .022097 .024458 .027000
14 .008038 .009072 .010208 .011452 .012811
15 .003395 .003886 .004434 .005042 .005717
16 .001352 .001569 .001816 .002094 .002407
17 .000509 .000599 .000703 .000822 .000958
18 .000182 .000217 .000258 .000306 .000362
19 .000062 .000075 .000090 .000108 .000130
20 .000020 .000024 .000030 .000037 .000044
21 .000006 .000008 .000010 .000012 .000014
22 .000002 .000002 .000003 .000004 .000005
23 .000001 .000001 .000001 .000001 .000001
c a = 7.1 a = 7.2 a = 7.3 a = 7.4 a = 7.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999175 .999253 .999324 .999389 .999447
2 .993317 .993878 .994393 .994865 .995299
3 .972520 .974526 .976393 .978129 .979743
4 .923301 .928083 .932594 .936847 .940855
5 .835937 .844484 .852660 .860475 .867938
6 .711881 .724103 .735957 .747443 .758564
7 .565080 .579644 .593968 .608038 .621845
8 .416183 .431059 .445893 .460667 .475361
9 .284036 .297332 .310776 .324349 .338033
10 .179788 .190350 .201180 .212265 .223592
11 .105771 .113323 .121175 .129323 .137762
12 .057997 .062906 .068081 .073526 .079241
13 .029730 .032655 .035782 .039117 .042666
14 .014292 .015901 .017645 .019531 .021565
15 .006463 .007285 .008188 .009178 .010260
16 .002757 .003149 .003586 .004071 .004608
17 .001113 .001288 .001486 .001709 .001959
18 .000426 .000500 .000584 .000680 .000790
19 .000155 .000184 .000218 .000258 .000303
20 .000054 .000065 .000078 .000093 .000111
21 .000018 .000022 .000026 .000032 .000039
22 .000006 .000007 .000009 .000011 .000013
23 .000002 .000002 .000003 .000003 .000004
24 .000001 .000001 .000001 .000001
c a = 7.6 a = 7.7 a = 7.8 a = 7.9 a = 8.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999500 .999547 .999590 .999629 .999665
2 .995696 .996060 .996394 .996700 .996981
(Continued)
Appendix 645

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

3 .981243 .982636 .983930 .985131 .986246

4 .944629 .948181 .951523 .954666 .957620
5 .875061 .881855 .888330 .894497 .900368
6 .769319 .779713 .789749 .799431 .808764
7 .635379 .648631 .661593 .674260 .686626
8 .489958 .504440 .518791 .532996 .547039
9 .351808 .365657 .379559 .393497 .407453
10 .235149 .246920 .258891 .271048 .283376
11 .146487 .155492 .164770 .174314 .184114
12 .085230 .091493 .098030 .104841 .111924
13 .046434 .050427 .054649 .059104 .063797
14 .023753 .026103 .028620 .031311 .034181
15 .011441 .012725 .014118 .015627 .017257
16 .005202 .005857 .006577 .007367 .008231
17 .002239 .002552 .002901 .003289 .003718
18 .000915 .001055 .001215 .001393 .001594
19 .000355 .000415 .000484 .000562 .000650
20 .000132 .000156 .000184 .000216 .000253
21 .000046 .000056 .000067 .000079 .000094
22 .000016 .000019 .000023 .000028 .000033
23 .000005 .000006 .000008 .000009 .000011
24 .000002 .000002 .000002 .000003 .000004
25 .000001 .000001 .000001 .000001
c a = 8.1 a = 8.2 a = 8.3 a = 8.4 a = 8.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999696 .999725 .999751 .999775 .999797
2 .997238 .997473 .997689 .997886 .998067
3 .987280 .988239 .989129 .989953 .990717
4 .960395 .963000 .965446 .967740 .969891
5 .905951 .911260 .916303 .921092 .925636
6 .817753 .826406 .834727 .842723 .850403
7 .698686 .710438 .721879 .733007 .743822
8 .560908 .574591 .588074 .601348 .614403
9 .421408 .435347 .449252 .463106 .476895
10 .295858 .308481 .321226 .334080 .347026
11 .194163 .204450 .214965 .225699 .236638
12 .119278 .126900 .134787 .142934 .151338
13 .068731 .073907 .079330 .084999 .090917
14 .037236 .040481 .043923 .047564 .051411
15 .019014 .020903 .022931 .025103 .027425
16 .009174 .010201 .011316 .012525 .013833
17 .004192 .004715 .005291 .005922 .006613
18 .001819 .002070 .002349 .002659 .003002
19 .000751 .000864 .000992 .001136 .001297
(Continued)
646 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

20 .000296 .000344 .000400 .000463 .000535

21 .000111 .000131 .000154 .000180 .000211
22 .000040 .000048 .000057 .000067 .000079
23 .000014 .000017 .000020 .000024 .000029
24 .000005 .000006 .000007 .000008 .000010
25 .000001 .000002 .000002 .000003 .000003
26 .000001 .000001 .000001 .000001
c a = 8.6 a = 8.7 a = 8.8 a = 8.9 a = 9.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999816 .999833 .999849 .999864 .999877
2 .998233 .998384 .998523 .998650 .998766
3 .991424 .992080 .992686 .993248 .993768
4 .971907 .973797 .975566 .977223 .978774
5 .929946 .934032 .937902 .941567 .945036
6 .857772 .864840 .871613 .878100 .884309
7 .754324 .764512 .774390 .783958 .793819
8 .627229 .639819 .652166 .664262 .676103
9 .490603 .504216 .517719 .531101 .544347
10 .360049 .373132 .386260 .399419 .412592
11 .247772 .259089 .270577 .282222 .294012
12 .159992 .168892 .178030 .187399 .196992
13 .097084 .103499 .110162 .117072 .124227
14 .055467 .059736 .064221 .068925 .073851
15 .029902 .032540 .035343 .038317 .041466
16 .015245 .016767 .018402 .020157 .022036
17 .007367 .008190 .009084 .010055 .011106
18 .003382 .003800 .004261 .004766 .005320
19 .001478 .001679 .001903 .002151 .002426
20 .000616 .000707 .000811 .000926 .001056
21 .000245 .000285 .000330 .000381 .000439
22 .000094 .000110 .000129 .000150 .000175
23 .000034 .000041 .000048 .000057 .000067
24 .000012 .000014 .000017 .000021 .000025
25 .000004 .000005 .000006 .000007 .000009
26 .000001 .000002 .000002 .000002 .000003
27 .000001 .000001 .000001 .000001
c a = 9.1 a = 9.2 a = 9.3 a = 9.4 a = 9.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999888 .999899 .999909 .999917 .999925
2 .998872 .998969 .999058 .999140 .999214
3 .994249 .994693 .995105 .995485 .995836
4 .980224 .981580 .982848 .984033 .985140
5 .948318 .951420 .954353 .957122 .959737
6 .890249 .895926 .901350 .906529 .911472
(Continued)
Appendix 647

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

7 .802177 .810835 .819197 .827267 .835051

8 .687684 .699000 .710050 .720829 .731337
9 .557448 .570391 .583166 .595765 .608177
10 .425765 .438924 .452054 .465142 .478174
11 .305933 .317974 .330119 .342356 .354672
12 .206800 .216815 .227029 .237430 .248010
13 .131624 .139261 .147133 .155238 .163570
14 .079001 .084376 .089978 .095807 .101864
15 .044795 .048309 .052010 .055903 .059992
16 .024044 .026188 .028470 .030897 .033473
17 .012242 .013468 .014788 .016206 .017727
18 .005924 .006584 .007302 .008083 .008928
19 .002731 .003066 .003435 .003840 .004284
20 .001201 .001362 .001542 .001742 .001962
21 .000505 .000579 .000662 .000755 .000859
22 .000203 .000235 .000272 .000314 .000361
23 .000078 .000092 .000107 .000125 .000145
24 .000029 .000034 .000041 .000048 .000056
25 .000010 .000012 .000015 .000018 .000021
26 .000004 .000004 .000005 .000006 .000007
27 .000001 .000001 .000002 .000002 .000003
28 .000001 .000001 .000001
c a = 9.6 a = 9.7 a = 9.8 a = 9.9 a = 10.0
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999932 .999939 .999945 .999950 .999955
2 .999282 .999344 .999401 .999453 .999501
3 .996161 .996461 .996738 .996994 .997231
4 .986174 .987139 .988040 .988880 .989664
5 .962205 .964533 .966729 .968798 .970747
6 .916185 .920678 .924959 .929035 .932914
7 .842553 .849779 .856735 .863426 .869859
8 .741572 .751533 .761221 .770636 .779779
9 .620394 .632410 .644217 .655809 .667180
10 .491138 .504021 .516812 .529498 .542070
11 .367052 .379484 .391955 .404451 .416960
12 .258759 .269665 .280719 .291909 .303224
13 .172124 .180895 .189876 .199062 .208444
14 .108148 .114659 .121395 .128355 .135536
15 .064279 .068767 .073458 .078355 .083458
16 .036202 .039090 .042139 .045355 .048740
17 .019357 .021098 .022956 .024936 .027042
18 .009844 .010832 .011898 .013045 .014278
19 .004770 .005300 .005877 .006505 .007187
20 .002207 .002476 .002772 .003098 .003454
(Continued)
648 Appendix

TABLE T3.5 (Continued)

ax e -a
å
¥
Molina Table of the Poisson Distribution, 1 – F(c– 1), Pr ( x ³ c|a ) =
x =c x !

21 .000976 .001106 .001250 .001411 .001588

22 .000414 .000473 .000540 .000616 .000700
23 .000168 .000194 .000224 .000258 .000296
24 .000066 .000077 .000089 .000104 .000120
25 .000025 .000029 .000034 .000040 .000047
26 .000009 .000011 .000013 .000015 .000018
27 .000003 .000004 .000004 .000005 .000006
28 .000001 .000001 .000002 .000002 .000002
29 .000001 .000001 .000001
c a = 10.1 a = 10.2 a = 10.3 a = 10.4 a = 10.5
0 1.000000 1.000000 1.000000 1.000000 1.000000
1 .999959 .999963 .999966 .999970 .999972
2 .999544 .999584 .999620 .999653 .999683
3 .997449 .997650 .997836 .998007 .998165
4 .990395 .991076 .991711 .992302 .992853

Source: Reprinted from Molina, E.C., Poisson’s Exponential Binomial Limit, Van Nostrand,
New York. Copyright 1942, Bell Telephone Laboratories. With permission.
Appendix 649

TABLE T5.1
Cameron Table of Unity Values for Constructing Single-Sampling Plans
Values of p2/p1 for Values of p2/p1 for
α = .05 α = .05 α = .05 α = .01 α = .01 α = .01
c β = .10 β = .05 β = .01 np1 c β = .10 β = .05 β = .01 np1
0 44.890 58.404 89.781 0.052 0 229.105 298.073 458.210 0.010
1 10.946 13.349 18.681 0.355 1 26.184 31.933 44.686 0.149
2 6.509 7.699 10.280 0.818 2 12.206 14.439 19.278 0.436
3 4.890 5.675 7.352 1.366 3 8.115 9.418 12.202 0.823
4 4.057 4.646 5.890 1.970 4 6.249 7.156 9.072 1.279
5 3.549 4.023 5.017 2.613 5 5.195 5.889 7.343 1.785
6 3.206 3.604 4.435 3.286 6 4.520 5.082 6.253 2.330
7 2.957 3.303 4.019 3.981 7 4.050 4.524 5.506 2.906
8 2.768 3.074 3.707 4.695 8 3.705 4.115 4.962 3.507
9 2.618 2.895 3.462 5.426 9 3.440 3.803 4.548 4.130
10 2.497 2.750 3.265 6.169 10 3.229 3.555 4.222 4.771
11 2.397 2.630 3.104 6.924 11 3.058 3.354 3.959 5.428
12 2.312 2.528 2.968 7.690 12 2.915 3.188 3.742 6.099
13 2.240 2.442 2.852 8.464 13 2.795 3.047 3.559 6.782
14 2.177 2.367 2.752 9.246 14 2.692 2.927 3.403 7.477
15 2.122 2.302 2.665 10.035 15 2.603 2.823 3.269 8.181
16 2.073 2.244 2.588 10.831 16 2.524 2.732 3.151 8.895
17 2.029 2.192 2.520 11.633 17 2.455 2.652 3.048 9.616
18 1.990 2.145 2.458 12.442 18 2.393 2.580 2.956 10.346
19 1.954 2.103 2.403 13.254 19 2.337 2.516 2.874 11.082
20 1.922 2.065 2.352 14.072 20 2.287 2.458 2.799 11.825
21 1.892 2.030 2.307 14.894 21 2.241 2.405 2.733 12.574
22 1.865 1.999 2.265 15.719 22 2.200 2.357 2.671 13.329
23 1.840 1.969 2.226 16.548 23 2.162 2.313 2.615 14.088
24 1.817 1.942 2.191 17.382 24 2.126 2.272 2.564 14.853
25 1.795 1.917 2.158 18.218 25 2.094 2.235 2.516 15.623
26 1.775 1.893 2.127 19.058 26 2.064 2.200 2.472 16.397
27 1.757 1.871 2.098 19.900 27 2.035 2.168 2.431 17.175
28 1.739 1.850 2.071 20.746 28 2.009 2.138 2.393 17.957
29 1.723 1.831 2.046 21.594 29 1.985 2.110 2.358 18.742
30 1.707 1.813 2.023 22.444 30 1.962 2.083 2.324 19.532
31 1.692 1.796 2.001 23.298 31 1.940 2.059 2.293 20.324
32 1.679 1.780 1.980 24.152 32 1.920 2.035 2.264 21.120
33 1.665 1.764 1.960 25.010 33 1.900 2.013 2.236 21.919
34 1.653 1.750 1.941 25.870 34 1.882 1.992 2.210 22.721
35 1.641 1.736 1.923 26.731 35 1.865 1.973 2.185 23.525
36 1.630 1.723 1.906 27.594 36 1.848 1.954 2.162 24.333
37 1.619 1.710 1.890 28.460 37 1.833 1.936 2.139 25.143
38 1.609 1.698 1.875 29.327 38 1.818 1.920 2.118 25.955
39 1.599 1.687 1.860 30.196 39 1.804 1.903 2.098 26.770
40 1.590 1.676 1.846 31.066 40 1.790 1.887 2.079 27.587
41 1.581 1.666 1.833 31.938 41 1.777 1.873 2.060 28.406
(Continued)
650 Appendix

TABLE T5.1 (Continued)

Cameron Table of Unity Values for Constructing Single-Sampling Plans
Values of p2/p1 for Values of p2/p1 for
α = .05 α = .05 α = .05 α = .01 α = .01 α = .01
c β = .10 β = .05 β = .01 np1 c β = .10 β = .05 β = .01 np1

42 1.572 1.656 1.820 32.812 42 1.765 1.859 2.043 29.228

43 1.564 1.646 1.807 33.686 43 1.753 1.845 2.026 30.051
44 1.556 1.637 1.796 34.563 44 1.742 1.832 2.010 30.877
45 1.548 1.628 1.784 35.441 45 1.731 1.820 1.994 31.704
46 1.541 1.619 1.773 36.320 46 1.720 1.808 1.980 32.534
47 1.534 1.611 1.763 37.200 47 1.710 1.796 1.965 33.365
48 1.527 1.603 1.752 38.082 48 1.701 1.785 1.952 34.198
49 1.521 1.596 1.743 38.965 49 1.691 1.775 1.938 35.032
Source: Reprinted from Cameron, J.M., Ind. Qual. Control, 9(1), 38, 1952. With permission.
TABLE T5.2
Cameron Table of Unity Values to Determine Probability of Acceptance
P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) =
Appendix

c .995 .990 .975 .950 .900 .750 .500 .250 .100 .050 .025 .010 .005

0 .00501 .0101 .0253 .0513 .105 .288 .693 1.386 2.303 2.996 3.689 4.605 5.298
1 .103 .149 .242 .355 .532 .961 1.678 2.693 3.890 4.744 5.572 6.638 7.430
2 .338 .436 .619 .818 1.102 1.727 2.674 3.920 5.322 6.296 7.224 8.406 9.274
3 .672 .823 1.090 1.366 1.745 2.535 3.672 5.109 6.681 7.754 8.768 10.045 10.978
4 1.078 1.279 1.623 1.970 2.433 3.369 4.671 6.274 7.994 9.154 10.242 11.605 12.594
5 1.537 1.785 2.202 2.613 3.152 4.219 5.670 7.423 9.275 10.513 11.668 13.108 14.150
6 2.037 2.330 2.814 3.286 3.895 5.083 6.670 8.558 10.532 11.842 13.060 14.571 15.660
7 2.571 2.906 3.454 3.981 4.656 5.956 7.669 9.684 11.771 13.148 14.422 16.000 17.134
8 3.132 3.507 4.115 4.695 5.432 6.838 8.669 10.802 12.995 14.434 15.763 17.403 18.578
9 3.717 4.130 4.795 5.426 6.221 7.726 9.669 11.914 14.206 15.705 17.085 18.783 19.998

10 4.321 4.771 5.491 6.169 7.021 8.620 10.668 13.020 15.407 16.962 18.390 20.145 21.398
11 4.943 5.428 6.201 6.924 7.829 9.519 11.668 14.121 16.598 18.208 19.682 21.490 22.779
12 5.580 6.099 6.922 7.690 8.646 10.422 12.668 15.217 17.782 19.442 20.962 22.821 24.145
13 6.231 6.782 7.654 8.464 9.470 11.329 13.668 16.310 18.958 20.668 22.230 24.139 25.496
14 6.893 7.477 8.396 9.246 10.300 12.239 14.668 17.400 20.128 21.886 23.490 25.446 26.836
15 7.566 8.181 9.144 10.035 11.135 13.152 15.668 18.486 21.292 23.098 24.741 26.743 28.166
16 8.249 8.895 9.902 10.831 11.976 14.068 16.668 19.570 22.452 24.302 25.984 28.031 29.484
17 8.942 9.616 10.666 11.633 12.822 14.986 17.668 20.652 23.606 25.500 27.220 29.310 30.792
18 9.644 10.346 11.438 12.442 13.672 15.907 18.668 21.731 24.756 26.692 28.448 30.581 32.092
19 10.353 11.082 12.216 13.254 14.525 16.830 19.668 22.808 25.902 27.879 29.671 31.845 33.383

20 11.069 11.825 12.999 14.072 15.383 17.755 20.668 23.883 27.045 29.062 30.888 33.103 34.668
21 11.791 12.574 13.787 14.894 16.244 18.682 21.668 24.956 28.184 30.241 32.102 34.355 35.947
22 12.520 13.329 14.580 15.719 17.108 19.610 22.668 26.028 29.320 31.416 33.309 35.601 37.219
23 13.255 14.088 15.377 16.548 17.975 20.540 23.668 27.098 30.453 32.586 34.512 36.841 38.485
24 13.995 14.853 16.178 17.382 18.844 21.471 24.668 28.167 31.584 33.752 35.710 38.077 39.745
25 14.740 15.623 16.984 18.218 19.717 22.404 25.667 29.234 32.711 34.916 36.905 39.308 41.000
26 15.490 16.397 17.793 19.058 20.592 23.338 26.667 30.300 33.836 36.077 38.096 40.535 42.252
(Continued)
651
 652

TABLE T5.2 (Continued)

Cameron Table of Unity Values to Determine Probability of Acceptance
P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) = P(A) =
c .995 .990 .975 .950 .900 .750 .500 .250 .100 .050 .025 .010 .005

27 16.245 17.175 18.606 19.900 21.469 24.273 27.667 31.365 34.959 37.234 39.284 41.757 43.497
28 17.004 17.957 19.422 20.746 22.348 25.209 28.667 32.428 36.080 38.389 40.468 42.975 44.738
29 17.767 18.742 20.241 21.594 23.229 26.147 29.667 33.491 37.198 39.541 41.649 44.190 45.976

30 18.534 19.532 21.063 22.444 24.113 27.086 30.667 34.552 38.315 40.690 42.827 45.401 47.210
31 19.305 20.324 21.888 23.298 24.998 28.025 31.667 35.613 39.430 41.838 44.002 46.609 48.440
32 20.079 21.120 22.716 24.152 25.885 28.966 32.667 36.672 40.543 42.982 45.174 47.813 49.666
33 20.856 21.919 23.546 25.010 26.774 29.907 33.667 37.731 41.654 44.125 46.344 49.015 50.888
34 21.638 22.721 24.379 25.870 27.664 30.849 34.667 38.788 42.764 45.266 47.512 50.213 52.108
35 22.422 23.525 25.214 26.731 28.556 31.792 35.667 39.845 43.872 46.404 48.676 51.409 53.324
36 23.208 24.333 26.052 27.594 29.450 32.736 36.667 40.901 44.978 47.540 49.840 52.601 54.538
37 23.998 25.143 26.891 28.460 30.345 33.681 37.667 41.957 46.083 48.676 51.000 53.791 55.748
38 24.791 25.955 27.733 29.327 31.241 34.626 38.667 43.011 47.187 49.808 52.158 54.979 56.956
39 25.586 26.770 28.576 30.196 32.139 35.572 39.667 44.065 48.289 50.940 53.314 56.164 58.160

40 26.384 27.587 29.422 31.066 33.038 36.519 40.667 45.118 49.390 52.069 54.469 57.347 59.363
41 27.184 28.406 30.270 31.938 33.938 37.466 41.667 46.171 50.490 53.197 55.622 58.528 60.563
42 27.986 29.228 31.120 32.812 34.839 38.414 42.667 47.223 51.589 54.324 56.772 59.717 61.761
43 28.791 30.051 31.970 33.686 35.742 39.363 43.667 48.274 52.686 55.449 57.921 60.884 62.956
44 29.598 30.877 32.824 34.563 36.646 40.312 44.667 49.325 53.782 56.572 59.068 62.059 64.150
45 30.408 31.704 33.678 35.441 37.550 41.262 45.667 50.375 54.878 57.695 60.214 63.231 65.340
46 31.219 32.534 34.534 36.320 38.456 42.212 46.667 51.425 55.972 58.816 61.358 64.402 66.529
47 32.032 33.365 35.392 37.200 39.363 43.163 47.667 52.474 57.065 59.936 62.500 65.571 67.716
48 32.848 34.198 36.250 38.082 40.270 44.115 48.667 53.522 58.158 61.054 63.641 66.738 68.901
49 33.664 35.032 37.111 38.965 41.179 45.067 49.667 54.571 59.249 62.171 64.780 67.903 70.084

Source: Reprinted from Cameron, J.M., Ind. Qual. Control, 9(1), 39, 1952. With permission.
Appendix
TABLE T6.1
Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Appendix

Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

0S Ac = 0 44.893 2.303 np .0101 .0513 .105 .288 .693 1.386 2.303 2.996 4.605 5.298 6.908 7.601 9.206
Re = 1 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
XD* Ac = # 1 32.655 1.636 np .0100 .0501 .101 .259 .573 1.053 1.636 2.057 2.995 3.389 4.286 4.668 5.542
Re = 1 2 ASN 1.990 1.951 1.904 1.772 1.564 1.349 1.195 1.128 1.050 1.034 1.014 1.009 1.004
n1
XM Ac = # # 0 0 1 33.254 .838 np .00501 .0252 .0508 .132 .294 .539 .838 1.057 1.566 1.788 2.312 2.541 3.071
23
Re = 1 1 2 2 3 ASN 2.995 2.973 2.941 2.821 2.538 2.119 1.732 1.536 1.271 1.205 1.111 1.086 1.049
44 n1
XXD Ac = 0 1 20.193 2.302 np .0459 .114 .176 .347 .713 1.388 2.302 2.993 4.571 5.201 6.815 7.490 9.048
Re = 2 2
é For this plan ù ASN 1.219 1.507 1.737 2.226 2.748 2.732 2.151 1.750 1.237 1.143 1.037 1.021 1.005
ê ú n1
ëonly use n2 = 5n1 û
XXM Ac = # # 0 0 1 20.204 .891 np .00968 .0441 .0817 .183 .357 .602 .891 1.102 1.593 1.808 2.321 2.546 3.074
23
Re = 1 2 2 2 3 ASN 3.018 3.067 3.095 3.072 2.834 2.383 1.927 1.685 1.345 1.259 1.135 1.103 1.056
44 n1
1S Ac = 1 10.958 3.890 np .149 .355 .532 .961 1.678 2.693 3.890 4.744 6.638 7.430 9.234 10.000 11.759
Re = 2 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
1D Ac = 0 1 12.029 2.490 np .0860 .207 .310 .566 1.006 1.661 2.490 3.124 4.649 5.324 6.914 7.604 9.209
Re = 2 2 ASN 1.079 1.168 1.228 1.321 1.368 1.316 1.206 1.137 1.045 1.026 1.007 1.004 1.001
n1
1M Ac = # # 0 0 1 8.903 .917 np .0459 .103 .148 .252 .416 .643 .917 1.121 1.602 1.815 2.325 2.549 3.075
12
(Continued)
653
TABLE T6.1 (Continued)
654

Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

Re = 2 2 2 3 3 ASN 3.254 3.501 3.637 3.774 3.640 3.169 2.601 2.270 1.761 1.618 1.388 1.319 1.205
33 n1
2S Ac = 2 6.506 5.322 np .436 .818 1.102 1.727 2.674 3.920 5.322 6.296 8.406 9.274 11.230 12.053 13.934
Re = 3 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
2D Ac = 0 3 5.357 3.402 np .363 .635 .827 1.231 1.816 2.566 3.402 3.986 5.290 5.852 7.201 7.810 9.295
Re = 3 4 ASN 1.298 1.443 1.511 1.581 1.564 1.450 1.306 1.222 1.097 1.066 1.025 1.016 1.005
n1
2M Ac = # 0 0 1 2 6.244 1.355 np .111 .217 .293 .451 .683 .988 1.355 1.635 2.343 2.671 3.458 3.803 4.602
34
Re = 2 3 3 4 4 ASN 2.432 2.789 2.983 3.207 3.165 2.776 2.261 1.950 1.470 1.344 1.167 1.122 1.060
55 n1
3S Ac = 3 4.891 6.681 np .823 1.366 1.745 2.535 3.672 5.109 6.681 7.754 10.045 10.978 13.062 13.935 15.922
Re = 4 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
3D Ac = 1 4 4.398 4.398 np .635 1.000 1.246 1.750 2.465 3.373 4.398 5.130 6.808 7.542 9.270 10.019 11.757
Re = 4 5 ASN 1.130 1.245 1.316 1.421 1.470 1.414 1.293 1.211 1.084 1.053 1.017 1.010 1.003
n1
3M Ac = # 0 1 2 3 4.672 1.626 np .200 .348 .446 .642 .910 1.246 1.626 1.901 2.553 2.848 3.566 3.887 4.650
46
Re = 3 3 4 5 6 ASN 2.461 2.820 3.026 3.286 3.288 2.935 2.450 2.156 1.693 1.559 1.340 1.274 1.163
67 n1
4S Ac = 4 4.058 7.994 np 1.279 1.970 2.433 3.369 4.671 6.274 7.994 9.154 11.605 12.594 14.795 15.711 17.792
Re = 5 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
4D Ac = 3 5 4.102 6.699 np 1.099 1.633 1.992 2.728 3.789 5.162 6.699 7.762 10.047 10.978 13.062 13.933 15.909
Appendix

(Continued)
TABLE T6.1 (Continued)
Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Appendix

Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

Re = 6 6 ASN 1.025 1.077 1.125 1.233 1.341 1.345 1.242 1.164 1.055 1.033 1.009 1.005 1.001
n1
4M Ac = # 1 2 3 4 4.814 2.118 np .266 .440 .558 .798 1.141 1.591 2.118 2.502 3.385 3.763 4.640 5.016 5.884
56
Re = 3 4 4 6 6 ASN 2.128 2.300 2.417 2.590 2.618 2.384 2.021 1.792 1.427 1.326 1.174 1.132 1.070
77 n1
5S Ac = 5 3.550 9.275 np 1.785 2.613 3.152 4.219 5.670 7.423 9.275 10.513 13.109 14.150 16.455 17.411 19.578
Re = 6 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
5D Ac = 2 6 3.547 5.781 np 1.116 1.630 1.959 2.607 3.490 4.579 5.781 6.627 8.537 9.357 11.253 12.066 13.928
Re = 5 7 ASN 1.097 1.199 1.263 1.360 1.405 1.352 1.243 1.171 1.064 1.039 1.012 1.007 1.002
n1
5M Ac = # 1 2 3 5 3.243 2.270 np .490 .700 .830 1.079 1.410 1.814 2.270 2.604 3.411 3.776 4.642 5.017 5.884
79
Re = 4 5 6 7 8 ASN 2.496 2.906 3.143 3.459 3.516 3.188 2.677 2.347 1.791 1.628 1.367 1.292 1.171
9 10 n1
6S Ac = 6 3.206 10.532 np 2.330 3.285 3.895 5.083 6.670 8.558 10.532 11.842 14.571 15.660 18.062 19.056 21.302
Re = 7 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
6D Ac = 3 7 3.217 6.914 np 1.559 2.149 2.525 3.262 4.268 5.519 6.914 7.898 10.087 11.000 13.068 13.936 15.903
Re = 8 8 ASN 1.073 1.169 1.243 1.393 1.548 1.608 1.525 1.422 1.203 1.138 1.051 1.032 1.011
n1
6M Ac = 0 2 4 5 7 3.452 3.134 np .604 .908 1.093 1.439 1.894 2.463 3.134 3.645 4.917 5.511 6.983 7.646 9.222
10 11
Re = 4 5 8 9 10 ASN 1.584 1.928 2.134 2.425 2.519 2.288 1.902 1.663 1.304 1.211 1.083 1.054 1.018
12 12 n1

(Continued)
655
TABLE T6.1 (Continued)
656

Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

7S Ac = 7 2.957 11.771 np 2.906 3.981 4.656 5.956 7.669 9.684 11.771 13.148 16.000 17.134 19.627 20.655 22.976
Re = 8 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
7D Ac = 3 8 2.951 7.162 np 1.796 2.427 2.822 3.584 4.599 5.826 7.162 8.093 10.174 11.057 13.085 13.946 15.914
Re = 7 9 ASN 1.106 1.215 1.288 1.409 1.492 1.467 1.352 1.262 1.110 1.072 1.024 1.014 1.004
n1
7M Ac = 0 1 3 5 7 2.892 2.959 np .713 1.023 1.200 1.518 1.921 2.403 2.959 3.400 4.686 5.337 6.915 7.604 9.210
10 13
Re = 4 6 8 10 ASN 2.022 2.586 2.882 3.255 3.325 2.966 2.397 2.019 1.406 1.261 1.091 1.057 1.018
11 12 14 n1
8S Ac = 8 2.768 12.995 np 3.507 4.695 5.432 6.838 8.669 10.802 12.995 14.435 17.403 18.578 21.157 22.218 24.600
Re = 9 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
8D Ac = 3 11 2.668 8.248 np 2.268 3.092 3.583 4.489 5.628 6.925 8.248 9.121 10.964 11.722 13.470 14.232 16.046
Re = 7 12 ASN 1.185 1.335 1.409 1.488 1.478 1.375 1.248 1.176 1.075 1.051 1.019 1.012 1.004
n1
8M Ac = 0 2 4 6 9 2.840 3.314 np .787 1.167 1.375 1.739 2.190 2.720 3.314 3.761 4.936 5.517 6.983 7.646 9.219
12 14
Re = 4 7 9 11 ASN 1.806 2.320 2.599 2.963 3.063 2.765 2.264 1.934 1.400 1.263 1.093 1.058 1.019
12 14 15 n1
9S Ac = 9 2.619 14.206 np 4.130 5.425 6.221 7.726 9.669 11.914 14.206 15.705 18.783 19.999 22.658 23.751 26.198
Re = 10 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
9D Ac = 5 11 2.587 9.533 np 2.871 3.685 4.184 5.134 6.385 7.893 9.533 10.670 13.152 14.174 16.460 17.412 19.564
Re = 12 12 ASN 1.071 1.167 1.243 1.401 1.584 1.694 1.662 1.573 1.328 1.241 1.105 1.071 1.026
n1
(Continued)
Appendix
TABLE T6.1 (Continued)
Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Appendix

Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

9M Ac = 1 3 5 8 11 2.813 4.219 np 1.117 1.500 1.719 2.123 2.659 3.349 4.219 4.924 6.682 7.454 9.239 10.000 11.754
13 15
Re = 5 8 10 12 ASN 1.526 1.928 2.167 2.521 2.667 2.414 1.937 1.635 1.230 1.145 1.048 1.029 1.009
14 16 16 n1
10S Ac = 10 2.497 15.407 np 4.771 6.169 7.021 8.620 10.669 13.020 15.407 16.962 20.145 21.398 24.135 25.257 27.768
Re = 11 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
10D Ac = 5 12 2.486 9.732 np 3.055 3.914 4.433 5.406 6.663 8.147 9.732 10.822 13.216 14.214 16.472 17.420 19.562
Re = 9 13 ASN 1.085 1.183 1.248 1.357 1.426 1.394 1.286 1.206 1.081 1.051 1.016 1.009 1.003
n1
10M Ac = 0 3 6 8 11 2.516 3.927 np 1.144 1.561 1.792 2.199 2.701 3.286 3.927 4.391 5.498 6.003 7.266 7.851 9.303
14 18
Re = 5 8 10 13 ASN 1.927 2.357 2.602 2.939 3.034 2.750 2.282 1.982 1.501 1.368 1.170 1.118 1.047
15 17 19 n1
11S Ac = 11 2.397 16.598 np 5.428 6.924 7.829 9.519 11.668 14.121 16.598 18.208 21.490 22.779 25.590 26.741 29.313
Re = 12 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
11D Ac = 5 12 2.438 9.766 np 3.165 4.006 4.517 5.477 6.723 8.194 9.766 10.847 13.226 14.221 16.474 17.420 19.566
Re = 13 13 ASN 1.101 1.216 1.299 1.463 1.642 1.752 1.737 1.664 1.429 1.332 1.163 1.115 1.047
n1
11M Ac = 1 4 7 10 2.567 4.657 np 1.362 1.814 2.074 2.539 3.125 3.834 4.657 5.285 6.847 7.561 9.273 10.022 11.758
13 16 20
Re = 6 8 11 14 ASN 1.548 1.912 2.132 2.452 2.575 2.357 1.970 1.726 1.348 1.246 1.101 1.066 1.024
17 21 21 n1
12S Ac = 12 2.312 17.782 np 6.099 7.690 8.646 10.422 12.668 15.217 17.782 19.443 22.821 24.145 27.027 28.206 30.836
(Continued)
657
TABLE T6.1 (Continued)
658

Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

Re = 13 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
12D Ac = 6 15 2.289 11.233 np 3.884 4.907 5.507 6.606 7.990 9.580 11.233 12.351 14.781 15.795 18.105 19.080 21.291
Re = 10 16 ASN 1.092 1.196 1.261 1.358 1.403 1.353 1.246 1.175 1.068 1.043 1.014 1.008 1.002
n1
12M Ac = 0 3 7 10 2.249 4.400 np 1.510 1.956 2.203 2.640 3.172 3.774 4.400 4.828 5.802 6.242 7.372 7.916 9.316
14 18 21
Re = 6 9 12 15 ASN 2.215 2.655 2.892 3.203 3.274 2.982 2.516 2.214 1.708 1.555 1.299 1.222 1.102
17 20 22 n1
13S Ac = 13 2.240 18.958 np 6.782 8.464 9.470 11.329 13.668 16.310 18.958 20.669 24.139 25.497 28.447 29.652 32.339
Re = 14 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
13D Ac = 5 14 2.227 10.474 np 3.797 4.704 5.246 6.249 7.520 8.976 10.474 11.474 13.624 14.523 16.608 17.509 19.589
Re = 15 15 ASN 1.184 1.332 1.427 1.591 1.750 1.842 1.838 1.789 1.603 1.511 1.312 1.242 1.122
n1
13M Ac = 1 3 5 8 11 2.474 4.298 np 1.244 1.737 1.987 2.408 2.920 3.537 4.298 4.947 6.682 7.454 9.239 10.001 11.752
14 19
Re = 5 9 12 15 ASN 1.669 2.326 2.691 3.188 3.368 3.009 2.315 1.852 1.266 1.161 1.050 1.030 1.009
18 20 20 n1
14S Ac = 14 2.177 20.128 np 7.477 9.246 10.300 12.239 14.668 17.400 20.128 21.886 25.446 26.836 29.853 31.084 33.824
Re = 15 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
14D Ac = 7 18 2.176 12.722 np 4.652 5.847 6.534 7.769 9.286 10.989 12.722 13.876 16.345 17.370 19.712 20.707 22.974
Re = 11 19 ASN 1.091 1.199 1.263 1.352 1.380 1.317 1.214 1.150 1.058 1.037 1.012 1.007 1.002
n1
(Continued)
Appendix
TABLE T6.1 (Continued)
Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Appendix

Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

14M Ac = 1 4 8 12 2.185 5.112 np 1.844 2.340 2.618 3.110 3.708 4.387 5.112 5.632 6.955 7.612 9.279 10.024 11.757
17 21 25
Re = 7 10 13 ASN 1.916 2.375 2.624 2.952 3.042 2.780 2.350 2.067 1.575 1.424 1.192 1.133 1.053
17 20 23 26 n1
15S Ac = 15 2.122 21.292 np 8.181 10.036 11.135 13.152 15.668 18.487 21.292 23.097 26.743 28.164 31.245 32.501 35.294
Re = 16 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
15D Ac = 5 16 2.091 11.405 np 4.476 5.455 6.033 7.094 8.419 9.908 11.405 12.381 14.405 15.224 17.088 17.890 19.782
Re = 17 17 ASN 1.293 1.463 1.559 1.710 1.838 1.904 1.898 1.861 1.716 1.640 1.459 1.384 1.235
n1
15M Ac = 2 7 13 18 2.142 6.795 np 2.553 3.173 3.529 4.167 4.953 5.850 6.795 7.453 8.971 9.658 11.362 12.132 13.948
23 28 30
Re = 9 12 16 ASN 1.606 1.914 2.089 2.339 2.443 2.301 2.028 1.842 1.507 1.397 1.205 1.148 1.064
21 26 31 31 n1
18S Ac = 18 1.990 24.756 np 10.346 12.442 13.672 15.907 18.668 21.731 24.756 26.692 30.581 32.091 35.353 36.679 39.622
Re = 19 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
18D Ac = 9 23 1.955 15.524 np 6.559 7.940 8.722 10.111 11.796 13.659 15.524 16.748 19.329 20.391 22.818 23.853 26.219
Re = 14 24 ASN 1.120 1.244 1.315 1.412 1.442 1.374 1.260 1.188 1.079 1.052 1.018 1.011 1.003
n1
18M Ac = 1 6 11 16 1.990 6.225 np 2.506 3.128 3.462 4.035 4.712 5.460 6.225 6.744 7.917 8.443 9.765 10.381 11.912
22 27 32
Re = 8 12 17 ASN 2.009 2.443 2.681 2.999 3.087 2.824 2.389 2.107 1.639 1.499 1.270 1.202 1.096
22 25 29 33 n1
21S Ac = 21 1.892 28.184 np 12.574 14.894 16.244 18.682 21.668 24.956 28.184 30.240 34.355 35.947 39.376 40.768 43.850
(Continued)
659
TABLE T6.1 (Continued)
660

Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

Re = 22 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
21D Ac = 11 26 1.882 18.909 np 7.843 9.329 10.170 11.666 13.486 15.510 17.555 18.909 21.792 22.978 25.656 26.777 29.312
Re = 16 27 ASN 1.094 1.201 1.268 1.367 1.413 1.363 1.256 1.185 1.075 1.048 1.016 1.009 1.003
n1
21M Ac = 2 7 13 19 1.893 7.083 np 3.071 3.741 4.100 4.713 5.440 6.246 7.083 7.664 9.044 9.696 11.367 12.133 13.948
25 31 37
Re = 9 14 19 ASN 1.912 2.370 2.621 2.962 3.077 2.830 2.392 2.102 1.606 1.457 1.219 1.155 1.065
25 29 33 38 n1
27S Ac = 27 1.757 34.959 np 17.175 19.901 21.469 24.273 27.667 31.365 34.959 37.234 41.757 43.497 47.231 48.740 52.077
Re = 28 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
27D Ac = 15 34 1.760 22.183 np 10.797 12.605 13.613 15.382 17.504 19.839 22.183 23.727 26.993 28.323 31.292 32.526 35.284
Re = 20 35 ASN 1.074 1.170 1.231 1.324 1.367 1.320 1.221 1.156 1.060 1.038 1.012 1.007 1.002
n1
27M Ac = 3 10 17 1.805 8.738 np 3.936 4.841 5.301 6.050 6.896 7.807 8.738 9.380 10.890 11.586 13.318 14.102 15.968
24 32 40 48
Re = 10 17 24 ASN 1.746 2.219 2.484 2.841 2.951 2.688 2.245 1.958 1.490 1.357 1.162 1.112 1.045
31 37 43 49 n1
30S Ac = 30 1.707 38.315 np 19.532 22.445 24.113 27.086 30.667 34.552 38.315 40.691 45.401 47.210 51.085 52.647 56.102
Re = 31 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
30D Ac = 17 37 1.724 24.257 np 12.177 14.072 15.130 16.995 19.243 21.735 24.257 25.928 29.453 30.876 34.015 35.305 38.169
Re = 22 38 ASN 1.063 1.148 1.205 1.297 1.349 1.311 1.216 1.152 1.056 1.035 1.011 1.006 1.002
n1
(Continued)
Appendix
TABLE T6.1 (Continued)
Unity Values for Construction and Evaluation of Single-, Double-, and Multiple-Sampling Plans
Probability of Acceptance
Appendix

Acceptance R=
Plan Numbers p2/p1 np2 .99 .95 .90 .75 .50 .25 .10 .05 .01 .005 .001 .0005 .0001

30M Ac = 4 11 19 1.708 9.660 np 4.817 5.656 6.096 6.841 7.713 8.669 9.660 10.356 12.058 12.869 14.873 15.756 17.785
27 36 45 53
Re = 12 19 27 ASN 1.840 2.320 2.586 2.951 3.084 2.847 2.411 2.114 1.596 1.441 1.206 1.145 1.061
34 40 47 54 n1
41S Ac = 41 1.581 50.490 np 28.406 31.938 33.938 37.466 41.667 46.171 50.490 53.197 58.528 60.564 64.904 66.648 70.488
Re = 42 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
41D Ac = 23 52 1.584 31.843 np 17.706 20.108 21.415 23.661 26.284 29.094 31.843 33.620 37.311 38.801 42.131 43.517 46.616
Re = 29 53 ASN 1.080 1.183 1.248 1.340 1.375 1.319 1.219 1.155 1.062 1.039 1.013 1.008 1.002
n1
41M Ac = 6 16 26 1.574 12.617 np 6.942 8.014 8.552 9.435 10.440 11.519 12.617 13.378 15.211 16.076 18.205 19.143 21.306
37 49 61 72
Re = 15 25 36 ASN 1.842 2.370 2.660 3.054 3.195 2.938 2.470 2.155 1.613 1.453 1.213 1.151 1.065
46 55 64 73 n1
44S Ac = 44 1.556 53.783 np 30.877 34.563 36.646 40.312 44.667 49.325 53.783 56.573 62.058 64.150 68.607 70.395 74.332
Re = 45 ASN 1 1 1 1 1 1 1 1 1 1 1 1 1
n1
44D Ac = 25 56 1.561 34.068 np 19.292 21.820 23.192 25.544 28.282 31.209 34.068 35.916 39.750 41.296 44.739 46.166 49.357
Re = 31 57 ASN 1.075 1.174 1.237 1.328 1.363 1.309 1.211 1.149 1.058 1.037 1.012 1.007 1.002
n1
44M Ac = 6 17 29 1.538 13.372 np 7.614 8.695 9.239 10.139 11.168 12.270 13.372 14.112 15.784 16.537 18.424 19.289 21.358
40 53 65 77
Re = 16 27 39 ASN 1.971 2.472 2.747 3.127 3.274 3.036 2.582 2.272 1.731 1.565 1.294 1.217 1.101
49 58 68 78 n1

Source: Reprinted from Schilling, E.G. and Johnson, L.I., J. Qual. Technol., 12(4), 220, 1980. With permission.
Note: n1 = n2 … nk, # indicates acceptance not allowed at a given stage.
661
TABLE T7.1
662

Statistical Research Group: Table of Sequential Sampling Plans

α = .05 α = .05

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.0002 .0020 .10 1.2543 .9770 .000782 1250 2 1487 1568 847 .001 .01 .10 1.2504 .9739 .003915 249 2 296 312 169
.0003 .0025 .10 1.3618 1.0607 .001038 1022 2 1273 1393 766 " " .50 .9961 .2777 " 71 2 73 71 59
" .0040 .10 1.1143 .8679 .001429 608 2 681 677 356 " .011 .10 1.2003 .9349 .004178 224 2 261 270 145
" .0055 .10 .9919 .7726 .001790 432 1 459 429 220 " .013 .10 1.1216 .8736 .004689 187 2 210 210 111
.0005 .002 .10 2.0827 1.6222 .001082 1499 3 2468 3125 1866 " .02 .10 .9587 .7467 .006369 118 1 123 113 58
" " .50 1.6592 .4625 " 428 2 612 710 652 " " .50 .7637 .2129 " 34 1 31 26 20
" .003 .10 1.6109 1.2547 .001396 899 2 1241 1540 826 " .03 .10 .8425 .6562 .008587 77 1 77 65 32
" " .50 1.2833 .3577 " 257 2 308 329 288 " " .50 .6712 .1871 " 22 1 19 15 11
" .004 .10 1.3876 1.0808 .001684 642 2 809 892 493 " .04 .10 .7752 .6038 .01068 57 1 55 44 22
" " .50 1.1054 .3081 " 183 2 201 203 172 " " .50 .6175 .1721 " 17 1 14 10 8
" .005 .10 1.2528 .9758 .001956 499 2 594 626 338 " .05 .10 .7295 .5682 .01269 45 1 43 33 16
" " .50 .9980 .2782 " 143 2 147 142 118 " " .50 .5811 .1620 " 13 1 11 8 6
" .006 .10 1.1606 .9040 .002216 408 2 467 475 252 " .06 .10 .6956 .5418 .01465 37 1 35 26 13
" " .50 .9246 .2577 " 117 1 116 108 88 " " .50 .5541 .1545 " 11 1 9 6 4
" .007 .10 1.0925 .8510 .002466 346 2 383 378 198 .0015 .0055 .10 2.2177 1.7274 .003080 561 3 968 1248 753
" " .50 .8704 .2426 " 99 1 95 86 69 " .0085 .10 1.6596 1.2926 .004039 321 2 451 533 306
" .008 .10 1.0937 .8098 .002709 299 2 325 312 162 " .0110 .10 1.4437 1.1245 .004775 236 2 304 342 191
" " .50 .8282 .2309 " 86 1 81 71 56 " .0130 .10 1.3313 1.0370 .005336 195 2 239 260 143
" .009 .10 .9971 .7766 .002946 264 1 281 264 135 " .0150 .10 1.2479 .9720 .005877 166 2 197 208 112
.00055 .0037 .10 1.5138 1.1791 .001653 714 2 947 1081 608 " .01875 .10 1.1365 .8852 .006852 130 2 147 148 79
.001 .004 .10 2.0804 1.6204 .002165 749 3 1232 1561 932 .00175 .0125 .10 1.4617 1.1385 .005476 208 2 271 306 171
" .005 .10 1.7914 1.3953 .002487 562 2 831 1008 586 .002 .007 .10 2.2980 1.7899 .003993 449 3 795 1034 628
" " .50 1.4271 .3978 " 160 2 206 229 205 " .010 .10 1.7870 1.3918 .004976 280 2 414 502 292
" .0065 .10 1.5396 1.1992 .002941 408 2 547 630 356 " .013 .10 1.5351 1.1957 .005886 204 2 273 314 177
" .007 .10 1.4808 1.1534 .003086 374 2 490 555 311 " .016 .10 1.3806 1.0753 .006748 160 2 201 222 123
" " .50 1.1796 .3288 " 107 2 121 126 109 " .019 .10 1.2741 .9924 .007574 132 2 158 168 92
" .009 .10 1.3107 1.0209 .003646 281 2 342 368 201 " .023 .10 1.1732 .9138 .008632 106 2 122 125 67
.002 .025 .10 1.1339 .8832 .009147 97 2 109 110 59 .005 .015 .10 2.6070 2.0305 .009111 223 3 438 586 364
.0023 .0188 .10 1.3649 1.0631 .007875 135 2 169 186 103 " " .50 2.0768 .5789 " 64 3 109 133 127
(Continued)
Appendix
TABLE T7.1 (Continued)
Statistical Research Group: Table of Sequential Sampling Plans
α = .05 α = .05
Appendix

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.0025 .0075 .10 2.6190 2.0399 .004553 449 3 880 1179 731 " .02 .10 2.0624 1.6064 .01084 149 3 244 309 185
" " .50 2.0864 .5816 " 128 3 218 268 255 " " .50 1.6430 .4580 " 43 2 60 70 65
" .01 .10 2.0737 1.6152 .005415 299 3 491 622 372 " .03 .10 1.5906 1.2389 .01400 89 2 122 143 82
" " .50 1.6520 .4605 " 86 2 122 141 130 " " .50 1.2671 .3532 " 26 2 30 32 29
" .015 .10 1.6019 1.2477 .006989 179 2 246 288 164 " .04 .10 1.3664 1.0643 .01693 63 2 79 87 49
" " .50 1.2761 .3557 " 51 2 61 65 57 " " .50 1.0886 .3034 " 18 2 20 20 17
" .02 .10 1.3782 1.0735 .008440 128 2 160 177 98 " .05 .10 1.2305 .9585 .01970 49 2 58 61 33
" " .50 1.0980 .3061 " 37 2 40 40 34 " " .50 .9803 .2733 " 14 2 14 14 12
" .03 .10 1.1502 .8959 .01113 81 2 92 94 50 " .06 .10 1.1371 .8857 .02237 40 2 45 46 25
" " .50 .9163 .2554 " 23 1 23 21 18 " " .50 .9059 .2525 " 12 1 11 10 9
" .04 .10 1.0283 .8009 .01363 59 2 64 61 32 " .067 .10 1.0868 .8465 .02419 35 2 39 39 21
" " .50 .8192 .2283 " 17 1 16 14 11 " .07 .10 1.0679 .8318 .02496 34 2 37 36 19
" .05 .10 .9494 .7395 .01603 47 1 48 45 23 " " .50 .8507 .2371 " 10 1 9 8 7
" " .50 .7563 .2108 " 14 1 12 10 8 .006 .012 .10 4.1338 3.2198 .008659 372 5 1073 1551 1017
" .06 .10 .8928 .6954 .01834 38 1 39 34 18 " .018 .10 2.6022 2.0268 .01093 186 3 364 488 303
" " .50 .7112 .1983 " 11 1 10 8 6 " .021 .10 2.2795 1.7755 .01199 149 3 262 342 208
.0027 .025 .10 1.2856 1.0014 .01006 100 2 121 129 71 " .024 .10 2.0578 1.6028 .01301 124 3 203 257 154
.003 .009 .10 2.6166 2.0380 .005464 373 3 733 981 608 " .030 .10 1.7690 1.3779 .01496 93 2 136 165 97
" .014 .10 1.8629 1.4510 .007151 203 2 310 381 224 " .036 .10 1.5860 1.2353 .01682 74 2 101 119 68
" .018 .10 1.5996 1.2459 .008390 149 2 205 240 137 " .042 .10 1.4577 1.1354 .01860 62 2 80 91 51
" .0186 .10 1.5706 1.2233 .008570 143 2 195 226 129 " .050 .10 1.3347 1.0396 .02091 50 2 62 68 38
" .022 .10 1.4368 1.1191 .009565 118 2 151 170 95 " .06 .10 1.2255 .9546 .02368 41 2 48 51 28
" .026 .10 1.3241 1.0314 .01069 97 2 119 129 71 .0065 .027 .10 2.0004 1.5581 .01443 108 3 174 219 131
" .030 .10 1.2405 .9662 .01178 83 2 97 103 56 .0075 .015 .10 4.1248 3.2128 .01082 297 5 856 1238 812
" .035 .10 1.1611 .9044 .01310 70 2 79 81 44 " " .50 3.2860 .9160 " 85 4 212 281 284
" .036 .10 1.1476 .8939 .01336 67 2 76 78 42 " .02 .10 2.9093 2.2660 .01276 178 3 382 523 330
.005 .01 .10 4.1398 3.2245 .007216 447 5 1289 1863 1222 " " .50 2.3176 .6461 " 51 3 95 119 115
" " .50 3.2980 .9193 " 128 4 320 423 427 " .03 .10 2.0510 1.5975 .01627 99 3 161 205 123
(Continued)
663
TABLE T7.1 (Continued)
664

Statistical Research Group: Table of Sequential Sampling Plans

α = .05 α = .05

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.0075 .03 .50 1.6339 .4555 " 28 2 40 46 43 .011 .040 .10 2.1884 1.7046 .02253 76 3 131 169 103
" .04 .10 1.6930 1.3186 .01950 68 2 97 117 68 " .048 .10 1.9123 1.4895 .02523 60 2 93 116 69
" " .50 1.3487 .3760 " 20 2 24 27 24 " .056 .10 1.7266 1.3448 .02782 49 2 71 86 50
" .05 .10 1.4892 1.1599 .02255 52 2 68 78 45 " .066 .10 1.5632 1.2176 .03095 40 2 54 63 37
" " .50 1.1864 .3307 " 15 2 17 18 16 " .076 .10 1.4446 1.1252 .03398 34 2 43 50 28
" .06 .10 1.3546 1.0551 .02547 42 2 52 58 32 " .094 .10 1.2944 1.0082 .03925 26 2 32 35 19
" " .50 1.0791 .3008 " 12 2 13 13 11 .0115 .043 .10 2.1391 1.6661 .02397 70 3 118 152 92
" .07 .10 1.2574 .9794 .02830 35 2 42 45 25 .0118 .1175 .10 1.1986 .9336 .04691 20 2 24 25 14
" " .50 1.0017 .2792 " 10 2 10 10 9 .012 .056 .10 1.8224 1.4195 .02872 50 2 75 93 55
" .08 .10 1.1831 .9215 .03105 30 2 35 36 20 .014 .102 .10 1.3900 1.0827 .04496 25 2 31 35 20
" " .50 .9425 .2627 " 9 1 9 8 7 .015 .025 .10 5.5474 4.3209 .01958 221 6 835 1248 842
.01 .02 .10 4.1097 3.2010 .01444 222 5 639 925 607 " " .50 4.4193 1.2319 " 63 5 207 284 294
" " .50 3.2740 .9126 " 64 4 159 210 212 " .03 .10 4.0796 3.1776 .02166 147 5 423 612 402
" .025 .10 3.1027 2.4167 .01639 148 4 335 465 296 " " .50 3.2500 .9059 " 42 4 105 139 141
" " .50 2.4718 .6890 " 43 3 83 106 104 " .04 .10 2.8716 2.2367 .02554 88 3 188 258 163
" .03 .10 2.5829 2.0118 .01824 111 3 216 290 181 " " .50 2.2876 .6377 " 25 3 47 59 57
" " .50 2.0577 .5736 " 32 3 54 66 63 " .05 .10 2.3307 1.8153 .02917 63 3 113 149 92
" .04 .10 2.0397 1.5887 .02172 74 3 120 153 92 " " .50 1.8567 .5176 " 18 2 28 34 32
" " .50 1.6249 .4529 " 21 2 30 35 32 " .06 .10 2.0169 1.5710 .03263 49 3 79 100 61
" .05 .10 1.7510 1.3639 .02499 55 2 81 98 58 " " .50 1.6068 .4479 " 14 2 20 23 21
" " .50 1.3949 .3888 " 16 2 20 22 20 " .07 .10 1.8089 1.4089 .03596 40 2 60 74 44
" .06 .10 1.5678 1.2211 .02811 44 2 60 70 40 " " .50 1.4410 .4017 " 12 2 15 17 15
" " .50 1.2490 .3482 " 13 2 15 16 14 .02 .03 .10 6.9527 5.4154 .02467 220 8 1027 1565 1073
" .07 .10 1.4391 1.1209 .03113 37 2 47 53 30 " " .50 5.5388 1.5440 " 63 6 255 355 375
" " .50 1.1465 .3196 " 11 2 12 12 11 " .035 .10 5.0264 3.9150 .02682 146 6 508 754 505
" .08 .10 1.3426 1.0458 .03406 31 2 38 43 24 " " .50 4.0042 1.1162 " 42 5 126 171 177
" " .50 1.0696 .2982 " 9 2 10 10 8 " .04 .10 4.0495 3.1541 .02889 110 5 314 455 300
(Continued)
Appendix
TABLE T7.1 (Continued)
Statistical Research Group: Table of Sequential Sampling Plans
α = .05 α = .05
Appendix

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.011 .020 .10 4.7619 3.7090 .01506 247 5 809 1191 793 " " .50 3.2260 .8992 " 32 4 78 103 105
" .025 .10 3.4605 2.6954 .01707 158 4 393 556 359 " .05 .10 3.0509 2.3763 .03282 73 4 164 228 146
" .032 .10 2.6534 2.0667 .01970 105 3 210 284 177 " " .50 2.4305 .6775 " 21 3 41 52 51
.02 .06 .10 2.5348 1.9743 .03655 55 3 106 142 89 .03 .045 .50 5.4687 1.5244 " 42 6 167 234 247
" " .50 2.0193 .5629 " 16 3 26 32 31 " .05 .10 5.4365 4.2345 .03919 109 6 408 611 413
" .07 .10 2.2146 1.7250 .04012 43 3 76 99 61 " " .50 4.3309 1.2073 " 31 5 101 139 144
" " .50 1.7643 .4918 " 13 2 19 23 21 " .06 .10 3.9891 3.1071 .04336 72 5 206 299 197
" .08 .10 1.9941 1.5532 .04359 36 3 58 74 45 " " .50 3.1779 .8858 " 21 4 51 68 69
" " .50 1.5886 .4428 " 11 2 14 17 16 " .07 .10 3.2498 2.5312 .04735 54 4 129 182 118
" .0820 .10 1.9578 1.5249 .04427 35 3 56 71 43 " " .50 2.5889 .7217 " 16 3 32 41 41
" .085 .10 1.9071 1.4855 .04528 33 2 52 66 39 " .08 .10 2.7960 2.1778 .05119 43 3 91 125 80
" .09 .10 1.8315 1.4265 .04696 31 2 47 58 35 " " .50 2.2274 .6209 " 13 3 23 28 28
" " .50 1.4590 .4067 " 9 2 12 13 12 " .086 .10 2.5978 2.0234 .05345 38 3 76 104 66
" .096 .10 1.7524 1.3650 .04894 28 2 42 51 31 " .09 .10 2.4864 1.9367 .05493 36 3 69 93 58
" .10 .10 1.7056 1.3285 .05025 27 2 39 47 28 " " .50 1.9808 .5522 " 11 3 17 21 20
" " .50 1.3588 .3788 " 8 2 10 11 10 " .10 .10 2.2601 1.7604 .05857 31 3 55 72 45
" .114 .10 1.5697 1.2227 .05476 23 2 31 37 22 " " .50 1.8005 .5019 " 9 2 14 16 16
" .172 .10 1.2457 .9703 .07264 14 2 16 18 10 " .11 .10 2.0864 1.6251 .06213 27 3 45 58 36
" .178 .10 1.2238 .9532 .07444 13 2 16 17 10 " " .50 1.6621 .4633 " 8 2 11 13 13
.021 .037 .10 4.9588 3.8624 .02827 137 6 471 697 467 " .118 .10 1.9735 1.5371 .06494 24 3 39 50 31
" .043 .10 3.9090 3.0447 .03074 100 5 277 399 262 " .12 .10 1.9481 1.5174 .06563 24 3 38 48 29
" .052 .10 3.0785 2.3978 .03427 70 4 160 223 143 " " .50 1.5520 .4326 " 7 2 9 11 10
" .062 .10 2.5683 2.0004 .03801 53 3 104 140 88 " .13 .10 1.8350 1.4293 .06908 21 2 32 41 25
.022 .04 .10 4.6890 3.6522 .03014 122 5 398 586 391 " " .50 1.4618 .4075 " 6 2 8 9 9
" .136 .10 1.4856 1.1571 .06370 19 2 25 29 17 " .15 .10 1.6597 1.2927 .07583 18 2 25 31 18
.0255 .175 .10 1.3812 1.0758 .07958 14 2 18 20 12 " " .50 1.3222 .3686 " 5 2 6 7 6
.026 .107 .10 1.9249 1.4993 .05782 26 3 42 53 32 " .20 .10 1.3831 1.0773 .09220 12 2 15 18 11
" .115 .10 1.8263 1.4225 .06055 24 2 36 46 28 " " .50 1.1018 .3071 " 4 2 4 4 4
.027 .178 .10 1.4068 1.0957 .08208 14 2 18 20 12 .031 .076 .10 3.0609 2.3841 .05036 47 4 109 153 98
(Continued)
665
TABLE T7.1 (Continued)
666

Statistical Research Group: Table of Sequential Sampling Plans

α = .05 α = .05

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.028 .225 .10 1.2510 .9744 .09803 10 2 12 14 8 " .155 .10 1.6551 1.2891 .07841 16 2 24 30 18
.03 .04 .10 9.6978 7.5535 .03477 218 11 1403 2183 1524 .032 .057 .10 4.7895 3.7305 .04336 86 5 291 431 289
" " .50 7.7256 2.1535 " 62 9 348 496 533 " .066 .10 3.8048 2.9635 .04707 63 4 174 251 165
" .045 .10 6.8647 5.3469 .03701 145 8 675 1030 707 .033 .145 .10 1.8027 1.4041 .07678 18 2 28 36 22
.035 .225 .10 1.3896 1.0823 .1054 11 2 14 16 10 .04 .25 .50 1.1073 .3073 " 3 2 3 3 3
.0375 .155 .10 1.8656 1.4531 .08404 18 3 28 35 22 " .317 .10 1.1991 .9340 .1412 7 2 8 9 6
.04 .06 .10 6.7767 5.2783 .04936 107 8 499 762 524 .041 .092 .10 3.3497 2.6091 .06333 42 4 103 147 96
" " .50 5.3986 1.5049 " 31 6 124 173 183 " .102 .10 2.9580 2.3040 .06726 35 4 78 109 70
" .07 .10 4.8876 3.8069 .05369 71 6 246 366 246 .042 .072 .10 5.0636 3.9440 .05574 71 6 254 379 256
" " .50 3.8937 1.0854 " 21 5 61 83 86 " .082 .10 4.0612 3.1633 .05993 53 5 156 228 151
" .08 .10 3.9287 3.0600 .05785 53 5 152 221 146 " .195 .10 1.6909 1.3171 .1018 13 2 20 24 15
" " .50 3.1298 .8724 " 16 4 38 50 51 .0475 .1975 .10 1.5106 1.4103 .1073 14 3 21 27 17
" .09 .10 3.3437 2.6044 .06188 43 4 105 150 98 .048 .192 .10 1.8644 1.4522 .1058 14 3 22 29 18
" " .50 2.6637 .7425 " 13 3 26 34 34 .049 .16 .10 2.2107 1.7219 .09493 19 3 33 44 28
" .10 .10 2.9469 2.2953 .06580 35 4 79 110 71 .05 .07 .10 8.0793 6.2929 .05948 106 9 588 909 631
" " .50 2.3476 .6544 " 10 3 20 25 25 " " .50 6.4363 1.7941 " 31 7 146 206 221
" .11 .10 2.6583 2.0705 .06963 30 3 62 85 54 " .08 .10 5.7567 4.4838 .06391 71 7 286 431 294
" " .50 2.1177 .5903 " 9 3 15 19 19 " " .50 4.5860 1.2784 " 21 5 71 98 103
" .118 .10 2.4777 1.9299 .07264 27 3 52 71 45 " .09 .10 4.5820 3.5689 .06819 53 5 174 257 173
.04 .12 .10 2.4378 1.8988 .07339 26 3 50 68 43 " " .50 3.6502 1.0175 " 15 4 43 58 60
" " .50 1.9421 .5414 " 8 3 12 15 15 " .10 .10 3.8682 3.0129 .07236 42 5 119 174 115
" .13 .10 2.2632 1.7628 .07708 23 3 42 56 35 " " .50 3.0816 .8590 " 12 4 30 39 40
" " .50 1.8030 .5026 " 7 2 10 13 12 " .11 .10 3.3857 2.6371 .07642 35 4 88 126 83
" .138 .10 2.1473 1.6725 .08000 21 3 37 49 30 " " .50 2.6972 .7519 " 10 3 22 29 29
" .14 .10 2.1210 1.6520 .08072 21 3 36 47 29 " .12 .10 3.0361 2.3648 .08040 30 4 69 97 63
" " .50 1.6896 .4710 " 6 2 9 11 10 " " .50 2.4187 .6742 " 9 3 17 22 22
" .15 .10 2.0024 1.5597 .08431 19 3 31 40 25 " .13 .10 2.7699 2.1575 .08430 26 4 56 77 50
" " .50 1.5952 .4447 " 6 2 8 9 9 " " .50 2.2066 .6151 " 8 3 14 18 17
" .17 .10 1.8151 1.4137 .09137 16 2 24 31 19 " .138 .10 2.5982 2.0237 .08738 24 3 48 66 42
(Continued)
Appendix
TABLE T7.1 (Continued)
Statistical Research Group: Table of Sequential Sampling Plans
α = .05 α = .05
Appendix

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

" " .50 1.4460 .4031 " 5 2 6 7 7 " .14 .10 2.5598 1.9938 .08815 23 3 46 63 41
" .20 .10 1.6131 1.2565 .1018 13 2 18 22 13 " " .50 2.0392 .5684 " 7 3 11 14 14
" " .50 1.2851 .3582 " 4 2 4 5 5 " .15 .10 2.3891 1.8608 .09193 21 3 39 53 34
" .23 .10 1.4674 1.1429 .1120 11 2 14 17 10 " " .50 1.9032 .5305 " 6 3 10 12 12
" .25 .10 1.3900 1.0826 .1187 10 2 12 14 9 " .16 .10 2.2472 1.7503 .09568 19 3 34 45 29
.05 .16 .50 1.7902 .4990 " 6 2 8 10 10 .06 .15 .10 2.8422 2.2138 .09897 23 4 50 71 46
" .17 .10 2.1271 1.6568 .09938 17 3 30 39 25 " " .50 2.2642 .6312 " 7 3 12 16 16
" " .50 1.6946 .4724 " 5 2 7 9 9 " .16 .10 2.6437 2.0592 .1029 21 3 43 59 38
" .20 .10 1.8550 1.4449 .1103 14 3 21 27 17 " " .50 2.1061 .5871 " 6 3 11 13 13
" " .50 1.4778 .4119 " 4 2 5 6 6 " .17 .10 2.4791 1.9309 .1067 19 3 37 50 32
" .23 .10 1.6648 1.2967 .1210 11 2 16 20 13 " " .50 1.9749 .5505 " 6 3 9 11 11
" .25 .10 1.5659 1.2197 .1281 10 2 14 17 11 " .18 .10 2.3400 1.8226 .1106 17 3 32 43 28
" " .50 1.2475 .3477 " 3 2 3 4 4 " " .50 1.8642 .5196 " 5 3 8 10 10
" .317 .10 1.3278 1.0342 .1516 7 2 9 11 7 " .20 .10 2.1171 1.6490 .1181 14 3 25 34 21
" .325 .10 1.3058 1.0170 .1544 7 2 9 10 6 " " .50 1.6866 .4701 " 4 2 6 8 7
.051 .12 .10 3.1041 2.4177 .08107 30 4 71 101 66 " .22 .10 1.9452 1.5151 .1256 13 3 20 27 17
.052 .10 .10 4.0947 3.1893 .07361 44 5 131 192 128 " " .50 1.5497 .4320 " 4 2 5 6 6
" .11 .10 3.5580 2.7713 .07771 36 4 95 138 91 " .25 .10 1.7486 1.3620 .1366 10 3 16 20 13
.0575 .1875 .10 2.1725 1.6922 .1116 16 3 28 37 24 " " .50 1.3930 .3883 " 3 2 4 5 4
.058 .33 .10 1.3900 1.0827 .1639 7 2 9 11 7 " .30 .10 1.5179 1.1823 .1548 8 2 11 14 9
.06 .08 .10 9.3483 7.2813 .06956 105 11 675 1052 736 " " .50 1.2092 .3371 " 3 2 3 3 3
" " .50 7.4472 2.0759 " 30 9 167 239 257 .061 .14 .10 3.1463 2.4506 .09567 26 4 63 89 58
" .09 .10 6.6005 5.1411 .07407 70 8 324 495 341 .062 .12 .10 3.9912 3.1087 .08814 36 5 105 154 103
" " .50 5.2582 1.4658 " 20 6 80 112 119 " .13 .10 3.5436 2.7601 .09227 30 4 81 117 77
" .10 .10 5.2144 4.0614 .07845 52 6 195 293 199 .063 .2125 .10 2.0800 1.6201 .1251 13 3 23 31 20
" " .50 4.1540 1.1579 " 15 5 48 67 70 .065 .245 .10 1.8760 1.4612 .1388 11 3 18 23 15
" .11 .10 4.3741 3.4069 .08272 42 5 133 196 132 " .560 .10 .9942 .7744 .2953 3 2 4 4 3
" " .50 3.4846 .9713 " 12 4 33 45 46 .07 .09 .10 10.58533 8.2448 .07962 104 12 759 1191 838
" .12 .10 3.8076 2.9657 .08689 35 5 98 142 95 " " .50 8.4327 2.3506 " 30 10 188 271 293
(Continued)
667
TABLE T7.1 (Continued)
668

Statistical Research Group: Table of Sequential Sampling Plans

α = .05 α = .05

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

" " .50 3.0333 .8455 " 10 4 24 32 33 " .10 .10 7.4214 5.7805 .08419 69 9 361 556 386
" .13 .10 3.3981 2.6468 .09098 30 4 76 109 72 " " .50 5.9122 1.6480 " 20 7 89 126 135
" " .50 2.7071 .7546 " 9 3 19 25 25 " .11 .10 5.8280 4.5394 .08864 52 7 216 327 224
" .136 .10 3.2022 2.4942 .09340 27 4 66 94 62 " " .50 4.6428 1.2942 " 15 6 53 74 78
" .14 .10 3.0872 2.4046 .09500 26 4 61 86 56 " .12 .10 4.8638 3.7884 .09299 41 6 146 218 148
" " .50 2.4594 .6856 " 8 3 15 20 20 " " .50 3.8747 1.0801 " 12 5 36 50 52
.07 .13 .10 4.2150 3.2831 .09726 34 5 107 158 106 .08 .14 .10 4.6094 3.5903 .1076 34 6 115 172 117
" " .50 3.3579 .9360 " 10 4 26 36 37 " " .50 3.6721 1.0236 " 10 5 29 39 41
" .14 .10 3.7469 2.9185 .1014 29 5 82 120 80 " .15 .10 4.0839 3.1809 .1118 29 5 89 131 88
" " .50 2.9849 .8321 " 9 4 20 27 28 " " .50 3.2534 .9069 " 9 4 22 30 31
" .15 .10 3.3921 2.6421 .1056 25 4 66 95 63 " .16 .10 3.6861 2.8711 .1160 25 5 71 103 69
" " .50 2.7023 .7533 " 7 3 16 22 22 " " .50 2.9365 .8186 " 8 4 18 23 24
" .16 .10 3.1131 2.4248 .1096 23 4 54 77 51 " .17 .10 3.3738 2.6278 .1202 22 4 58 84 56
" " .50 2.4800 .6913 " 7 3 13 18 18 " " .50 2.6877 .7492 " 7 4 14 19 19
" .17 .10 2.8873 2.2489 .1136 20 4 46 64 42 " .18 .10 3.1214 2.4312 .1243 20 4 49 70 46
" " .50 2.3001 .6412 " 6 3 11 15 15 " " .50 2.4866 .6931 " 6 3 12 16 16
" .18 .10 2.7004 2.1033 .1176 18 4 39 55 36 " .19 .10 2.9127 2.2687 .1283 18 4 42 59 39
" " .50 2.1513 .5997 " 6 3 10 12 12 " " .50 2.3204 .6468 " 6 3 10 13 14
" .20 .10 2.4079 1.8755 .1254 15 3 30 41 27 " .20 .10 2.7370 2.1318 .1323 17 4 36 51 33
" " .50 1.9182 .5347 " 5 3 7 9 9 " " .50 2.1804 .6078 " 5 3 9 12 12
" .22 .10 2.1880 1.7042 .1331 13 3 24 32 21 " .22 .10 2.4564 1.9133 .1403 14 3 28 39 25
" " .50 1.7430 .4859 " 4 3 6 7 7 " " .50 1.9568 .5455 " 4 3 7 9 9
" .25 .10 1.9424 1.5129 .1446 11 3 18 24 15 " .25 .10 2.1510 1.6754 .1520 12 3 21 28 18
" " .50 1.5474 .4313 " 3 2 4 5 5 " " .50 1.7136 .4777 " 4 3 5 6 6
" .29 .10 1.7090 1.3311 .1596 9 3 13 17 11 " .30 .10 1.8121 1.4114 .1713 9 3 14 18 12
" .30 .10 1.6617 1.2943 .1633 8 2 12 16 10 " " .50 1.4436 .4024 " 3 2 3 4 4
" " .50 1.3238 .3690 " 3 2 3 4 3 .09 .12 .10 8.9985 7.0089 .1044 68 11 432 675 473
.075 .450 .10 1.2504 .9739 .2249 5 2 6 7 5 " " .50 7.1686 1.9983 " 20 9 107 153 165
(Continued)
Appendix
TABLE T7.1 (Continued)
Statistical Research Group: Table of Sequential Sampling Plans
α = .05 α = .05
Appendix

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

.08 .10 .10 11.7915 9.1844 .08966 103 13 842 1327 938 " .13 .10 7.0040 5.4553 .1089 51 8 255 394 273
" " .50 9.3936 2.6185 " 30 11 209 301 328 " " .50 5.5796 1.5553 " 15 7 63 89 95
" .11 .10 8.2205 6.4029 .09429 68 10 397 616 430 " .14 .10 5.7999 4.5175 .1134 40 7 171 261 179
" " .50 6.5488 1.8255 " 20 8 98 140 150 " " .50 4.6205 1.2880 " 12 6 42 59 63
" .12 .10 6.4242 5.0038 .09880 51 8 236 361 249 " .15 .10 4.9917 3.8880 .1178 34 6 124 187 127
" " .50 5.1178 1.4266 " 15 6 58 82 87 " " .50 3.9766 1.1085 " 10 5 31 42 45
" .13 .10 5.3388 4.1584 .1032 41 6 159 240 164 " .16 .10 4.4100 3.4350 .1221 29 6 95 141 96
" " .50 4.2531 1.1856 " 12 5 39 54 57 " " .50 3.5132 .9793 " 9 5 23 32 33
.09 .17 .10 3.9702 3.0924 .1264 25 5 75 111 75 .10 .22 .10 3.1027 2.4167 .1536 16 4 40 58 38
" " .50 3.1729 .8817 " 7 4 19 25 26 " " .50 2.4718 .6890 " 5 3 10 13 13
" .18 .10 3.6253 2.8237 .1306 22 5 62 90 60 " .25 .10 2.6309 2.0492 .1660 13 4 28 39 26
" " .50 2.8880 .8050 " 7 4 15 20 21 " " .50 2.0959 .5842 " 4 3 7 9 9
" .19 .10 3.3468 2.6068 .1348 20 4 52 75 50 " .27 .10 2.4034 1.8720 .1741 11 3 22 31 21
" " .50 2.6662 .7432 " 6 4 13 17 17 " " .50 1.9147 .5337 " 4 3 6 7 7
" .20 .10 3.1168 2.4277 .1389 18 4 44 63 42 " .30 .10 2.1411 1.6677 .1862 9 3 17 24 15
" " .50 2.4830 .6921 " 5 3 11 14 15 " " .50 1.7057 .4755 " 3 3 4 5 5
" .22 .10 2.7581 2.1482 .1471 15 4 33 47 31 .12 .15 .10 11.2104 8.7317 .1345 65 13 532 841 596
" " .50 2.1972 .6125 " 5 3 8 11 11 " " .50 8.9307 2.4895 " 19 11 132 191 208
" .25 .10 2.3789 1.8529 .1592 12 3 24 33 22 " .16 .10 8.6486 6.7363 .1392 49 11 311 486 342
" " .50 1.8951 .5283 " 4 3 6 7 8 " " .50 6.8898 1.9206 " 14 9 77 110 119
" .30 .10 1.9712 1.5353 .1789 9 3 15 21 13 " .17 .10 7.1051 5.5341 .1437 39 9 206 319 223
" " .50 1.5703 .4377 " 3 2 4 5 5 " " .50 5.6602 1.5778 " 11 7 51 73 78
.10 .13 .10 9.7560 7.5989 .1144 67 12 466 732 515 " .18 .10 6.0712 4.7288 .1483 32 8 148 227 158
" " .50 7.7720 2.1665 " 19 9 116 166 180 " " .50 4.8365 1.3482 " 10 6 37 52 55
" .14 .10 7.5677 5.8944 .1190 50 9 274 425 297 " .20 .10 4.7685 3.7142 .1572 24 6 88 134 92
" " .50 6.0287 1.6805 " 15 7 68 97 104 " " .50 3.7988 1.0589 " 7 5 22 30 32
" .15 .10 6.2478 4.8664 .1236 40 8 183 281 194 " .22 .10 3.9770 3.0977 .1660 19 5 60 89 61
" " .50 4.9772 1.3874 " 12 6 45 64 68 " " .50 3.1683 .8832 " 6 4 15 20 21

(Continued)
669
 670

TABLE T7.1 (Continued)

Statistical Research Group: Table of Sequential Sampling Plans
α = .05 α = .05

p1 p2 β h2 h1 s n0 n1 np1 ns np2 p1 p2 β h2 h1 s n0 n1 np1 ns np2

" .16 .10 5.3625 4.1768 .1280 33 7 132 201 138 " .25 .10 3.2337 2.5187 .1788 15 4 38 55 37
" " .50 4.2720 1.1908 " 10 5 33 46 48 " " .50 2.5761 .7181 " 5 4 9 13 13
" .17 .10 4.7259 3.6810 .1324 28 6 101 151 103 " .28 .10 2.7581 2.1482 .1915 12 4 27 38 26
" " .50 3.7649 1.0495 " 8 5 25 34 36 " " .50 2.1972 .6125 " 4 3 7 9 9
" .18 .10 4.2451 3.3065 .1367 25 5 80 119 81 " .30 .10 2.5241 1.9660 .1998 10 4 22 31 21
" " .50 3.3818 .9427 " 7 4 20 27 28 " " .50 2.0108 .5605 " 3 3 5 7 7
" .19 .10 3.8682 3.0129 .1410 22 5 65 96 65 .15 .19 .10 10.1562 7.9106 .1694 47 13 362 571 405
" " .50 3.0816 .8590 " 7 4 16 22 23 " " .50 8.0909 2.2554 " 14 10 90 130 141
" .20 .10 3.5643 2.7762 .1452 20 5 54 80 54 " .20 .10 8.2984 6.4635 .1741 38 11 238 373 263
" " .50 2.8394 .7915 " 6 4 13 18 19 " " .50 6.6108 1.8428 " 11 9 59 85 92
.15 .22 .10 6.1637 4.8009 .1833 27 8 128 198 138 .20 .24 .10 12.3724 9.6368 .2196 44 16 436 696 498
" " .50 4.9102 1.3687 " 8 7 32 45 48 " " .50 9.8563 2.7475 " 13 13 108 158 174
" .25 .10 4.5447 3.5398 .1968 18 6 67 102 70 " .25 .10 10.0471 7.8256 .2243 35 13 285 452 322
" " .50 3.6205 1.0092 " 6 5 17 23 25 " " .50 8.0039 2.2311 " 10 11 71 103 112
" .30 .10 3.2575 2.5372 .2188 12 5 33 48 33 " .30 .10 5.3625 4.1768 .2477 17 8 77 120 84
" " .50 2.5950 .7234 " 4 4 8 11 12 " " .50 4.2720 1.1908 " 5 6 19 27 29
" .35 .10 2.5910 2.0181 .2405 9 4 20 29 19 " .35 .10 3.7672 2.9342 .2706 11 6 37 56 39
" " .50 2.0641 .5754 " 3 3 5 7 7 " " .50 3.0011 .8366 " 4 5 9 13 14

Source: Statistical Research Group, Sequential Analysis of Statistical Data: Applications, AMP Report 30.2R, Columbia University, New York, 1945.
Note: Characteristic quantities of sequential tests for the binomial distribution computed for various combinations of p1, p2, α = .05, β = .10 and .50.
Appendix
Appendix 671

TABLE T7.2
Statistical Research Group: Table of Values of a and b for Sequential Sampling
1-b 1-a
a = log b = log
a b
α for computing a, β for computing b

.001 .01 .02 .03 .04 .05 .10 .15 .20 .30 .40

.001 3.000 2.000 1.699 1.522 1.398 1.301 1.000 .823 .699 .522 .398
.01 2.996 1.996 1.695 1.519 1.394 1.297 .996 .820 .695 .519 .394
β for computing a, α for

.02 2.991 1.991 1.690 1.514 1.389 1.292 .991 .815 .690 .514 .389
.03 2.987 1.987 1.686 1.510 1.385 1.288 .987 .811 .686 .510 .385
­computing b

.04 2.982 1.982 1.681 1.505 1.380 1.283 .982 .806 .681 .505 .380
.05 2.978 1.978 1.677 1.501 1.376 1.279 .978 .802 .677 .501 .376
.10 2.954 1.954 1.653 1.477 1.352 1.255 .954 .778 .653 .477 .352
.15 2.929 1.929 1.628 1.452 1.327 1.230 .929 .753 .628 .452 .327
.20 2.903 1.903 1.602 1.426 1.301 1.204 .903 .727 .602 .426 .301
.30 2.845 1.845 1.544 1.368 1.243 1.146 .845 .669 .544 .368 .243
.40 2.778 1.778 1.477 1.301 1.176 1.079 .778 .602 .477 .301 .176

Source: Statistical Research Group, Sequential Analysis of Statistical Data: Applications, AMP
Report 30.2R, Columbia University, New York, 1945.
Note: a and b in terms of α and β using common logarithms.
TABLE T7.3
672

g1 and g2 in Terms of p1 and p2 Using Common Logarithms for Sequential Sampling

æp ö æ 1 - p1 ö
g1 = log çç 2 ÷÷ g 2 = log çç ÷÷
p
è 1ø è 1 - p2 ø
g1 p2
g2 .005 .01 .02 .03 .04 .05 .06 .07 .08 .09 .10 .15 .20

.001 .6990 1.0000 1.3010 1.4771 1.6021 1.6990 1.7782 1.8451 1.9031 1.9542 2.0000 2.1761 2.3010
.0017 .0039 .0083 .0128 .0173 .0218 .0264 .0311 .0358 .0405 .0453 .0701 .0965
.005 .3010 .6021 .7782 .9031 1.0000 1.0792 1.1461 1.2041 1.2553 1.3010 1.4771 1.6021
.0022 .0066 .0111 .0156 .0201 .0247 .0293 .0340 .0388 .0436 .0684 .0947
.01 .3010 .4771 .6021 .6990 .7782 .8451 .9031 .9542 1.0000 1.1761 1.3010
.0044 .0089 .0134 .0179 .0225 .0272 .0318 .0366 .0414 .0662 .0925
.02 .1761 .3010 .3979 .4771 .5441 .6021 .6532 .6990 .8751 1.0000
.0045 .0090 .0135 .0181 .0227 .0274 .0322 .0370 .0618 .0881
.03 .1249 .2218 .3010 .3680 .4260 .4771 .5229 .6990 .8239
.0045 .0090 .0136 .0183 .0230 .0277 .0325 .0574 .0837
.04 .0969 .1761 .2430 .3010 .3522 .3979 .5740 .6990
.0045 .0091 .0138 .0185 .0232 .0280 .0529 .0792
p1 .05 g1 .0792 .1461 .2041 .2553 .3010 .4771 .6021
g2 .0046 .0092 .0139 .0187 .0235 .0483 .0746
.06 .0669 .1249 .1761 .2218 .3979 .5229
.0046 .0093 .0141 .0189 .0437 .0700
.07 .0580 .1091 .1549 .3310 .4559
.0047 .0094 .0142 .0391 .0654
.08 .0512 .0969 .2730 .3979
.0047 .0095 .0344 .0607
.09 .0458 .2218 .3468
.0048 .0296 .0560
.10 .1761 .3010
.0248 .0512
.15 .1249
.0263
Appendix
Appendix 673

1.00

0.80
Probability of accepting HO

0.60

0.40

0.20
n=
40 50

1
15 20
75

4
100

5
10

2
3
9
30

8
7
0
–1.00 –0.50 0.0 0.50 1.00 1.50 2.00 2.50 3.00
d

FIGURE F8.1
Operating characteristics of the one-sided normal test for a level of significance equal to 0.05. (Reprinted from
Bowker, A.H. and Lieberman, G.J., Engineering Statistics, Prentice–Hall, Inc., Englewood Cliffs, NJ, 1959, p. 118.
With permission.)

1.0

0.8
Probability of accepting HO

n=
0.6 1

0.4
2
3
4
6 7
8 10

5
15 20

0.2
30 40
50 75
100

0
0 1 2 3 4 5
d

FIGURE F8.2
Operating characteristics of the two-sided normal test for a level of significance equal to 0.05. (Reprinted from
Ferris, C.L. et al., Ann. Math. Stat., 17, 190, June 1946. With permission.)
674 Appendix

1.00

0.90

0.80
Probability of accepting HO

0.70

0.60

0.50

0.40
n = 100
n = 50 n = 7
n = 30 n = 4
0.30
n= n=
n = 20

n=
n=
4 3

n=
n=
0.20

5
15
10

7
0.10
0
0

0
–0.8 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2
d

FIGURE F8.3
Operating characteristics of the one-sided t-test for a level of significance equal to 0.05. (Reprinted from
Bowker, A.H. and Lieberman, G.J., Engineering Statistics, Prentice–Hall, Inc., Englewood Cliffs, NJ, 1959, p. 132.
With permission.)

1.0

n=2
0.8
Probability of accepting HO

0.6
4
5

0.4
7
10
15
30
50
100

20

0.2
40
75

0
0 1 2 3
d

FIGURE F8.4
Operating characteristics of the two-sided t-test for a level of significance equal to 0.05. (Reprinted from Ferris, C.L.
et al., Ann. Math. Stat., 17, 195, June 1946. With permission.)
Appendix 675

1.00

0.80
Probability of accepting HO

0.60 n=
2

0.40 3
4
5

6
0.20
7
8
10
100

40 50
75

15
20
30

0
0 1.0 2.0 3.0 4.0
λ

FIGURE F8.5
Operating characteristics of the one-sided (upper tail) chi-square test for a level of significance equal to 0.05.
(Reprinted from Ferris, C.L. et al., Ann. Math. Stat., 17, 181, June 1946. With permission.)
676 Appendix

TABLE T8.6
Factors for Acceptance Control Limits
Rejectable process level (RPL) – – – Factor A0 is used when σ′ is known
Acceptance control limit (ACL) α A0,βσ x–A1,βσ –A2,βR–A3,βS Factor A1 is used when
– –
Acceptable process level (APL) –
( )
2
β A0,ασ x–A1,ασ –A2,αR–A3,αS 1 å X-X
s= å is computed
m n
Factor A2 is used when

Nominal value 1
R= å R is computed
m
Factor A3 is used when
Acceptable process level (APL) – –
–
( )
2

Acceptance control limit (ACL) β A0,βσ x–A1,βσ –A2,βR–A3,βS 1 å X-X

s= å is computed
Rejectable process level (RPL) – – – m n-1
α A0,ασ x–A1,ασ –A2,αR–A3,αS
For n greater than 25,
4 ( n - 1) 4 ( n - 1)
c2 @ and c3 @ .
4n - 1 4n - 3

α (β) = 5% α (β) = 1%
n A0.05 A1.05 A2.05 A3.05 A0.01 A1.01 A2.01 A3.01

2 1.163 2.062 1.031 1.458 1.644 2.915 1.458 2.062

3 0.950 1.313 0.561 1.071 1.343 1.856 0.793 1.515
4 0.822 1.031 0.400 0.893 1.163 1.458 0.565 1.262
5 0.736 0.875 0.316 0.782 1.040 1.237 0.447 1.106

6 0.672 0.773 0.265 0.706 0.950 1.093 0.374 0.998

7 0.622 0.700 0.230 0.648 0.879 0.990 0.325 0.916
8 0.582 0.644 0.205 0.603 0.823 0.911 0.289 0.852
9 0.548 0.600 0.185 0.566 0.775 0.848 0.261 0.800
10 0.520 0.564 0.169 0.535 0.736 0.797 0.239 0.756

11 0.496 0.533 0.156 0.508 0.702 0.754 0.221 0.719

12 0.475 0.507 0.146 0.486 0.671 0.717 0.206 0.687
13 0.456 0.485 0.137 0.466 0.645 0.685 0.193 0.659
14 0.440 0.465 0.129 0.448 0.622 0.657 0.182 0.633
15 0.425 0.447 0.122 0.433 0.601 0.633 0.173 0.612

16 0.411 0.432 0.418 0.581 0.611 0.592

17 0.399 0.418 0.405 0.564 0.591 0.573
18 0.388 0.405 0.394 0.548 0.572 0.557
19 0.377 0.393 0.383 0.533 0.556 0.541
20 0.368 0.382 0.373 0.520 0.540 0.527

21 0.359 0.372 0.364 0.508 0.526 0.514

22 0.351 0.363 0.355 0.496 0.513 0.502
23 0.343 0.355 0.347 0.485 0.502 0.491
24 0.336 0.347 0.339 0.474 0.490 0.480
25 0.329 0.339 0.332 0.465 0.480 0.470
>25 1.645/ n 1.645/c2 n 1.645/c3 n 2.326/ n 2.326/c2 n 2.326/c3 n
(Continued)
Appendix 677

TABLE T8.6 (Continued)

Factors for Acceptance Control Limits
α (β) = 5% α (β) = 1%
n A0.05 A1.05 A2.05 A3.05 A0.01 A1.01 A2.01 A3.01

2 1.821 3.229 1.614 2.283 2.185 3.874 1.937 2.740

3 1.487 2.056 0.878 1.678 1.784 2.467 1.054 2.013
4 1.288 1.614 0.626 1.398 1.545 1.937 0.751 1.677
5 1.152 1.370 0.495 1.225 1.382 1.644 0.594 1.470

6 1.052 1.211 0.415 1.105 1.262 1.453 0.498 1.326

7 0.974 1.097 0.360 1.015 1.168 1.316 0.432 1.218
8 0.911 1.009 0.320 0.944 1.093 1.211 0.384 1.132
9 0.859 0.939 0.289 0.886 1.030 1.127 0.347 1.063
10 0.815 0.883 0.265 0.837 0.977 1.059 0.316 1.005

11 0.777 0.836 0.245 0.796 0.932 1.002 0.294 0.955

12 0.744 0.794 0.228 0.761 0.892 0.953 0.274 0.913
13 0.714 0.759 0.214 0.730 0.857 0.911 0.257 0.876
14 0.688 0.728 0.202 0.702 0.826 0.874 0.242 0.842
15 0.665 0.701 0.192 0.678 0.798 0.841 0.230 0.813

16 0.644 0.677 0.655 0.773 0.812 0.786

17 0.625 0.654 0.635 0.750 0.785 0.761
18 0.607 0.634 0.617 0.728 0.760 0.740
19 0.591 0.616 0.599 0.709 0.739 0.719
20 0.576 0.599 0.584 0.691 0.718 0.701

21 0.562 0.583 0.569 0.675 0.700 0.683

22 0.550 0.568 0.556 0.659 0.682 0.667
23 0.538 0.556 0.544 0.645 0.667 0.652
24 0.526 0.543 0.532 0.631 0.651 0.638
25 0.515 0.532 0.520 0.618 0.638 0.624
>25 2.576/ n 2.576/c2 n 2.576/c3 n 3.090/ n 3.090/c2 n 3.090/c3 n

Source: Reprinted from Freund, R.A., Ind. Qual. Control, 14, 18, 1957. With permission.
Note: If the acceptance control limits lie so close to the nominal value that two-tail probabilities must be used
(within ±2.5s¢/ n for a = 5%; ±3.0 s¢/ n for a = 1%; ±3.2s¢/ n for a = 0.5%; ±3.5s¢/ n for α = 0.1%), refer to
Table III for correction terms to be applied to the factors in Table II.
678 Appendix

TABLE T8.7
Correction Terms for Acceptance Control Factors
α = 5% α = 1% α = 0.5% α = 0.1%
Δ1 Δ2 C.T. Δ1 Δ2 C.T. Δ1 Δ2 C.F. Δ1 Δ2 C.F.

1.960 0.000 1.1916 2.576 0.000 1.1072 2.807 0.000 1.0898 3.291 0.000 1.0648
1.970 0.100 1.1366 2.589 0.100 1.0700 2.821 0.100 1.0563 3.307 0.100 1.0378
1.999 0.200 1.0935 2.625 0.200 1.0426 2.862 0.200 1.0333 3.352 0.200 1.0202
2.045 0.300 1.0610 2.685 0.300 1.0252 2.922 0.300 1.0179 3.421 0.300 1.0100
2.107 0.400 1.0377 2.757 0.400 1.0134 3.000 0.400 1.0094 3.492 0.400 1.0005
2.182 0.500 1.0223 2.842 0.500 1.0068 3.088 0.500 1.0046 3.500 0.409 1.0004
2.267 0.600 1.0132 2.933 0.600 1.0032 3.181 0.600 1.0021
2.356 0.700 1.0067 3.000 0.670 1.0018 3.200 0.619 1.0018
2.451 0.800 1.0034
2.500 0.851 1.0023

Source: Reprinted from Freund, R.A., Ind. Qual. Control, 14(4), 19, 1957. With permission.
Notes: When the acceptance control limits are too close to the nominal value (within ±2.5s¢/ n for α = 5%;
±3.0s¢/ n for α = 1%; ±3.2s¢/ n for α = 0.5%; ±3.5s¢/ n for α = 0.1%), corrections to the factors in
Table II are required since two-tail probabilities must replace the one-tail probabilities otherwise
applicable. The factors in Table II should be multiplied by the correction term (C.T.).*
Δ1 = Deviations of the acceptance control limit from the nominal value in terms of ±s¢/ n . To be
used when the APL values are to be determined from the acceptance control limits.
Δ2 = Deviations of the APL values from the nominal value in terms of ±s¢/ n . To be used when the
acceptance control limits are to be determined from the APL values.
t
C.T. = a1
t
where
α1 is the risk of an average from a process centered at the APL value falling outside the nearer
Acceptance Control Limit
α2 is the risk of that average falling outside the farther Acceptance Control Limit
α = α1 + α2
tα2 = 2Δ2 + tα1
TABLE T8.8
Boundary Values for Barnard’s Sequential t-Test
Appendix

α = .05 β = .05

D = .10 D = .25 D = .50 D = .75 D = 1.0 D = 1.5 D = 2.0 D = 3.0

k Y1 Y2 Y1 Y2 Y1 Y2 Y1 Y2 Y1 Y2 Y1 Y2 Y1 Y2 Y1 Y2

2 [–6.96] [–3.90] [2.60] [–2.14] [2.13] –0.47 [1.69] 0.37 [1.56] 0.95 [1.46]
4 [–3.13] [3.01] –1.49 [2.30] –0.53 [2.03] 0.51 1.84 1.03 1.82 1.50 1.85
6 –2.07 [2.73] –0.76 2.20 0.03 2.04 0.91 2.01 1.43 2.06 1.90 2.19
8 [–4.32] [4.24] –1.51 2.56 –0.35 2.16 0.37 2.09 1.23 2.18 1.74 2.29 2.22 2.47
10 [–3.67] [3.91] –1.15 2.46 –0.07 2.16 0.63 2.16 1.49 2.34 2.00 2.49 2.50 2.73

15 –2.72 3.39 –0.57 2.34 0.44 2.23 1.11 2.34 2.01 2.70 2.54 2.94 3.10 3.29
20 [–6.68] –2.17 3.10 –0.21 2.31 0.78 2.33 1.47 2.52 2.42 3.02 2.97 3.32
25 [–5.87] [6.00] –1.77 2.90 0.07 2.30 1.05 2.44 1.76 2.70 2.78 3.32 3.36 3.67
30 –5.27 [5.55] –1.50 2.77 0.29 2.32 1.28 2.55 2.02 2.88 3.09 3.59 3.71 3.99
35 –4.81 5.19 –1.28 2.67 0.48 2.36 1.49 2.66 2.24 3.05 3.38 3.84 4.03 4.29

40 –4.44 4.91 –1.09 2.60 0.65 2.40 1.67 2.76 2.45 3.21 3.64 4.07 4.32 4.57
45 –4.14 4.67 –0.93 2.55 0.79 2.44 1.84 2.87 2.64 3.36 3.89 4.29 4.60 4.83
50 –3.88 4.47 –0.79 2.51 0.92 2.49 1.99 2.97 2.82 3.50 4.12 4.50 4.86 5.08
60 –3.47 4.15 –0.56 2.44 1.16 2.58 2.27 3.17 3.16 3.77
70 –3.16 3.90 –0.37 2.41 1.36 2.68 2.52 3.35 3.45 4.03

80 –2.88 3.70 –0.20 2.39 1.54 2.78 2.76 3.53 3.73 4.27
90 –2.66 3.55 –0.06 2.39 1.71 2.88 2.97 3.71 3.99 4.49
100 –2.47 3.41 0.07 2.39 1.87 2.97 3.17 3.87 4.24 4.70
150 –1.80 2.99 0.57 2.46 2.51 3.42 4.00 4.59 5.27 5.65
200 –1.38 2.77 0.93 2.57 3.03 3.83 4.75 5.23 6.15 6.48

k1 29 12 6 4 3 2 2 2
k2 31 13 7 5 5 4 3 3
k1 600 100 30 20 10 <10 <10 <5
k2 600 100 30 20 10 <10 <10 <5
679
TABLE T10.1
680

d2* Factors and Degrees of Freedom ν for Estimating the Standard Deviation for the Average Range of k
Samples of n

R /d2* ® s

Size of Samples, n
No. of
Samples, k 2 3 4 5 6 7 8 9 10 11 12 13 14 15

d2* 1.41 1.91 2.24 2.48 2.67 2.83 2.96 3.08 3.18 3.27 3.35 3.42 3.49 3.55
1
ν 1.00 1.98 2.93 3.83 4.68 5.48 6.25 6.98 7.68 8.35 8.99 9.61 10.2 10.8
d2* 1.28 1.81 2.15 2.40 2.60 2.77 2.91 3.02 3.13 3.22 3.30 3.38 3.45 3.51
2
ν 1.92 3.83 5.69 7.47 9.16 10.8 12.3 13.8 15.1 16.5 17.8 19.0 20.2 21.3
d2* 1.23 1.77 2.12 2.38 2.58 2.75 2.89 3.01 3.11 3.21 3.29 3.37 3.44 3.50
3
ν 2.82 5.66 8.44 11.1 13.6 16.0 18.3 20.5 22.6 24.6 26.5 28.4 30.1 31.9
d2* 1.21 1.75 2.11 2.37 2.57 2.74 2.88 3.00 3.10 3.20 3.28 3.36 3.43 3.49
4
ν 3.71 7.49 11.2 14.7 18.1 21.3 24.4 27.3 30.1 32.7 35.3 37.7 40.1 42.4
d2* 1.19 1.74 2.10 2.36 2.56 2.73 2.87 2.99 3.10 3.19 3.28 3.35 3.42 3.49
5
ν 4.59 9.31 13.9 18.4 22.6 26.6 30.4 34.0 37.5 40.8 44.0 47.1 50.1 52.9
d2* 1.18 1.73 2.09 2.35 2.56 2.73 2.87 2.99 3.09 3.19 3.27 3.35 3.42 3.49
6
ν 5.47 11.1 16.7 22.0 27.0 31.8 36.4 40.8 45.0 49.0 52.8 56.5 60.1 63.5
d2* 1.17 1.73 2.09 2.35 2.55 2.72 2.86 2.99 3.09 3.19 3.27 3.35 3.42 3.48
7
ν 6.35 12.9 19.4 25.6 31.5 37.1 42.5 47.6 52.4 57.1 61.6 65.9 70.0 74.0
d2* 1.17 1.72 2.08 2.35 2.55 2.72 2.86 2.98 3.09 3.19 3.27 3.35 3.42 3.48
8
ν 7.23 14.8 22.1 29.2 36.0 42.4 48.5 54.3 59.9 65.2 70.3 75.2 80.0 84.6
d2* 1.16 1.72 2.08 2.34 2.55 2.72 2.86 2.98 3.09 3.18 3.27 3.35 3.42 3.48
9
ν 8.11 16.6 24.9 32.9 40.4 47.7 54.5 61.1 67.3 73.3 79.1 84.6 90.0 95.1
d2* 1.16 1.72 2.08 2.34 2.55 2.72 2.86 2.98 3.09 3.18 3.27 3.34 3.42 3.48
10
ν 8.99 18.4 27.6 36.5 44.9 52.9 60.6 67.8 74.8 81.5 87.8 94.0 99.9 106
Appendix

(Continued)
TABLE T10.1 (Continued)
Appendix

d2* Factors and Degrees of Freedom ν for Estimating the Standard Deviation for the Average Range of k
Samples of n

R /d2* ® s

Size of Samples, n
No. of
Samples, k 2 3 4 5 6 7 8 9 10 11 12 13 14 15

d2* 1.16 1.71 2.08 2.34 2.55 2.72 2.86 2.98 3.09 3.18 3.27 3.34 3.41 3.48
11
ν 9.87 20.2 30.4 40.1 49.4 58.2 66.6 74.6 82.2 89.6 96.6 103 110 116
d2* 1.15 1.71 2.07 2.34 2.55 2.72 2.86 2.98 3.09 3.18 3.27 3.34 3.41 3.48
12
ν 10.7 22.0 33.1 43.7 53.8 63.5 72.6 81.3 89.7 97.7 105 113 120 127
d2* 1.15 1.71 2.07 2.34 2.55 2.71 2.86 2.98 3.09 3.18 3.27 3.34 3.41 3.48
13
ν 11.6 23.8 35.8 47.3 58.3 68.7 78.6 88.1 97.2 106 114 122 130 137
d2* 1.15 1.71 2.07 2.34 2.54 2.71 2.86 2.98 3.08 3.18 3.27 3.34 3.41 3.48
14
ν 12.5 25.7 38.6 51.0 62.8 74.0 84.7 94.9 105 114 123 131 140 148
d2* 1.15 1.71 2.07 2.34 2.54 2.71 2.86 2.98 3.08 3.18 3.26 3.34 3.41 3.48
15
ν 13.4 27.5 41.3 54.6 67.2 79.3 90.7 102 112 122 132 141 150 158
d2* 1.14 1.70 2.07 2.33 2.54 2.71 2.85 2.98 3.08 3.18 3.26 3.34 3.41 3.48
20
ν 17.8 36.5 55.0 72.7 89.6 106 121 135 149 163 175 188 200 211
d2* 1.14 1.70 2.07 2.33 2.54 2.71 2.85 2.97 3.08 3.18 3.26 3.34 3.41 3.47
30
ν 26.5 54.7 82.4 109 134 158 181 203 224 244 263 281 299 316
d2* 1.13 1.70 2.06 2.33 2.54 2.71 2.85 2.97 3.08 3.17 3.26 3.34 3.41 3.47
50
ν 44.0 91.0 137 181 224 264 302 338 373 406 438 469 499 527
d2 1.13 1.69 2.06 2.33 2.53 2.70 2.85 2.97 3.08 3.17 3.26 3.34 3.41 3.47
c.d. 0.876 1.82 2.74 3.62 4.47 5.27 6.03 6.76 7.45 8.12 8.76 9.37 9.97 10.54
Source: Reprinted from Nelson, L.S., J. Qual. Technol., 7(1), 48, 1975. With permission.
681
682 Appendix

TABLE T10.2
Matched Single and Double, Known (σ) and Unknown (s) Standard
Deviation, Variables Sampling Plans for Values of p1 and p2 with α = .05,
β = .10 (n1 = n2, kt = kr)
Single Double
p1 p2 nσ ns k nσ ns ka kr ASNσ ASNs

.001 .0015 572 3180 3.02 422 2334 3.04 3.01 464.5 2568.5
.002 191 1032 2.97 138 739 3.01 2.95 154.9 829.1
.0025 107 567 2.93 75 391 3.00 2.90 87.5 455.6
.003 74 381 2.90 51 260 2.98 2.86 59.4 302.4
.004 45 226 2.84 33 163 2.92 2.80 36.8 181.2
.005 33 160 2.80 24 115 2.89 2.75 26.8 128.0
.006 26 124 2.77 18 84 2.90 2.70 20.9 97.8
.007 22 102 2.73 15 69 2.90 2.66 17.7 81.8
.008 19 87 2.71 14 64 2.79 2.65 15.1 69.4
.009 17 76 2.68 12 53 2.81 2.61 13.4 59.5
.01 15 67 2.66 10 44 2.88 2.57 12.0 53.1
.012 13 55 2.62 9 39 2.76 2.54 10.0 43.5
.015 11 44 2.57 7 29 2.84 2.46 8.4 35.1
.02 8 34 2.51 6 24 2.65 2.41 6.6 26.4
.025 7 27 2.46 5 19 2.62 2.35 5.5 21.0
.03 6 23 2.41 4 15 2.80 2.26 4.9 18.5
.035 6 20 2.37 4 15 2.53 2.26 4.3 16.3
.04 5 18 2.34 4 15 2.42 2.26 4.2 15.7
.05 5 15 2.28 3 10 2.51 2.14 3.3 11.2
.06 4 13 2.23 3 10 2.32 2.14 3.1 10.5
.0025 .004 357 1678 2.72 286 1337 2.74 2.71 308.1 1439.7
.005 161 736 2.68 111 501 2.74 2.65 132.1 596.0
.006 99 443 2.64 71 313 2.70 2.61 80.6 354.9
.0075 62 267 2.60 45 193 2.66 2.56 50.1 214.2
.01 38 157 2.54 27 111 2.62 2.49 30.1 123.5
.012 29 117 2.50 21 84 2.59 2.44 23.4 93.9
.015 22 85 2.45 15 58 2.58 2.38 17.1 66.4
.02 16 59 2.38 11 41 2.51 2.31 12.2 45.8
.025 12 45 2.33 9 32 2.45 2.25 9.9 35.2
.03 10 37 2.29 7 24 2.51 2.18 8.2 28.2
.035 9 31 2.25 6 20 2.51 2.13 7.1 23.8
.04 8 27 2.21 6 20 2.32 2.13 6.4 21.5
.05 7 22 2.15 5 16 2.26 2.07 5.3 17.1
.06 6 18 2.10 4 12 2.31 1.98 4.4 13.4
.005 .0075 417 1714 2.50 293 1195 2.53 2.48 341.2 1390.6
.01 138 547 2.44 100 391 2.48 2.41 112.0 437.1
.012 85 327 2.40 60 228 2.47 2.36 69.5 263.0
.015 53 196 2.35 39 144 2.41 2.31 43.0 157.6
.02 32 114 2.28 23 81 2.37 2.23 25.6 90.5
.025 23 79 2.23 16 54 2.37 2.16 18.5 62.7
(Continued)
Appendix 683

TABLE T10.2 (Continued)

Matched Single and Double, Known (σ) and Unknown (s) Standard
Deviation, Variables Sampling Plans for Values of p1 and p2 with α = .05,
β = .10 (n1 = n1, kt = kr)
Single Double
p1 p2 nσ ns k nσ ns ka kr ASNσ ASNs

.03 18 61 2.19 13 42 2.32 2.11 14.7 47.8

.035 15 49 2.15 11 35 2.25 2.08 12.0 38.3
.04 13 41 2.11 9 28 2.29 2.02 10.3 32.3
.05 10 31 2.05 7 21 2.25 1.95 8.0 24.2
.06 9 25 2.00 6 17 2.17 1.90 6.7 19.0
.07 8 21 1.96 5 14 2.19 1.84 5.7 16.1
.0075 .01 763 2909 2.37 511 1935 2.41 2.36 640.4 2423.7
.012 279 1040 2.33 213 787 2.36 2.32 233.1 860.2
.015 125 450 2.29 91 324 2.33 2.26 101.4 360.2
.02 60 208 2.22 43 146 2.30 2.18 49.2 166.0
.025 39 129 2.17 28 91 2.25 2.12 31.3 102.1
.03 29 92 2.12 20 63 2.25 2.06 23.2 73.3
.035 23 71 2.08 16 49 2.20 2.02 18.0 55.4
.04 19 58 2.05 14 42 2.14 1.99 15.2 45.9
.05 14 42 1.99 10 29 2.13 1.91 11.2 32.7
.06 12 33 1.94 8 22 2.10 1.85 9.0 24.9
.07 10 27 1.90 7 19 2.03 1.81 7.7 20.9
.08 9 23 1.86 6 16 2.01 1.76 6.6 17.7
.01 .015 351 1231 2.24 246 853 2.28 2.22 291.4 1009.5
.02 116 388 2.17 87 289 2.21 2.15 94.7 313.6
.025 64 208 2.12 45 143 2.21 2.08 52.6 166.2
.03 44 137 2.08 31 95 2.16 2.03 35.0 107.5
.035 33 100 2.04 23 69 2.15 1.98 26.5 79.6
.04 26 78 2.00 19 55 2.11 1.94 21.4 62.2
.045 22 64 1.97 15 43 2.13 1.90 17.6 50.7
.05 19 54 1.94 13 36 2.09 1.87 14.9 41.5
.06 15 41 1.89 11 30 1.98 1.83 11.8 32.4
.07 12 33 1.85 9 24 1.96 1.77 9.8 26.3
.08 11 27 1.81 7 18 2.05 1.70 8.3 21.4
.09 9 23 1.77 6 15 2.09 1.65 7.4 18.5
.10 8 20 1.74 6 15 1.87 1.65 6.5 16.3
.015 .02 633 2036 2.11 448 1427 2.14 2.09 544.3 1733.0
.025 195 603 2.05 142 435 2.09 2.03 159.3 487.4
.03 103 309 2.01 75 223 2.06 1.98 84.0 248.6
.035 67 197 1.97 47 135 2.05 1.93 54.3 155.4
.04 49 140 1.93 35 98 2.02 1.89 39.8 111.9
.045 39 107 1.90 27 74 2.01 1.85 31.2 85.7
.05 32 86 1.88 23 62 1.97 1.82 25.8 69.8
.06 23 61 1.82 17 44 1.91 1.77 18.6 48.3
.07 18 46 1.78 13 33 1.90 1.71 14.5 37.0
(Continued)
684 Appendix

TABLE T10.2 (Continued)

Matched Single and Double, Known (σ) and Unknown (s) Standard
Deviation, Variables Sampling Plans for Values of p1 and p2 with α = .05,
β = .10 (n1 = n1, kt = kr)
Single Double
p1 p2 nσ ns k nσ ns ka kr ASNσ ASNs

.08 15 37 1.74 10 24 1.96 1.65 12.0 29.0

.09 13 31 1.70 9 21 1.85 1.62 10.1 23.6
.10 11 26 1.67 8 18 1.82 1.58 8.9 20.2
.11 10 23 1.64 7 16 1.81 1.54 7.9 18.1
.12 9 20 1.61 6 13 1.89 1.49 7.2 15.6
.13 8 18 1.58 6 13 1.72 1.49 6.5 14.2
.14 8 16 1.56 5 11 1.84 1.43 5.9 13.1
.15 7 15 1.53 5 11 1.69 1.43 5.5 12.1
.02 .03 287 835 1.96 208 600 1.99 1.94 234.5 675.8
.035 147 416 1.92 102 285 1.98 1.89 120.2 334.8
.04 94 259 1.88 66 179 1.95 1.85 75.9 205.8
.045 67 182 1.85 46 122 1.95 1.81 54.8 144.5
.05 52 137 1.82 37 96 1.91 1.78 42.3 110.1
.06 35 89 1.77 25 62 1.87 1.72 28.3 70.4
.07 26 64 1.73 19 46 1.83 1.67 21.2 51.6
.08 21 50 1.69 15 35 1.81 1.62 16.9 39.6
.09 17 40 1.65 12 27 1.82 1.57 13.9 31.5
.10 15 34 1.62 10 22 1.82 1.53 11.8 26.1
.11 13 29 1.59 9 20 1.75 1.50 10.2 22.8
.12 12 25 1.56 8 17 1.72 1.47 9.0 19.2
.13 10 22 1.53 7 15 1.74 1.43 8.1 17.4
.15 9 18 1.48 6 12 1.66 1.38 6.7 13.5
.17 8 15 1.44 5 10 1.69 1.31 5.8 11.7
.20 6 12 1.37 4 8 1.70 1.23 4.8 9.6
.03 .04 506 1333 1.81 411 1077 1.82 1.80 434.6 1138.7
.045 250 643 1.78 185 472 1.81 1.76 206.6 526.6
.05 154 389 1.75 105 261 1.82 1.72 127.7 316.5
.06 81 197 1.70 56 134 1.78 1.66 65.8 156.5
.07 53 124 1.65 37 85 1.74 1.61 42.4 97.7
.08 38 88 1.61 28 63 1.69 1.57 31.0 69.9
.09 30 66 1.58 22 48 1.65 1.53 24.0 52.5
.10 24 53 1.54 17 36 1.67 1.48 19.4 41.3
.11 20 43 1.51 14 29 1.67 1.44 16.3 34.0
.12 18 37 1.48 13 27 1.58 1.42 14.2 29.6
.13 16 32 1.46 11 22 1.59 1.38 12.3 24.8
.15 13 24 1.41 9 17 1.52 1.33 9.8 18.7
.20 8 15 1.30 6 11 1.44 1.20 6.6 12.1
.25 6 11 1.20 4 7 1.61 1.05 5.0 8.7
.30 5 8 1.12 4 7 1.18 1.05 4.1 7.2
(Continued)
Appendix 685

TABLE T10.2 (Continued)

Matched Single and Double, Known (σ) and Unknown (s) Standard
Deviation, Variables Sampling Plans for Values of p1 and p2 with α = .05,
β = .10 (n1 = n1, kt = kr)
Single Double
p1 p2 nσ ns k nσ ns ka kr ASNσ ASNs

.04 .06 224 524 1.64 159 368 1.68 1.62 180.6 417.6
.07 114 258 1.60 83 186 1.64 1.57 91.8 205.0
.08 72 159 1.56 51 110 1.63 1.52 58.4 125.8
.09 51 110 1.52 37 78 1.59 1.48 41.2 87.2
.10 39 82 1.49 28 58 1.57 1.44 31.3 65.1
.11 32 65 1.46 22 44 1.57 1.40 25.3 50.7
.12 26 53 1.43 19 37 1.52 1.37 21.1 41.2
.13 22 44 1.40 17 33 1.47 1.35 18.3 35.6
.14 20 38 1.37 14 27 1.48 1.31 15.5 30.0
.15 17 33 1.35 12 22 1.50 1.27 13.7 25.3
.17 14 25 1.30 10 18 1.42 1.23 11.0 19.9
.20 11 19 1.24 8 14 1.35 1.16 8.7 15.2
.25 8 13 1.15 5 8 1.52 1.01 6.3 10.0
.30 6 9 1.06 4 6 1.35 0.92 4.7 7.0
.35 5 7 .98 4 6 1.04 0.92 4.1 6.2
.40 4 6 .91 3 4 1.05 0.80 3.2 4.3
.05 .07 300 660 1.55 204 443 1.60 1.53 246.7 535.4
.08 149 319 1.51 113 239 1.54 1.49 122.3 258.0
.09 93 194 1.47 66 135 1.54 1.44 75.8 154.4
.10 65 133 1.44 46 92 1.52 1.40 52.9 106.1
.11 49 98 1.41 36 70 1.48 1.37 40.0 78.1
.12 39 76 1.38 28 53 1.48 1.33 32.0 60.9
.13 32 62 1.35 23 43 1.45 1.30 25.9 48.6
.14 27 51 1.33 20 37 1.42 1.27 22.2 41.3
.15 24 43 1.30 17 31 1.41 1.24 19.0 34.8
.16 21 37 1.28 15 27 1.38 1.22 16.5 29.9
.17 18 33 1.26 13 23 1.40 1.18 14.8 26.4
.20 14 23 1.19 10 17 1.30 1.12 10.9 18.6
.25 10 15 1.10 7 11 1.21 1.02 7.5 11.9
.30 7 11 1.01 5 8 1.21 0.90 5.6 9.0
.35 6 8 .94 4 6 1.12 0.82 4.4 6.6
.40 5 7 .86 3 4 1.31 0.69 3.7 4.9

Source: Reprinted from Sommers, D.J., J. Qual. Technol., 13(1), 26, 1981. With
permission.
686 Appendix

TABLE T10.3
Comparison of Approximate and Exact Values of N and k for Variables
Sampling Plans
Exact Approximate
N k P2 N k True p1 True p2
(1) (2) (3) (4) (5) (6) (7)

5 0.5445 0.5428 5 0.5150 0.0541 0.5519

10 0.8037 0.3774 10 0.7856 0.0524 0.3830
15 0.9292 0.3033 15 0.9163 0.0517 0.3071
20 1.0083 0.2604 20 0.9983 0.0513 0.2633
25 1.0643 0.2332 25 1.0560 0.0510 0.2344
30 1.1069 0.2119 30 1.1000 0.0509 0.2137
35 1.1409 0.1966 35 1.1348 0.0507 0.1981
40 1.1688 0.1846 40 1.1633 0.0507 0.1859
45 1.1922 0.1748 45 1.1874 0.0506 0.1759
50 1.2125 0.1666 50 1.2082 0.0505 0.1676
65 1.2592 0.1489 64 1.2556 0.0501 0.1500
75 1.2828 0.1404 74 1.2797 0.0501 0.1413
100 1.3264 0.1256 99 1.3241 0.0501 0.1262

Source: Reprinted from Statistical Research Group, Techniques of Statistical

Analysis, McGraw-Hill, New York, 1947.
Notes: The exact values of k in column (2) have been computed from the non-
central t distribution, taking p1 = 0.05, α = 0.01, and N as shown in column
(1). The exact values of p2 for which β = 0.10 have then been computed
and entered in column (3). From these p2, taking β = 0.10, p1 = 0.05, and
α = 0.01, approximate values of N and k were computed from the Wallis
­approximation. For the approximation, the true values of p1 and p2 for
which α = 0.01 and β = 0.10 have been computed from the noncentral
t distribution.
Appendix 687

TABLE T10.4
Odeh-Owen Table 5: Two-Sided Sampling Plan Factors to Control Equal Tails
γ = 0.900
N P = 0.20 P = 0.10 P = 0.05 P = 0.025 P = 0.02 P = 0.01 P = 0.005

2 6.987 10.253 13.090 15.586 16.331 18.500 20.486

3 3.039 4.258 5.311 6.244 6.523 7.340 8.092
4 2.295 3.188 3.957 4.637 4.841 5.438 5.988
5 1.976 2.742 3.400 3.981 4.156 4.666 5.136
6 1.806 2.494 3.092 3.620 3.779 4.243 4.669
7 1.721 2.334 2.894 3.389 3.538 3.972 4.372
8 1.666 2.227 2.755 3.227 3.369 3.783 4.164
9 1.626 2.158 2.652 3.106 3.242 3.641 4.009
10 1.595 2.112 2.576 3.012 3.144 3.532 3.888
11 1.570 2.075 2.520 2.938 3.066 3.444 3.792
12 1.550 2.045 2.479 2.879 3.004 3.371 3.712
13 1.533 2.020 2.446 2.833 2.953 3.312 3.646
14 1.519 1.999 2.419 2.796 2.912 3.261 3.589
15 1.506 1.981 2.395 2.767 2.880 3.219 3.541
16 1.496 1.965 2.374 2.742 2.853 3.184 3.499
17 1.486 1.950 2.356 2.720 2.830 3.155 3.463
18 1.478 1.938 2.340 2.701 2.810 3.130 3.433
19 1.470 1.927 2.325 2.683 2.791 3.109 3.406
20 1.463 1.916 2.312 2.667 2.775 3.090 3.383
21 1.457 1.907 2.300 2.653 2.760 3.073 3.364
22 1.451 1.899 2.290 2640 2.746 3.057 3.346
23 1.446 1.891 2.280 2.628 2.733 3.043 3.330
24 1.441 1.884 2.270 2.617 2.721 3.029 3.315
25 1.437 1.877 2.262 2.606 2.711 3.017 3.301
30 1.419 1.851 2.227 2.565 2.667 2.967 3.245
35 1.406 1.831 2.202 2.534 2.634 2.929 3.203
40 1.396 1.816 2.182 2.510 2.609 2.901 3.171
45 1.387 1.804 2.166 2.491 2.589 2.878 3.146
50 1.381 1.794 2.154 2.476 2.573 2.859 3.124
60 1.370 1.778 2.133 2.451 2.547 2.829 3.091
70 1.362 1.766 2.118 2.433 2.528 2.807 3.066
80 1.356 1.757 2.106 2.418 2.513 2.790 3.047
90 1.351 1.750 2.097 2.407 2.500 2.776 3.031
100 1.347 1.744 2.089 2.397 2.490 2.764 3.018
120 1.341 1.734 2.076 2.382 2.474 2.746 2.997
150 1.334 1.723 2.062 2.365 2.457 2.726 2.975
300 1.317 1.699 2.030 2.326 2.416 2.678 2.922
500 1.309 1.686 2.013 2.306 2.394 2.654 2.895
600 1.306 1.682 2.008 2.300 2.388 2.647 2.887
700 1.304 1.679 2.005 2.295 2.383 2.641 2.880
800 1.303 1.677 2.002 2.292 2.379 2.637 2.875
900 1.301 1.675 1.999 2.289 2.376 2.633 2.871
(Continued)
688 Appendix

TABLE T10.4 (Continued)

Odeh-Owen Table 5: Two-Sided Sampling Plan Factors to Control Equal Tails
γ = 0.900
N P = 0.20 P = 0.10 P = 0.05 P = 0.025 P = 0.02 P = 0.01 P = 0.005

1000 1.300 1.673 1.997 2.286 2.374 2.630 2.868

1500 1.297 1.668 1.990 2.278 2.365 2.620 2.856
2000 1.295 1.665 1.986 2.273 2.359 2.614 2.850
3000 1.292 1.661 1.981 2.267 2.353 2.607 2.842
5000 1.290 1.657 1.976 2.261 2.347 2.600 2.834
10000 1.287 1.654 1.971 2.255 2.341 2.593 2.826
∞ 1.282 1.645 1.960 2.241 2.326 2.576 2.807
Source: Reprinted from Odeh, R.E. and Owen, D.B., Tables for Normal Tolerance Limits,
Sample Plans, and Screening, Marcel-Dekker, Inc., 1980, p. 146.
Appendix 689

TABLE T10.5
Odeh-Owen Table 6: Two-Sided Sampling Plan Factors to Control
Tails Separately
γ = 0.900
N P = 0.20 P = 0.10 P = 0.05 P = 0.025 P = 0.02 P = 0.01 P = 0.005

2 6.987 10.253 13.090 15.586 16.331 18.500 20.486

3 3.039 4.258 5.311 6.244 6.523 7.340 8.092
4 2.295 3.188 3.957 4.637 4.841 5.438 5.988
5 1.976 2.742 3.400 3.981 4.156 4.666 5.136
6 1.795 2.494 3.092 3.620 3.779 4.243 4.669
7 1.676 2.333 2.894 3.389 3.538 3.972 4.372
8 1.590 2.219 2.754 3.227 3.369 3.783 4.164
9 1.525 2.133 2.650 3.106 3.242 3.641 4.009
10 1.474 2.066 2.568 3.011 3.144 3.532 3.888
11 1.433 2.011 2.503 2.935 3.065 3.443 3.792
12 1.398 1.966 2.448 2.872 3.000 3.371 3.712
13 1.368 1.928 2.402 2.820 2.945 3.309 3.645
14 1.343 1.895 2.363 2.774 2.898 3.257 3.588
15 1.321 1.867 2.329 2.735 2.857 3.212 3.538
16 1.301 1.842 2.299 2.701 2.821 3.172 3.495
17 1.284 1.819 2.272 2.670 2.789 3.137 3.456
18 1.268 1.800 2.249 2.643 2.761 3.105 3.422
19 1.254 1.782 2.227 2.618 2.736 3.077 3.391
20 1.241 1.765 2.208 2.596 2.712 3.052 3.363
21 1.229 1.750 2.190 2.576 2.691 3.028 3.338
22 1.218 1.737 2.174 2.557 2.672 3.007 3.315
23 1.208 1.724 2.159 2.540 2.654 2.987 3.293
24 1.199 1.712 2.145 2.525 2.638 2.969 3.273
25 1.190 1.702 2.132 2.510 2.623 2.952 3.255
30 1.154 1.657 2.080 2.450 2.561 2.884 3.180
35 1.127 1.624 2.041 2.406 2.515 2.833 3.125
40 1.106 1.598 2.010 2.371 2.479 2.793 3.082
45 1.089 1.577 1.986 2.343 2.450 2.761 3.047
50 1.075 1.559 1.965 2.320 2.426 2.735 3.018
60 1.052 1.532 1.933 2.284 2.389 2.694 2.973
70 1.035 1.511 1.909 2.256 2.360 2.662 2.940
80 1.022 1.495 1.890 2.235 2.338 2.638 2.913
90 1.011 1.481 1.874 2.217 2.320 2.618 2.891
100 1.001 1.470 1.861 2.203 2.304 2.601 2.873
120 0.986 1.452 1.841 2.179 2.280 2.574 2.844
150 0.970 1.433 1.818 2.154 2.254 2.546 2.813
300 0.931 1.386 1.765 2.094 2.192 2.477 2.739
500 0.910 1.362 1.736 2.062 2.159 2.442 2.701
600 0.904 1.355 1.728 2.053 2.150 2.431 2.689
700 0.899 1.349 1.722 2.046 2.142 2.423 2.680
800 0.896 1.344 1.717 2.040 2.136 2.417 2.673
(Continued)
690 Appendix

TABLE T10.5 (Continued)

Odeh-Owen Table 6: Two-Sided Sampling Plan Factors to Control
Tails Separately
γ = 0.900
N P = 0.20 P = 0.10 P = 0.05 P = 0.025 P = 0.02 P = 0.01 P = 0.005

900 0.892 1.341 1.712 2.035 2.132 2.411 2.668

1,000 0.890 1.338 1.709 2.031 2.127 2.407 2.663
1,500 0.881 1.327 1.697 2.018 2.114 2.392 2.646
2,000 0.875 1.321 1.690 2.010 2.105 2.383 2.637
3,000 0.869 1.314 1.681 2.001 2.096 2.372 2.625
5,000 0.863 1.306 1.673 1.991 2.086 2.362 2.614
10,000 0.857 1.299 1.665 1.982 2.077 2.351 2.603
∞ 0.842 1.282 1.645 1.960 2.054 2.326 2.576

Source: Reprinted from Odeh, R.E. and Owen, D.B., Tables for Normal Tolerance Limits,
Sample Plans, and Screening, Marcel-Dekker, 1980, p. 147.
 Appendix

TABLE T11.1
MIL-STD-105E Table VIII—Limit Numbers for Reduced Inspection
Number of Acceptable Quality Level
Sample Units
from Last
10 Lots or
Batches 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

20–29 * * * * * * * * * * * * * * * 0 0 2 4 8 14 22 40 68 115 181

30–49 * * * * * * * * * * * * * * 0 0 1 3 7 13 22 36 63 105 178 277
50–79 * * * * * * * * * * * * * 0 0 2 3 7 14 25 40 63 110 181 301
80–129 * * * * * * * * * * * * 0 0 2 4 7 14 24 42 68 105 181 297
130–199 * * * * * * * * * * * 0 0 2 4 7 13 25 42 72 115 177 301 490
200–319 * * * * * * * * * * 0 0 2 4 8 14 22 40 68 115 181 277 471
320–499 * * * * * * * * * 0 0 1 4 8 14 24 39 68 113 189
500–799 * * * * * * * * 0 0 2 3 7 14 25 40 63 110 181
800–1,249 * * * * * * * 0 0 2 4 7 14 24 42 68 105 181
1,250–1,999 * * * * * * 0 0 2 4 7 13 24 40 69 110 169
2,000–3,149 * * * * * 0 0 2 4 8 14 22 40 68 115 181
3,150–4,999 * * * * 0 0 1 4 8 14 24 38 67 111 186
5,000–7,999 * * * 0 0 2 3 7 14 25 40 63 110 181
8,000–12,499 * * 0 0 2 4 7 14 24 42 68 105 181
12,500–19,999 * 0 0 2 4 7 13 24 40 69 110 169
20,000–31,499 0 0 2 4 8 14 22 40 68 115 181
31,500 & Over 0 1 4 8 14 24 38 67 111 186

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 32.
691
692 Appendix

TABLE T11.2
MIL-STD-105E Table I—Sample Size Code Letters
Special Inspection Levels General Inspection Levels
Lot or Batch Size S-1 S-2 S-3 S-4 I II III
2–8 A A A A A A B
9–15 A A A A A B C
16–25 A A B B B C D
26–50 A B B C C D E
51–90 B B C C C E F
91–150 B B C D D F G
151–280 B C D E E G H
281–500 B C D E F H J
501–1,200 C C E F G J K
1,201–3,200 C D E G H K L
3,201–10,000 C D F G J L M
10,001–35,000 C D F H K M N
35,001–150,000 D E G J L N P
150,001–500,000 D E G J M P Q
500,001 and over D E H K N Q R

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection
by Attributes, MIL-STD-105E, U.S. Government Printing Office, Washington, DC, 1989, p. 13.
 Appendix

TABLE T11.3
MIL-STD-105E Table II-A—Single-Sampling Plans for Normal Inspection (Master Table)
Acceptable Quality Levels (Normal Inspection)
Sample
Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample
Letter Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31
B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31 44 45
C 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31 44 45 ⇧
D 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31 44 45 ⇧ ⇧
E 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31 44 45 ⇧ ⇧ ⇧
F 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H 50 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K 125 ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L 200 ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M 315 ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N 500 ⇩ ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P 800 ⇩ 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q 1250 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R 2000 ⇧ ⇧ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 14.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
693
 694

TABLE T11.4
MIL-STD-105E Table II-B—Single-Sampling Plans for Tightened Inspection (Master Table)
Acceptable Quality Levels (Tightened Inspection)
Sample
Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample
Letter Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 27 28

B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 27 28 41 42
C 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 27 28 41 42 ⇧
D 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 27 28 41 42 ⇧ ⇧
E 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 27 28 41 42 ⇧ ⇧ ⇧
F 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H 50 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K 125 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L 200 ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M 315 ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N 500 ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P 800 ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q 1250 ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R 2000 0 1 ⇧ ⇩ 1 2 2 3 3 4 5 6 8 9 12 13 18 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
S 3150 ⇧ ⇧ 1 2 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 15.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
Appendix
 Appendix

TABLE T11.5
MIL-STD-105E Table II-C—Single-Sampling Plans for Reduced Inspection (Master Table)
Acceptable Quality Levels (Reduced Inspection)a
Sample
Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample
Letter Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇩ ⇩ 1 2 2 3 3 4 5 6 7 8 10 11 14 15 21 22 30 31
B 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 2 4 3 5 5 6 7 8 10 11 14 15 21 22 30 31
C 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 14 17 21 24 ⇧
D 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 14 17 21 24 ⇧ ⇧
E 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 14 17 21 24 ⇧ ⇧ ⇧
F 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K 50 ⇩ ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L 80 ⇩ ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M 125 ⇩ ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N 200 ⇩ ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P 315 ⇩ 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q 500 0 1 ⇧ ⇩ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R 800 ⇧ ⇧ 0 2 1 3 1 4 2 5 3 6 5 8 7 10 10 13 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 16.
a If the acceptance number has been exceeded, but the rejection number has not been reached, accept the lot, but reinstate normal inspection.

⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
695
TABLE T11.6
696

MIL-STD-105E Table III-A—Double-Sampling Plans for Normal Inspection (Master Table)

Acceptable Quality Levels (Normal Inspection)

Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Sample Size Sample Sample
Code Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * *
B First 2 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 17 22 25 31
Second 2 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 37 38 56 57
C First 3 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 17 22 25 31 ⇧
Second 3 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 37 38 56 57 ⇧
D First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 17 22 25 31 ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 37 38 56 57 ⇧ ⇧
E First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 17 22 25 31 ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 37 38 56 57 ⇧ ⇧ ⇧
F First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 50 50 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K First 80 80 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 80 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 125 125 ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M First 200 200 ⇩ ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 315 315 ⇩ ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇩ ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P First 500 500 ⇩ * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇩ ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 800 800 * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 800 1600 ⇧ ⇩ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R First 1250 1250 ⇧ ⇧ 0 2 0 3 1 4 2 5 3 7 5 9 7 11 11 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 1250 2500 ⇧ ⇧ 1 2 3 4 4 5 6 7 8 9 12 13 18 19 26 27 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing Office, Washington, DC, 1989, p. 17.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
Appendix

* = Use corresponding single-sampling plan (or alternatively use double-sampling plan below-where available).
TABLE T11.7
MIL-STD-105E Table III-B—Double-Sampling Plans for Tightened Inspection (Master Table)
Acceptable Quality Levels (Tightened Inspection)
Sample
Appendix

Size Cumu­lative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * * * * * * * *
B First 2 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 15 20 23 29
Second 2 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 34 35 52 53
C First 3 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 15 20 23 29 ⇧
Second 3 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 34 35 52 53 ⇧
D First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 15 20 23 29 ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 34 35 52 53 ⇧ ⇧
E First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 15 20 23 29 ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 34 35 52 53 ⇧ ⇧ ⇧
F First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 50 50 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K First 80 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 80 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 125 125 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M First 200 200 ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 315 315 ⇩ ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇩ ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P First 500 500 ⇩ ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇩ ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 800 800 ⇩ * ⇩ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 800 1600 ⇩ ⇩ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R First 1250 1250 * ⇧ ⇩ 0 2 0 3 1 4 2 5 3 7 6 10 9 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 1250 2500 ⇧ ⇩ 1 2 3 4 4 5 6 7 11 12 15 16 23 24 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
S First 2000 2000 0 2
Second 2000 4000 1 2

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing Office, Washington, DC, 1989, p. 18.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
* = Use corresponding single-sampling plan (or alternatively use double-sampling plan below, where available).
697
TABLE T11.8
698

MIL-STD-105E Table III-C—Double-Sampling Plans for Reduced Inspection (Master Table)

Acceptable Quality Levels (Reduced Inspection)
Sample Cumu­
Size lative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * *
B ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * * *
C ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * * * ⇧
D First 2 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 7 12 11 17 ⇧ ⇧
Second 2 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 18 22 26 30 ⇧ ⇧
E First 3 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 7 12 11 17 ⇧ ⇧ ⇧
Second 3 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 18 22 26 30 ⇧ ⇧ ⇧
F First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 45 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 50 50 ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 0 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 0 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M First 80 80 ⇩ ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 80 160 ⇩ ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 125 125 ⇩ ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P First 200 200 ⇩ * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 315 315 * ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R First 500 500 ⇩ ⇩ 02 03 04 04 15 27 38 5 10 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇩ ⇩ 02 04 15 36 47 69 8 12 12 16 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source:  United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing Office, Washington, DC, 1989, 19.
a If, after the second sample, the acceptance number has been exceeded, but the rejection number has not been reached, accept the lot, but reinstate normal inspection.

⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
* = Use corresponding single-sampling plan (or alternatively use double-sampling plan below, where available).
Appendix
TABLE T11.9
MIL-STD-105E Table IV-A—Multiple-Sampling Plans for Normal Inspection (Master Table)
Acceptable Quality Levels (Normal Inspection)a
Appendix

Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * *
B ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++ ++
C ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++ ++ ⇧
D First 2 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 4 12 6 16 ⇧ ⇧
Second 2 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 11 19 17 27 ⇧ ⇧
Third 2 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 19 27 29 39 ⇧ ⇧
Fourth 2 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 27 34 40 49 ⇧ ⇧
Fifth 2 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 36 40 53 58 ⇧ ⇧
Sixth 2 12 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 45 47 65 68 ⇧ ⇧
Seventh 2 14 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 53 54 77 78 ⇧ ⇧
E First 3 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 4 12 6 16 ⇧ ⇧ ⇧
Second 3 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 11 19 17 27 ⇧ ⇧ ⇧
Third 3 9 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 19 27 29 39 ⇧ ⇧ ⇧
Fourth 3 12 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 27 34 40 49 ⇧ ⇧ ⇧
Fifth 3 15 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 36 40 53 58 ⇧ ⇧ ⇧
Sixth 3 18 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 45 47 65 68 ⇧ ⇧ ⇧
Seventh 3 21 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 53 54 77 78 ⇧ ⇧ ⇧
F First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 5 15 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 5 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 5 25 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 5 30 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 5 35 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 8 24 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 8 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 8 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 8 48 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 8 56 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
699
 700

TABLE T11.9 (Continued)

MIL-STD-105E Table IV-A—Multiple-Sampling Plans for Normal Inspection (Master Table)
Acceptable Quality Levels (Normal Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

H First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 13 39 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 13 52 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 13 65 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 13 78 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 13 91 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 20 60 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 20 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 20 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 20 120 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 20 140 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 32 96 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 32 128 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 32 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 32 192 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 32 224 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 50 50 ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 50 150 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 50 200 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 50 250 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 50 300 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 50 350 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
Appendix
TABLE T11.9 (Continued)
Appendix

MIL-STD-105E Table IV-A—Multiple-Sampling Plans for Normal Inspection (Master Table)

Acceptable Quality Levels (Normal Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

M First 80 80 ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Second 80 160 ⇩ ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 80 240 ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 80 320 ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 80 400 ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 80 480 ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 80 560 ⇩ ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 125 125 ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 125 375 ⇩ ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 125 500 ⇩ ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 125 625 ⇩ ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 125 750 ⇩ ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 125 875 ⇩ ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P First 200 200 ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 200 600 ⇩ ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 200 800 ⇩ ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 200 1000 ⇩ ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 200 1200 ⇩ ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 200 1400 ⇩ ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 315 315 * ⇧ ⇩ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇧ ⇩ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 315 945 ⇧ ⇩ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 315 1260 ⇧ ⇩ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 315 1575 ⇧ ⇩ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 315 1890 ⇧ ⇩ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 315 2205 ⇧ ⇩ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
701
 702

TABLE T11.9 (Continued)

MIL-STD-105E Table IV-A—Multiple-Sampling Plans for Normal Inspection (Master Table)
Acceptable Quality Levels (Normal Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

R First 500 500 ⇧ ⇧ # 2 # 2 # 3 # 4 0 4 0 5 1 7 2 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇧ ⇧ # 2 0 3 0 3 1 5 1 6 3 8 4 10 7 14 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 500 1500 ⇧ ⇧ 0 2 0 3 1 4 2 6 3 8 6 10 8 13 13 19 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 500 2000 ⇧ ⇧ 0 3 1 4 2 5 3 7 5 10 8 13 12 17 19 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 500 2500 ⇧ ⇧ 1 3 2 4 3 6 5 8 7 11 11 15 17 20 25 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 500 3000 ⇧ ⇧ 1 3 3 5 4 6 7 9 10 12 14 17 21 23 31 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 500 3500 ⇧ ⇧ 2 3 4 5 6 7 9 10 13 14 18 19 25 26 37 38 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, pp. 20–21.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
* = Use corresponding single-sampling plan (or alternatively use multiple-sampling plan below, where available).
++ = Use corresponding double-sampling plan (or alternatively use multiple-sampling plan below, where available).
# = Acceptance not permitted at this sample size.
Appendix
TABLE T11.10
MIL-STD-105E Table IV-B—Multiple-Sampling Plans for Tightened Inspection (Master Table)
Acceptable Quality Levels (Tightened Inspection)a
Appendix

Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * * * * * * * *
B ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++
C ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++ ⇧
D First 2 2 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 3 10 6 15 ⇧ ⇧
Second 2 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 10 17 16 25 ⇧ ⇧
Third 2 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 17 24 26 36 ⇧ ⇧
Fourth 2 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 24 31 37 46 ⇧ ⇧
Fifth 2 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 32 37 49 55 ⇧ ⇧
Sixth 2 12 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 40 43 61 64 ⇧ ⇧
Seventh 2 14 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 48 49 72 73 ⇧ ⇧
E First 3 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 3 10 6 15 ⇧ ⇧ ⇧
Second 3 6 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 10 17 16 25 ⇧ ⇧ ⇧
Third 3 9 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 17 24 26 36 ⇧ ⇧ ⇧
Fourth 3 12 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 24 31 37 46 ⇧ ⇧ ⇧
Fifth 3 15 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 32 37 49 55 ⇧ ⇧ ⇧
Sixth 3 18 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 40 43 61 64 ⇧ ⇧ ⇧
Seventh 3 21 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 48 49 72 73 ⇧ ⇧ ⇧
F First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 5 15 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 5 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 5 25 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 5 30 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 5 35 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 8 24 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 8 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 8 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 8 48 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 8 56 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
703
 704

TABLE T11.10 (Continued)

MIL-STD-105E Table IV-B—Multiple-Sampling Plans for Tightened Inspection (Master Table)
Acceptable Quality Levels (Tightened Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

H First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 13 39 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 13 52 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 13 65 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 13 78 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 13 91 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 20 60 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 20 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 20 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 20 120 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 20 140 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
K First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 32 96 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 32 128 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 32 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 32 192 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 32 224 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 50 50 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 50 150 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 50 200 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 50 250 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 50 300 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 50 350 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
Appendix
TABLE T11.10 (Continued)
Appendix

MIL-STD-105E Table IV-B—Multiple-Sampling Plans for Tightened Inspection (Master Table)

Acceptable Quality Levels (Tightened Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

M First 80 80 ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Second 80 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 80 240 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 80 320 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 80 400 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 80 480 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 80 560 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 125 125 ⇩ ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 125 375 ⇩ ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 125 500 ⇩ ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 125 625 ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 125 750 ⇩ ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 125 875 ⇩ ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
P First 200 200 ⇩ ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 200 600 ⇩ ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 200 800 ⇩ ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 200 1000 ⇩ ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 200 1200 ⇩ ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 200 1400 ⇩ ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 315 315 ⇩ * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇩ ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 315 945 ⇩ ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 315 1260 ⇩ ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 315 1575 ⇩ ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 315 1890 ⇩ ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 315 2205 ⇩ ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
705
 706

TABLE T11.10 (Continued)

MIL-STD-105E Table IV-B—Multiple-Sampling Plans for Tightened Inspection (Master Table)
Acceptable Quality Levels (Tightened Inspection)a
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

R First 500 500 * ⇩ ⇩ # 2 # 2 # 3 # 4 0 4 0 6 1 8 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇩ ⇩ # 2 0 3 0 3 1 5 2 7 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 500 1500 ⇩ ⇩ 0 2 0 3 1 4 2 6 4 9 7 12 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 500 2000 ⇩ ⇩ 0 3 1 4 2 5 3 7 6 11 10 15 16 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 500 2500 ⇩ ⇩ 1 3 2 4 3 6 5 8 9 12 14 17 22 25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 500 3000 ⇩ ⇩ 1 3 3 5 4 6 7 9 12 14 18 20 27 29 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 500 3500 ⇩ ⇩ 2 3 4 5 6 7 9 10 14 15 21 22 32 33 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
S First 800 800 # 2
Second 800 1600 # 2
Third 800 2400 0 2
Fourth 800 3200 0 3
Fifth 800 4000 1 3
Sixth 800 4800 1 3
Seventh 800 5600 2 3

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, pp. 22–23.
⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
* = Use corresponding single-sampling plan (or alternatively use multiple-sampling plan below, where available).
++ = Use corresponding double-sampling plan (or alternatively use multiple-sampling plan below, where available).
# = Acceptance not permitted at this sample size.
Appendix
TABLE T11.11
MIL-STD-105E Table IV-C—Multiple-Sampling Plans for Reduced Inspection (Master Table)
Acceptable Quality Levels (Reduced Inspection)
Appendix

Sample
Size Cumulative0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

A ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇩ ⇩ * * * * * * * * *
B ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ * * * * * * * * * *
C ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ * * * * * * * * * * ⇧
D ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++ ++ ⇧ ⇧
E ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ ++ ++ ++ ++ ++ ++ ++ ++ ++ ++ ⇧ ⇧ ⇧
F First 5 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 5 10 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 5 15 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 5 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 5 25 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 5 30 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 5 35 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
G First 8 8 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 8 16 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 8 24 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 8 32 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 8 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 8 48 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 8 56 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
H First 13 13 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 13 26 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 13 39 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 13 52 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 13 65 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 13 78 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 13 91 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
J First 20 20 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 20 40 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 20 60 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 20 80 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 20 100 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 20 120 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 20 140 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
707

(Continued)
 708

TABLE T11.11 (Continued)

MIL-STD-105E Table IV-C—Multiple-Sampling Plans for Reduced Inspection (Master Table)
Acceptable Quality Levels (Reduced Inspection)
Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

K First 32 32 ⇩ ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Second 32 64 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 32 96 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 32 128 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 32 160 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 32 192 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 32 224 ⇩ ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
L First 50 50 ⇩ ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 50 100 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 50 150 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 50 200 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 50 250 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 50 300 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 50 350 ⇩ ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
M First 80 80 ⇩ ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 80 160 ⇩ ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 80 240 ⇩ ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 80 320 ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 80 400 ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 80 480 ⇩ ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 80 560 ⇩ ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
N First 125 125 ⇩ ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 125 250 ⇩ ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 125 375 ⇩ ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 125 500 ⇩ ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 125 625 ⇩ ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 125 750 ⇩ ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 125 875 ⇩ ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

(Continued)
Appendix
TABLE T11.11 (Continued)
MIL-STD-105E Table IV-C—Multiple-Sampling Plans for Reduced Inspection (Master Table)
Appendix

Acceptable Quality Levels (Reduced Inspection)

Sample
Size Cumulative 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000
Code Sample Sample
Letter Sample Size Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re

P First 200 200 ⇩ * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 200 400 ⇩ ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 200 600 ⇩ ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 200 800 ⇩ ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 200 1000 ⇩ ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 200 1200 ⇩ ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 200 1400 ⇩ ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Q First 315 315 * ⇧ ⇩ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 315 630 ⇧ ⇩ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 315 945 ⇧ ⇩ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 315 1260 ⇧ ⇩ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 315 1575 ⇧ ⇩ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 315 1890 ⇧ ⇩ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 315 2205 ⇧ ⇩ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
R First 500 500 ⇧ ⇧ # 2 # 2 # 3 # 3 # 4 # 4 0 5 0 6 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Second 500 1000 ⇧ ⇧ # 2 # 3 # 3 0 4 0 5 1 6 1 7 3 9 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Third 500 1500 ⇧ ⇧ 0 2 0 3 0 4 0 5 1 6 2 8 3 9 6 12 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fourth 500 2000 ⇧ ⇧ 0 3 0 4 0 5 1 6 2 7 3 10 5 12 8 15 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Fifth 500 2500 ⇧ ⇧ 0 3 0 4 1 6 2 7 3 8 5 11 7 13 11 17 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Sixth 500 3000 ⇧ ⇧ 0 3 1 5 1 6 3 7 4 9 7 12 10 15 14 20 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧
Seventh 500 3500 ⇧ ⇧ 1 3 1 5 2 7 4 8 6 10 9 14 13 17 18 22 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, pp. 24–25.
a If, after the final sample, use acceptance number has been exceeded, but the rejection number has not been reached, accept the lot, but reinstate normal inspection.

⇩ = Use first sampling plan below arrow. If sample size equals, or exceeds, lot or batch size, do 100% inspection.
⇧ = Use first sampling plan above arrow.
Ac = Acceptance number.
Re = Rejection number.
* = Use corresponding single-sampling plan (or alternatively use multiple-sampling plan below, where available).
++ = Use corresponding double-sampling plan (or alternatively use multiple-sampling plan below, where available).
# = Acceptance not permitted at this sample size.
709
 710

TABLE T11.12
MIL-STD-105E Table V-A—Average Outgoing Quality Limit Factors for Normal Inspection (Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A 2 18 42 69 97 160 220 330 470 730 1100

B 3 12 28 46 65 110 150 220 310 490 720 1100
C 5 7.4 17 27 39 63 90 130 190 290 430 660
D 8 4.6 11 17 24 40 56 82 120 180 270 410
E 13 2.8 6.5 11 15 24 34 50 72 110 170 250
F 20 1.8 4.2 6.9 9.7 16 22 33 47 73
G 32 1.2 2.6 4.3 6.1 9.9 14 21 29 46
H 50 0.74 1.7 2.7 3.9 6.3 9.0 13 19 29
J 80 0.46 1.1 1.7 2.4 4.0 5.6 8.2 12 18
K 125 0.29 0.67 1.1 1.6 2.5 3.6 5.2 7.5 12
L 200 0.18 0.42 0.69 0.97 1.6 2.2 3.3 4.7 7.3
M 315 0.12 0.27 0.44 0.62 1.00 1.4 2.1 3.0 4.7
N 500 0.074 0.17 0.27 0.39 0.63 0.90 1.3 1.9 2.9
P 800 0.046 0.11 0.17 0.24 0.40 0.56 0.82 1.2 1.8
Q 1250 0.029 0.067 0.11 0.16 0.25 0.36 0.52 0.75 1.2
R 2000 0.042 0.069 0.097 0.16 0.22 0.33 0.47 0.73

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government
Printing Office, Washington, DC, 1989, p. 26.
æ sample size ö
Note:  For the exact AOQL, the above values must be multiplied by çç 1 - ÷.
è lot or batch size ÷ø
Appendix
 Appendix

TABLE T11.13
MIL-STD-105E Table V-B—Average Outgoing Quality Limit Factors for Tightened Inspection (Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A 2 42 69 97 160 260 400 620 970

B 3 12 28 46 65 110 170 270 410 650 1100
C 5 7.4 17 27 39 63 100 160 250 390 610
D 8 4.6 11 17 24 40 64 99 160 240 380
E 13 2.8 6.5 11 15 24 40 61 95 150 240
F 20 1.8 4.2 6.9 9.7 16 26 40 62
G 32 1.2 2.6 4.3 6.1 9.9 16 25 39
H 50 0.74 1.7 2.7 3.9 6.3 10 16 25
J 80 0.46 1.1 1.7 2.4 4.0 6.4 9.9 16
K 125 0.29 0.67 1.1 1.6 2.5 4.1 6.4 9.9
L 200 0.18 0.42 0.69 0.97 1.6 2.6 4.0 6.2
M 315 0.12 0.27 0.44 0.62 1.0 1.6 2.5 3.9
N 500 0.074 0.17 0.27 0.39 0.63 1.0 1.6 2.5
P 800 0.046 0.11 0.17 0.24 0.40 0.64 0.99 1.6
Q 1250 0.029 0.067 0.11 0.16 0.25 0.41 0.64 0.99
R 2000 0.018 0.042 0.069 0.097 0.16 0.26 0.40 0.62
S 3150 0.027

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 27.
æ sample size ö
Note:  For the exact AOQL, the above values must be multiplied by çç 1 - ÷.
è lot or batch size ø÷
711
 712

TABLE T11.14
MIL-STD-105E Table VI-A—Limiting Quality (in Percent Defective) for Which Pa = 10% (for Normal Inspection, Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

A 2 68
B 3 54
C 5 37 58
D 8 25 41 54
E 13 16 27 36 44
F 20 11 18 25 30 42
G 32 6.9 12 16 20 27 34
H 50 4.5 7.6 10 13 18 22 29
J 80 2.8 4.8 6.5 8.2 11 14 19 24
K 125 1.8 3.1 4.3 5.4 7.4 9.4 12 16 23
L 200 1.2 2.0 2.7 3.3 4.6 5.9 7.7 10 14
M 315 0.73 1.2 1.7 2.1 2.9 3.7 4.9 6.4 9.0
N 500 0.46 0.78 1.1 1.3 1.9 2.4 3.1 4.0 5.6
P 800 0.29 0.49 0.67 0.84 1.2 1.5 1.9 2.5 3.5
Q 1250 0.18 0.31 0.43 0.53 0.74 0.94 1.2 1.6 2.3
R 2000 0.20 0.27 0.33 0.46 0.59 0.77 1.0 1.4
Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government
Printing Office, Washington, DC, 1989, p. 28.
Appendix
 Appendix

TABLE T11.15
MIL-STD-105E Table VI-B—Limiting Quality (in Defects per Hundred Units) for Which Pa = 10% (for Normal Inspection, Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A 2 120 200 270 330 460 590 770 1000 1400 1900
B 3 77 130 180 220 310 390 510 670 940 1300 1800
C 5 46 78 110 130 190 240 310 400 560 770 1100
D 8 29 49 67 84 120 150 190 250 350 480 670
E 13 18 30 41 51 71 91 120 160 220 300 410
F 20 12 20 27 33 46 59 77 100 140
G 32 7.2 12 17 21 29 37 48 63 88
H 50 4.6 7.8 11 13 19 24 31 40 56
J 80 2.9 4.9 6.7 8.4 12 15 19 25 35
K 125 1.8 3.1 4.3 5.4 7.4 9.4 12 16 23
L 200 1.2 2.0 2.7 3.3 4.6 5.9 7.7 10 14
M 315 0.73 1.2 1.7 2.1 2.9 3.7 4.9 6.4 9.0
N 500 0.46 0.78 1.1 1.3 1.9 2.4 3.1 4.0 5.6
P 800 0.29 0.49 0.67 0.84 1.2 1.5 1.9 2.5 3.5
Q 1250 0.18 0.31 0.43 0.53 0.74 0.94 1.2 1.6 2.3
R 2000 0.20 0.27 0.33 0.46 0.59 0.77 1.0 1.4

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 29.
713
 714

TABLE T11.16
MIL-STD-105E Table VII-A—Limiting Quality (in Percent Defective) for Which Pa = 5% (for Normal Inspection,
Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

A 2 78
B 3 63
C 5 45 66
D 8 31 47 60
E 13 21 32 41 50
F 20 14 22 28 34 46
G 32 8.9 14 18 23 30 37
H 50 5.8 9.1 12 15 20 25 32
J 80 3.7 5.8 7.7 9.4 13 16 20 26
K 125 2.4 3.8 5.0 6.2 8.4 11 14 18 24
L 200 1.5 2.4 3.2 3.9 5.3 6.6 8.5 11 15
M 315 0.95 1.5 2.0 2.5 3.3 4.2 5.4 7.0 9.6
N 500 0.60 0.95 1.3 1.6 2.1 2.6 3.4 4.4 6.1
P 800 0.38 0.59 0.79 0.97 1.3 1.6 2.1 2.7 3.8
Q 1250 0.24 0.38 0.50 0.62 0.84 1.1 1.4 1.8 2.4
R 2000 0.24 0.32 0.39 0.53 0.66 0.85 1.1 1.5

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E,
U.S. Government Printing Office, Washington, DC, 1989, p. 30.
Appendix
 Appendix

TABLE T11.17
MIL-STD-105E Table VII-B—Limiting Quality (in Defects per Hundred Units) for Which Pa = 5% (for Normal Inspection, Single Sampling)
Acceptable Quality Level
Code Sample
Letter Size 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A 2 150 240 320 390 530 660 850 1100 1500 2000
B 3 100 160 210 260 350 440 570 730 1000 1400 1900
C 5 60 95 130 160 210 260 340 440 610 810 1100
D 8 38 59 79 97 130 160 210 270 380 510 710
E 13 23 37 48 60 81 100 130 170 230 310 440
F 20 15 24 32 39 53 66 85 110 150
G 32 9.4 15 20 24 33 41 53 68 95
H 50 6.0 9.5 13 16 21 26 34 44 61
J 80 3.8 5.9 7.9 9.7 13 16 21 27 38
K 125 2.4 3.8 5.0 6.2 8.4 11 14 18 24
L 200 1.5 2.4 3.2 3.9 5.3 6.6 8.5 11 15
M 315 0.95 1.5 2.0 2.5 3.3 4.2 5.4 7.0 9.6
N 500 0.60 0.95 1.3 1.6 2.1 2.6 3.4 4.4 6.1
P 800 0.38 0.59 0.79 0.97 1.3 1.6 2.1 2.7 3.8
Q 1250 0.24 0.38 0.50 0.62 0.84 1.1 1.4 1.8 2.4
R 2000 0.24 0.32 0.39 0.53 0.66 0.85 1.1 1.5

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing
Office, Washington, DC, 1989, p. 31.
715
 716

n Mult Do Dou
iple ubl Do Dou ble
e ub ble
e le M
ubl Mu M M
3/4n Do ltip ult ult ul
tip
le ipl
e
ipl
e le
1/2n

1/4n
c=1 c=2 c=3 c=5 c=7
0
0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 0 5 10 0 3 6 9 10
n × proportion defective
n Do Do Dou
ub Dou ub Dou ble
le ble le ble
3/4n M M M M
ul M ult ult ul
tip ult ipl
e ip le
tip
le ipl
e le
1/2n

1/4n

Averave sample size

c=8 c = 10 c = 12 c = 14 c = 18
0
0 7 14 0 4 8 12 16 0 3 6 9 12 15 18 0 3 6 9 12 15 18 21 0 5 10 15 20 25
n × proportion defective
n Do
ub Do Do
ubl Double Double
le ubl
e e
M

3/4n
M
ul

M M
ul

M
tip

ul ul
tip

ul
le

tip tip
le

le tip le
1/2n le

1/4n
c = 21 c = 27 c = 30 c = 41 c = 44
0
0 5 10 15 20 25 30 0 9 18 27 36 0 10 20 30 40 0 13 26 39 52 0 9 18 27 36 45 54
n × proportion defective

FIGURE F11.18
MIL-STD-105E Table IX—average sample size curves for double- and multiple-sampling (normal and tightened inspection). n is the equivalent single-sample size, c is
the single-sample acceptance number, ↑ is the AQL for normal inspection. (From United States Department of Defense, Military Standard, Sampling Procedures and Tables
for Inspection by Attributes, MIL-STD-105E, U.S. Government Printing Office, Washington, DC, 1989, p. 33.)
Appendix
TABLE T11.19
MIL-STD-105E X-F Tables for Sample Size Code Letter: F
Percent of lots
Appendix

expected to CHART F—Operating characteristic curves for single-sampling plans

reaccepted (Pa) (curves for double and multiple sampling are matched as closely as practicable)
100 100
90 90
80 80
70 70
60 60
50 50
40 40
30 30
20 0.65 2.5 4.0 6.5 10 20 15 25 40 65
10 10
0 0
10 20 30 40 50 20 40 60 80 100 120 140 160
Quality of submitted lots (p, in percent defective for AQL’s ≤ 10; in defects per hundred units for AQL’s > 10)
Note: Figures on curves are acceptable quality levels (AQL’s) for normal inspection.
TABLE X-F-1
Tabulated Values for Operating Characteristic Curves for Single-Sampling Plans
Acceptable Quality Levels (Normal Inspection)
0.65 2.5 4.0 6.5 10 0.65 2.5 4.0 6.5 10 15 X 25 X 40 X 65
Pa p (in Percent Defective) p (in Defects per Hundred Units)
99.0 0.050 0.75 2.25 4.31 9.75 0.051 0.75 2.18 4.12 8.92 14.5 17.5 23.9 30.5 37.4 517 62.9
95.0 0.256 1.80 4.22 7.13 14.0 0.257 1.78 4.09 6.83 13.1 19.9 23.5 30.8 38.5 46.2 62.2 74.5
90.0 0.525 2.69 5.64 9.03 16.6 0.527 2.66 5.51 8.73 15.8 23.3 27.2 35.1 43.2 51.5 68.4 81.2
75.0 1.43 4.81 8.70 12.8 21.6 1.44 4.81 8.68 12.7 21.1 29.8 34.2 43.1 52.1 61.2 79.5 93.4
50.0 3.41 8.25 13.1 18.1 27.9 3.47 8.39 13.4 18.4 28.4 38.3 43.3 53.3 63.3 73.3 93.3 108
25.0 6.70 12.9 18.7 24.2 34.8 6.93 13.5 19.6 25.5 37.1 48.4 54.0 65.1 76.1 87.0 109 125
10.0 10.9 18.1 24.5 30.4 41.5 11.5 19.5 26.6 33.4 46.4 58.9 65.0 77.0 88.9 101 124 141
5.0 13.9 21.6 28.3 34.4 45.6 15.0 23.7 31.5 38.8 52.6 65.7 72.2 84.8 97.2 109 133 151
1.0 20.6 28.9 35.6 42.0 53.4 23.0 33.2 42.0 50.2 65.5 80.0 87.0 101 114 127 153 172
1.0 4.0 6.5 10 X 1.0 4.0 6.5 10 15 X 25 X 40 X 65 X
Acceptable quality levels (tightened inspection)

Note: Binomial distributions used for percent defective computations; Poisson for defects per hundred units.
717

(Continued)
 718

TABLE T11.19 (Continued)

MIL-STD-105E X-F Tables for Sample Size Code Letter: F
Acceptable Quality Levels (Normal Inspection)

Less Higher
Type of Cumulative than 0.65 0.65 1.0 X 1.5 2.5 4.0 6.5 10 15 X 25 X 40 X 65 than 65 Cumulative
Sampling Sample Sample
Plan Size Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Ac Re Size

Single 20 ∇ 0 1 1 2 2 3 3 4 5 6 7 8 8 9 10 11 12 13 14 15 18 19 21 22 ∆ 20

Use Use Use

13 ∇ * 0 2 0 3 1 4 2 5 3 7 3 7 5 9 6 10 7 11 9 14 11 16 ∆ 13
Double Letter Letter Letter
26 1 2 3 4 4 5 6 7 8 9 11 12 12 13 15 16 18 19 23 24 26 27 26
E H G
5 ∇ * # 2 # 2 # 3 # 4 0 4 0 4 0 5 0 6 1 7 1 8 2 9 ∆ 5
10 # 2 0 3 0 3 1 5 1 6 2 7 3 8 3 9 4 10 6 12 7 14 10
15 0 2 0 3 1 4 2 6 3 8 4 9 6 10 7 12 8 13 11 17 13 19 15
Multiple 20 0 3 1 4 2 5 3 7 5 10 6 11 8 13 10 15 12 17 16 22 19 25 20
25 1 3 2 4 3 6 5 8 7 11 9 12 11 15 14 17 17 20 22 25 25 29 25
30 1 3 3 5 4 6 7 9 10 12 12 14 14 17 18 20 21 23 27 29 31 33 30
35 2 3 4 5 6 7 9 10 13 14 14 15 18 19 21 22 25 26 32 33 37 38 35
Less 1.0 X 1.5 2.5 4.0 6.5 10 15 X 25 X 40 X 65 X Higher
than 1.0 than 65
Acceptable quality levels (tightened inspections)

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Attributes, MIL-STD-105E, U.S. Government
Printing Office, Washington, DC, 1989, pp. 44–45.
∇ = use next preceding sample size code letter for which acceptance and rejection numbers are available.
∇ = use next subsequent sample size code letter for which acceptance and rejection numbers are available.
Ac = acceptance number
Re = rejection number
* = use single-sampling plan above (or alternatively use letter J).
# = acceptance not permitted at this sample size.
Appendix
TABLE T11.20
Appendix

MIL-STD-105E Scheme Average Outgoing Quality Limit Factors (in Defects per Hundred Units)
Acceptable Quality Level
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A (11)
13 30 48 78 130 200 310 450 710 1100
B (6.8)
7.5 19 32 52 84 130 210 300 480 710 1100
C (4.4) (12)
4.7 12 20 31 51 78 130 180 290 430 660
D (2.8) (7.0) (13)
2.9 7.0 12 20 32 49 76 120 180 270 410
E (1.9) (4.5) (7.5) (13)
1.9 4.5 7.4 12 20 30 47 69 110 170 260
F (1.2) (2.9) (4.9) (7.9) (14)
1.2 2.9 4.8 7.8 13 20 31 45 71
G (.74) (1.8) (3.0) (4.9) (8.1) (13)
.75 1.8 3.0 4.9 7.9 13 19 28 45
H (.47) (1.2) (2.0) (3.2) (5.1) (8.0) (13)
.47 1.2 2.0 3.1 5.1 7.8 13 18 29
J (.30) (.72) (1.2) (2.0) (3.2) (5.0) (7.7) (12)
.30 .72 1.2 2.0 3.2 4.9 7.6 12 18
K .19 .46 .77 1.3 2.1 3.2 4.9 7.2 12
L .12 .29 .48 .78 1.3 2.0 3.1 4.5 7.1
M .075 .18 .31 .50 .80 1.3 2.0 2.9 4.5
N .047 .12 .20 .31 .51 .78 1.3 1.8 2.9
P .030 .072 .12 .20 .32 .49 .76 1.2 1.8
Q .019 .046 .077 .13 .21 .32 .49 .72 1.2
R .029 .048 .078 .13 .20 .31 .45 .71

Source: Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10(3), 106, 1978. With permission.
Note: For a better approximation to the AOQL, the values must be multiplied by (1—normal plan sample size/lot or batch size). Also applicable to percent
defective for AQL less than 15 with specific values for percent defective shown in parenthesis.
719
 720

TABLE T11.21
MIL-STD-105E Scheme Limiting Quality (in Defects per Hundred Units) for Which Pa = 10%
Acceptable Quality Level
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A (53.6)
76.7 130 194 266 334 464 650 889 1240 1750
B (36.9)
46.0 77.8 130 177 223 309 433 593 825 1170 1680
C (25.0) (40.6)
28.8 48.6 77.8 106 134 185 260 356 495 699 1010
D (16.2) (26.8) (4..6)
17.7 29.9 48.6 66.5 83.5 116 162 222 309 437 631
E (10.9) (18.1) (26.8) (36.0)
11.5 19.4 29.9 40.9 51.4 71.3 100 137 190 269 388
F (6.94) (11.6) (18.1) (24.5) (30.4)
7.19 12.2 19.4 26.6 33.4 46.4 65.0 88.9 124
G (4.50) (7.56) (11.6) (15.8) (19.7) (27.1)
4.60 7.78 12.2 16.6 20.9 29.0 40.6 55.6 77.4
H (2.84) (4.77) (7.56) (10.3) (12.9) (17.8) (24.7)
2.88 4.86 7.78 10.6 13.4 18.5 26.0 35.6 49.5
J (1.83) (3.08) (4.77) (6.52) (8.16) (11.3) (15.7) (21.4)
1.84 3.11 4.86 6.65 8.35 11.6 16.2 22.2 30.9
K 1.15 1.94 3.11 4.26 5.34 7.42 10.4 14.2 19.8
L .731 1.23 1.94 2.66 3.34 4.64 6.50 8.89 12.4
M .460 .778 1.23 1.69 2.12 2.94 4.13 5.64 7.86
N .288 .486 .778 1.06 1.34 1.85 2.60 3.56 4.95
P .184 .311 .486 .665 .865 1.16 1.62 2.22 3.09
Q .115 .194 .311 .426 .534 .742 1.04 1.42 1.98
R .123 .194 .266 .334 .464 .650 .889 1.24

Source: Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10(3), 107, 1978. With permission.
Note: Also applicable to percent defective for AQL less than 15 with specific values for percent defective shown in parenthesis.
Appendix
TABLE T11.22
Appendix

MIL-STD-105E Scheme Limiting Quality (in Defects per Hundred Units) for Which Pa = 5%
Acceptable Quality Level
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A (63.2)
99.8 158 237 315 388 526 722 972 1340 1860
B (45.1)
59.9 94.9 158 210 258 350 481 648 890 1240 1770
C (31.2) (47.1)
37.4 59.3 94.9 126 155 210 289 389 534 745 1060
D (20.6) (31.6) (47.1)
23.0 36.5 59.3 78.7 96.9 131 180 243 334 465 665
E (13.9) (21.6) (31.6) (41.0)
15.0 23.7 36.5 48.4 59.6 80.9 111 150 205 286 409
F (8.94) (14.0) (21.6) (28.3) (34.4)
9.36 14.8 23.7 31.5 38.8 52.6 72.2 97.2 133
G (5.81) (9.14) (14.0) (18.4) (22.5) (30.1)
5.99 9.49 14.8 19.7 24.2 32.9 45.1 60.8 83.4
H (3.68) (5.79) (9.14) (12.1) (14.8) (19.9) (27.0)
3.74 5.93 9.49 12.6 15.5 21.0 28.9 38.9 53.4
J (2.37) (3.74) (5.79) (7.66) (9.41) (12.7) (17.3) (23.2)
2.40 3.79 5.93 7.87 9.69 13.1 18.0 24.3 33.4
K 1.50 2.37 .379 5.04 6.20 8.41 11.5 15.6 21.4
L .951 1.51 2.37 3.15 3.88 5.26 7.22 9.72 13.3
M .599 .949 1.51 2.00 2.46 3.34 4.58 6.17 8.47
N .374 .593 .949 1.26 1.55 2.10 2.89 3.89 5.34
P .240 .379 .593 .787 .969 1.31 1.80 2.43 3.34
Q .150 .237 .379 .504 .620 .841 1.15 1.56 2.14
R .151 .237 .315 .388 .526 .722 .972 1.33

Source: Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10(3), 108, 1978. With permission.
Note: Also applicable to percent defective for AQL less than 15 with specific values for percent defective shown in parenthesis.
721
 722

TABLE T11.23
Scheme Measures of Performance for MIL-STD-105E, Code F

TABLE A4-F-1
Tabulated Values for Operating Characteristic Curves for Scheme
Acceptable Quality Levels (Normal Inspection)
0.65 2.5 4.0 6.5 10 .65 2.5 4.0 6.5 10 15 25 40 65
Pa p (in Percent Defective) p (in Defects per Hundred Units)

99.0 0.104 .978 2.94 4.93 10.1 0.104 .958 2.84 4.72 9.41 15.0 25.0 39.5 64.9
95.0 0.357 1.85 4.11 6.94 13.0 0.358 1.82 4.02 6.69 12.3 19.2 30.2 45.7 73.4
90.0 0.571 2.47 4.91 8.24 14.4 0.572 2.45 4.82 8.00 13.8 21.4 33.3 49.7 78.3
75.0 1.11 3.66 6.40 10.4 16.5 1.11 3.66 6.37 10.3 16.2 24.8 38.0 56.0 85.5
50.0 2.22 5.40 8.71 13.6 19.2 2.24 5.46 8.85 13.8 19.5 29.4 44.3 64.2 94.8
25.0 4.24 8.21 12.9 18.7 24.3 4.34 8.43 13.5 19.6 25.6 37.2 54.0 76.1 109
10.0 6.94 11.6 18.1 24.5 30.4 7.19 12.2 19.4 26.6 33.4 46.4 65.0 88.9 124
5.0 8.94 14.0 21.6 28.3 34.4 9.36 14.8 23.7 31.5 38.8 52.6 72.2 97.2 133
1.0 13.4 19.0 28.9 35.8 42.1 14.4 20.7 33.2 42.0 50.2 65.5 87.1 114 153
Appendix
 Appendix

TABLE A4-F-2
Tabulated Values for Average Sample Number Curves for Scheme
Acceptable Quality Levels (Normal Inspection)
.65 2.5 4.0 6.5 10 .65 2.5 4.0 6.5 10 15 25 40 65
Pa p (in Percent Defective) p (in Defects per Hundred Units)
99.0 9.5 14.6 13.4 15.7 17.9 9.5 14.5 13.2 15.3 16.8 17.8 16.2 15.1 15.7
95.0 14.4 19.1 18.5 19.5 19.9 14.4 19.0 18.3 19.3 19.8 20.0 19.9 19.8 19.9
90.0 18.6 21.5 19.7 19.9 20.0 18.6 21.5 19.6 19.9 20.0 20.0 20.0 20.0 20.0
75.0 26.1 26.2 20.0 20.0 20.0 26.0 26.2 20.0 20.0 20.0 20.0 20.0 20.0 20.0
50.0 31.0 30.9 20.0 20.0 20.0 31.0 30.9 20.0 20.0 20.0 20.0 20.0 20.0 20.0
25.0 32.0 32.0 20.0 20.0 20.0 32.0 32.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
10.0 32.0 32.0 20.0 20.0 20.0 32.0 32.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
5.0 32.0 32.0 20.0 20.0 20.0 32.0 32.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
1.0 32.0 32.0 20.0 20.0 20.0 32.0 32.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
723
 724

TABLE A4-F-3
Tabulated Values for Average Outgoing Quality Curves for Scheme (Lot Size = 120)
Acceptable Quality Levels (Normal Inspection)
.65 2.5 4.0 6.5 10 .65 2.5 4.0 6.5 10 15 25 40 65
Pa p (in Percent Defective) p (in Defects per Hundred Units)

99.0 0.88 0.85 2.6 4.2 8.5 0.088 0.83 2.5 4.1 8.0 13 21 34 56
95.0 0.30 1.5 3.3 5.5 10 0.30 1.5 3.2 5.3 9.8 15 24 36 58
90.0 0.44 1.8 3.7 6.2 11 0.44 1.8 3.6 6.0 10 16 25 37 59
75.0 0.65 2.2 4.0 6.5 10 0.66 2.2 4.0 6.5 10 15 24 35 53
50.0 0.83 2.0 3.6 5.7 8.0 0.83 2.0 3.7 5.8 8.1 12 18 27 39
25.0 0.78 1.5 2.7 3.9 5.1 0.80 1.5 2.8 4.1 5.3 7.7 11 16 23
10.0 0.51 0.85 1.5 2.0 2.5 0.53 0.89 1.6 2.2 2.8 3.9 5.4 7.4 10
5.0 0.33 0.51 0.90 1.2 1.4 0.34 0.54 0.99 1.3 1.6 2.2 3.0 4.1 5.6
1.0 0.99 0.14 0.24 0.30 0.35 0.11 0.15 0.28 0.35 0.42 0.55 0.72 0.94 1.3
AOQL 0.85 2.2 4.1 6.6 11 0.86 2.2 4.0 6.5 11 17 26 38 59
Appendix
 Appendix

TABLE A4-F-4
Tabulated Values for Average Total Inspection Curves for Scheme (Lot Size = 120)
Acceptable Quality Levels (Normal Inspection)
.65 2.5 4.0 6.5 10 .65 2.5 4.0 6.5 10 15 25 40 65
Pa p (in Percent Defective) p (in Defects per Hundred Units)
99.0 10.5 15.6 14.4 16.7 18.9 10.5 15.5 14.2 16.3 17.8 18.8 17.2 16.1 16.8
95.0 19.5 24.1 23.5 24.5 25.0 19.5 24.0 23.3 24.3 24.8 25.0 24.9 24.8 24.9
90.0 28.5 31.2 29.7 29.9 30.0 28.5 31.1 29.6 29.9 30.0 30.0 30.0 30.0 30.0
75.0 49.2 49.2 45.0 45.0 45.0 49.2 49.2 45.0 45.0 45.0 45.0 45.0 45.0 45.0
50.0 75.4 75.2 70.0 70.0 70.0 75.4 75.2 70.0 70.0 70.0 70.0 70.0 70.0 70.0
25.0 98.0 98.0 95.0 95.0 95.0 98.0 98.0 95.0 95.0 95.0 95.0 95.0 95.0 95.0
10.0 111 111 110 110 110 111 111 110 110 110 110 110 110 110
5.0 116 116 115 115 115 116 116 115 115 115 115 115 115 115
1.0 119 119 119 119 119 119 119 119 119 119 119 119 119 119

Source: Reprinted from Schilling, E.G. and Sheesley, J.H., J. Qual. Technol., 10(3), 114, 1978. With permission.
725
 726

TABLE T11.24
Operating Ratios for the MIL-STD-105E Scheme (R = p.10/p.95, Calculated Using Poisson Distribution)
Acceptable Quality Level
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10 15 25 40 65 100 150 250 400 650 1000

A 30.32 7.43 5.01 4.02 2.72 2.42 2.15 1.95 1.69 1.58
B 23.23 6.71 5.04 3.99 2.72 2.41 2.15 1.95 1.69 1.58 1.46
C 20.14 6.67 4.96 3.98 2.72 2.41 2.15 1.95 1.68 1.57 1.47
D 19.34 6.54 4.81 4.01 2.71 2.42 2.14 1.95 1.68 1.57 1.46
E 20.54 6.88 4.83 4.01 2.72 2.41 2.15 1.94 1.68 1.57 1.46
F 20.08 6.70 4.83 3.98 2.72 2.42 2.15 1.95 1.69
G 20.63 7.01 4.86 3.97 2.72 2.42 2.15 1.94 1.69
H 20.14 6.67 4.96 3.98 2.72 2.41 2.15 1.95 1.68
J 20.51 6.81 4.86 4.01 2.71 2.42 2.14 1.95 1.68
K 20.07 6.64 4.84 4.02 2.71 2.42 2.15 1.94 1.69
L 20.42 6.72 4.83 3.98 2.72 2.42 2.15 1.95 1.69
M 20.18 6.88 4.82 3.98 2.71 2.41 2.15 1.94 1.69
N 20.14 6.67 4.96 3.98 2.72 2.41 2.15 1.95 1.68
P 20.42 6.81 4.81 4.01 2.71 2.42 2.14 1.95 1.68
Q 20.07 6.64 4.84 4.02 2.71 2.42 2.15 1.94 1.69
R 6.72 4.83 3.98 2.72 2.42 2.15 1.95 1.69

Source: Reprinted from Schilling, E.G. and Johnson, L.I., J. Qual. Technol., 12(4), 226, 1980. With permission.
Appendix
Appendix 727

TABLE T12.1
MIL-STD-414 Table B-6—Values of T for Tightened Inspection: Standard Deviation Method
Sample Acceptable Quality Levels (in Percent Defective)
Size
Code Number
Letter .040 .065 .10 .15 .25 .40 .65 1.0 1.5 2.5 4.0 6.5 10.0 15.0 of Lots

2 3 4 4 4 5
B * * * * * * * * * 4 5 6 7 8 10
5 6 8 9 11 15
2 2 3 3 4 4 4 5
C * * * * * * * 3 4 5 6 7 7 8 10
5 6 7 8 9 10 11 15
2 3 3 3 4 4 4 4 5
D * * * * * * 4 4 5 6 6 7 7 8 10
5 6 7 8 9 10 10 11 15
2 3 3 3 4 4 4 4 4 4 5
E * * * * 4 4 5 5 6 6 7 7 8 8 10
5 6 6 7 8 9 9 10 11 11 15
3 3 3 3 4 4 4 4 4 4 4 5
F * * * 4 5 5 6 6 6 7 7 8 8 8 10
6 6 7 8 8 9 9 10 11 11 11 15
3 3 3 3 3 4 4 4 4 4 4 4 4 4 5
G 4 5 5 5 6 6 6 7 7 7 7 8 8 8 10
6 6 6 7 7 8 9 9 9 10 10 11 11 11 15
3 3 3 3 4 4 4 4 4 4 4 4 4 4 5
H 5 5 5 6 6 6 7 7 7 7 8 8 8 8 10
6 7 7 8 8 9 9 9 10 10 11 11 11 11 15
3 3 4 4 4 4 4 4 4 4 4 4 4 4 5
I 5 6 6 6 6 7 7 7 7 7 8 8 8 8 10
7 7 8 8 9 9 9 10 10 10 11 11 11 11 15
3 4 4 4 4 4 4 4 4 4 4 4 4 4 5
J 6 6 6 6 7 7 7 7 7 8 8 8 8 8 10
8 8 8 9 9 9 10 10 10 11 11 11 11 11 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
K 6 6 6 6 7 7 7 7 8 8 8 8 8 8 10
8 8 9 9 9 9 10 10 10 11 11 11 11 11 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
L 6 6 6 7 7 7 7 7 8 8 8 8 8 8 10
8 9 9 9 9 10 10 10 10 11 11 11 11 11 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
M 6 7 7 7 7 7 7 7 8 8 8 8 8 8 10
9 9 9 9 10 10 10 10 11 11 11 11 11 11 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
N 7 7 7 7 7 7 8 8 8 8 8 8 8 8 10
9 9 10 10 10 10 11 11 11 11 11 11 11 11 15
(Continued)
728 Appendix

TABLE T12.1 (Continued)

MIL-STD-414 Table B-6—Values of T for Tightened Inspection: Standard Deviation Method
Sample Acceptable Quality Levels (in Percent Defective)
Size
Code Number
Letter .040 .065 .10 .15 .25 .40 .65 1.0 1.5 2.5 4.0 6.5 10.0 15.0 of Lots
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
O 7 7 7 7 7 8 8 8 8 8 8 8 8 8 10
10 10 10 10 10 11 11 11 11 11 11 11 11 11 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
P 7 7 7 8 8 8 8 8 8 8 8 8 8 8 10
10 10 10 10 11 11 11 11 11 11 11 11 11 12 15
4 4 4 4 4 4 4 4 4 4 4 4 4 4 5
Q 7 8 8 8 8 8 8 8 8 8 8 8 8 8 10
10 11 11 11 11 11 11 11 11 11 11 11 11 12 15

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by
Variables for Percent Defective, MIL-STD-414, U.S. Government Printing Office, Washington, DC, 1957,
pp. 54, 55.
* There are no sampling plans provided in this Standard for these code letters and AQL values.
The top figure in each block refers to the preceding 5 lots, the middle figure to the preceding 10 lots, and the
­bottom figure to the preceding 15 lots.
Tightened inspection is required when the number of lots with estimate of percent defective above the AQL from
the preceding 5, 10, or 15 lots is greater than the given value of T in the table, and the process average from
these lots exceeds the AQL.
All estimates of the lot percent defective are obtained from Table B-5.
TABLE T12.2
MIL-STD-414 Table B-7—Limits of Estimated Lot Percent Defective for Reduced Inspection: Standard Deviation Method
Sample Acceptable Quality Levels (in Percent Defective)
Appendix

Size
Code Number
Letter .040 .065 .10 .15 .25 .40 .65 1.0 1.5 2.5 4.0 6.5 10.0 15.0 of Lots

B * * * * * * * * * [42]** [28]** [18]** [12]** [9]**

.77 5
C * * * * * * * [45]** [31]** [22]** [15]** [10]** [7]** 15.00 10
⇧ 15
0.00 .74 6.06 5
D * * * * * * [33]** [25]** [18]** [13]** [9]** 4.40 9.96 15.00 10
6.50 10.00 ⇧ 15
.00 .00 .13 1.38 4.24 9.09 5
E * * * * [25]** [18]** [14]** [11]** .10 .88 2.65 5.96 10.00 15.00 10
.88 2.49 4.00 6.50 ⇧ ⇧ 15
⇩ .000 .000 .000 .003 .044 .306 1.05 2.81 5.79 10.47 5
F * * * .000 .001 .016 .101 .317 .74 1.80 3.56 6.50 10.00 15.00 10
.002 .029 .123 .369 .81 1.50 2.50 4.00 ⇧ ⇧ ⇧ 15
⇩ .000 .000 .000 .002 .011 .047 .136 .323 .84 1.84 3.80 6.86 11.52 5
G .000 .002 .006 .018 .057 .143 .330 .643 1.14 2.23 3.94 6.50 10.00 15.00 10
.003 .010 .028 .062 .151 .315 .626 1.00 1.50 2.50 4.00 ⇧ ⇧ ⇧ 15
.000 .000 .002 .005 .017 .048 .123 .266 .521 1.14 2.24 4.29 7.40 12.07 5
H .004 .010 .023 .048 .111 .225 .445 .785 1.31 2.40 4.00 6.50 10.00 15.00 10
.013 .029 .058 .105 .215 .396 .65 1.00 1.50 2.50 ⇧ ⇧ ⇧ ⇧ 15
.001 .002 .006 .014 .037 .083 .185 .360 .653 1.33 2.49 4.59 7.74 12.43 5
I .009 .020 .039 .071 .146 .274 .509 .863 1.39 2.48 4.00 6.50 10.00 15.00 10
.021 .043 .077 .133 .248 .40 .65 1.00 1.50 2.50 ⇧ ⇧ ⇧ ⇧ 15
.002 .005 .012 .023 .054 .113 .233 .431 .750 1.47 2.66 4.81 7.98 12.69 5
J .013 .027 .050 .087 .169 .306 .550 .909 1.44 2.50 4.00 6.50 10.00 15.00 10
.027 .052 .089 .146 .25 .40 .65 1.00 1.50 ⇧ ⇧ ⇧ ⇧ ⇧ 15
(Continued)
729
TABLE T12.2 (Continued)
730

MIL-STD-414 Table B-7—Limits of Estimated Lot Percent Defective for Reduced Inspection: Standard Deviation Method
Sample Acceptable Quality Levels (in Percent Defective)
Size
Code Number
Letter .040 .065 .10 .15 .25 .40 .65 1.0 1.5 2.5 4.0 6.5 10.0 15.0 of Lots
.004 .008 .017 .032 .069 .137 .270 .483 .821 1.57 2.79 4.96 8.15 12.88 5
K .017 .033 .059 .099 .186 .328 .577 .940 1.47 2.50 4.00 6.50 10.00 15.00 10
.032 .058 .097 .15 .25 .40 .65 1.00 1.50 ⇧ ⇧ ⇧ ⇧ ⇧ 15
.005 .011 .022 .040 .082 .157 .300 .525 .876 1.64 2.88 5.08 8.29 13.03 5
L .020 .038 .065 .108 .199 .343 .596 .961 1.49 2.50 4.00 6.50 10.00 15.00 10
.035 .063 .10 .15 .25 .40 .65 1.00 1.50 ⇧ ⇧ ⇧ ⇧ ⇧ 15
.008 .016 .030 .052 .102 .187 .345 .587 .959 1.76 3.03 5.27 8.50 13.25 5
M .025 .045 .075 .120 .215 .364 .621 .989 1.50 2.50 4.00 6.50 10.00 15.00 10
.04 .065 .10 .15 .25 .40 .65 1.00 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ 15
.014 .026 .044 .072 .134 .235 .414 .681 1.082 1.92 3.24 5.52 8.81 13.60 5
N .031 .054 .087 .136 .236 .389 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00 10
.04 .065 .10 .15 .25 .40 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ 15
.018 .032 .053 .085 .153 .261 .453 .733 1.149 2.01 3.36 5.67 8.98 13.80 5
O .034 .058 .093 .143 .245 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00 10
.04 .065 .10 .15 .25 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ 15
.023 .039 .064 .101 1.77 2.96 .501 7.99 1.237 2.13 3.52 5.87 9.22 14.07 5
P .038 .064 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00 10
.04 .065 ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ 15
.025 .044 .069 .108 .188 .312 .525 .830 1.276 2.19 3.59 5.96 9.32 14.19 5
Q .04 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00 10
⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ ⇧ 15
Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC, 1957, pp. 56, 57.
Notes: *There are no sampling plans provided in this Standard for these code letters and AQL values. All AQL and table values, expect those in the brackets,
are in percent defective.
⇧⇩ Use the first figure in direction of arrow and corresponding number of lots. In each block, the top figure refers to the preceding 5 lots, the middle figure to
the preceding 10 lots, and the bottom figure to the preceding 15 lots.
Reduced inspection may be instituted when every estimated lot percent defective from the preceding 5, 10, or 15 lots is below the figure given in the table;
reduced inspection for sampling plans marked (**) in the table requires that the estimated lot percent defective is equal to zero for the number of consecutive
lots indicated in brackets. In addition, all other conditions for reduced inspection, in Part III of Section B, must be satisfied.
Appendix

All estimates of the lot percent defective are obtained from Table B-5.
Appendix 731

TABLE T12.3
MIL-STD-414 Table A-1—AQL Conversion Table
For Specified AQL Values Falling
within These Ranges Use This AQL Value

—– to 0.049 0.04
0.050–0.069 0.065
0.070–0.109 0.10
0.110–0.164 0.15
0.165–0.279 0.25
0.280–0.439 0.40
0.440–0.699 0.65
0.700–1.09 1.0
1.10–1.64 1.5
1.65–2.79 2.5
2.80–4.39 4.0
4.40–6.99 6.5
7.00–10.9 10.0
11.00–16.4 15.0

Source: United States Department of Defense, Military

Standard, Sampling Procedures and Tables for Inspection
by  Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC,
1957, p. 4.
732 Appendix

TABLE T12.4
MIL-STD-414 Table A-2—Sample Size Code Lettersa
Inspection Levels
Lot Size I II III IV V
3–8 B B B B C
9–15 B B B B D
16–25 B B B C E
26–40 B B B D F
41–65 B B C E G
66–110 B B D F H
111–180 B C E G I
181–300 B D F H J
301–500 C E G I K
501–800 D F H J L
801–1,300 E G I K L
1,301–3,200 F H J L M
3,201–8,000 G I L M N
8,001–22,000 H J M N O
22,001–110,000 I K N O P
110,001–550,000 I K O P Q
550,001 and over I K P Q Q

Source: United States Department of Defense, Military Standard, Sampling Procedures

and Tables for Inspection by Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC, 1957, p. 4.
a Sample size code letters given in body of table are applicable when the indicated

inspection levels are to be used.

TABLE T12.5
Appendix

MIL-STD-414 Table B-3—Master Table for Normal and Tightened Inspection for Plans Based on Variability Unknown:
Standard Deviation Method (Double Specification Limit and Form 2, Single-Specification Limit)

Sample Acceptable Quality Levels (Normal Inspection)

Size .040 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
Code Sample
Letter Size M M M M M M M M M M M M M M

B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 7.59 18.86 26.94 33.69 40.47

C 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.53 5.50 10.92 16.45 22.86 29.45 36.90
D 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.33 3.32 5.83 9.80 14.39 20.19 26.56 33.99
E 7 ⇩ ⇩ ⇩ ⇩ 0.422 1.06 2.14 3.55 5.35 8.40 12.20 17.35 23.29 30.50
F 10 ⇩ ⇩ ⇩ 0.349 0.716 1.30 2.17 3.26 4.77 7.29 10.54 15.17 20.74 27.57
G 15 0.099 0.186 0.312 0.503 0.818 1.31 2.11 3.05 4.31 6.56 9.46 13.71 18.94 25.61
H 20 0.135 0.228 0.365 0.544 0.846 1.29 2.05 2.95 4.09 6.17 8.92 12.99 18.03 24.53
I 25 0.155 0.250 0.380 0.551 0.877 1.29 2.00 2.86 3.97 5.97 8.63 12.57 17.51 23.97
J 30 0.179 0.280 0.413 0.581 0.879 1.29 1.98 2.83 3.91 5.86 8.47 12.36 17.24 23.58
K 35 0.170 0.264 0.388 0.535 0.847 1.23 1.87 2.68 3.70 5.57 8.10 11.87 16.65 22.91
L 40 0.179 0.275 0.401 0.566 0.873 1.26 1.88 2.71 3.72 5.58 8.09 11.85 16.61 22.86
M 50 0.163 0.250 0.363 0.503 0.789 1.17 1.71 2.49 3.45 5.20 7.61 11.23 15.87 22.00
N 75 0.147 0.228 0.330 0.467 0.720 1.07 1.60 2.29 3.20 4.87 7.15 10.63 15.13 21.11
O 100 0.145 0.220 0.317 0.447 0.689 1.02 1.53 2.20 3.07 4.69 6.91 10.32 14.75 20.66
P 150 0.134 0.203 0.293 0.413 0.638 0.949 1.43 2.05 2.89 4.43 6.57 9.88 14.20 20.02
Q 200 0.135 0.204 0.294 0.414 0.637 0.945 1.42 2.04 2.87 4.40 6.53 9.81 14.12 19.92
.065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
Acceptability Quality Levels (tightened inspection)

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent
Defective, MIL-STD-414, U.S. Government Printing Office, Washington, DC, 1957, p. 45.
All AQL and table values are in percent defective.
⇩ Use first sampling plan below arrow, that is, both sample size as well as M value. When sample size equals or exceeds lot size, every
item in the lot must be inspected.
733
 734

TABLE T12.6
MIL-STD-414 Table B-4—Master Table for Reduced Inspection for Plans Based on Variability Unknown:
Standard Deviation Method (Double Specification Limit and Form 2, Single-Specification Limit)
Acceptable Quality Levels (Reduced Inspection)
Sample .040 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00
Size Code Sample
Letter Size M M M M M M M M M M M M M

B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 7.59 18.86 26.94 33.69 40.47

C 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 7.59 18.86 26.94 33.69 40.47
D 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 7.59 18.86 26.94 33.69 40.47
E 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 7.59 18.86 26.94 33.69 40.47
F 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.53 5.50 10.92 16.45 22.86 29.45 36.90
G 5 ⇩ ⇩ ⇩ ⇩ ⇩ 1.33 3.32 5.83 9.80 14.39 20.19 26.56 33.99
H 7 ⇩ ⇩ ⇩ 0.422 1.06 2.14 3.55 5.35 8.40 12.20 17.35 23.29 30.50
I 10 ⇩ ⇩ 0.349 0.716 1.30 2.17 3.26 4.77 7.29 10.54 15.17 20.74 27.57
J 10 ⇩ ⇩ 0.349 0.716 1.30 2.17 3.26 4.77 7.29 10.54 15.17 20.74 27.57
K 15 0.186 0.312 0.503 0.8.18 1.31 2.11 3.05 4.31 6.56 9.46 13.71 18.94 25.61
L 20 0.228 0.365 0.544 0.846 1.29 2.05 2.95 4.09 6.17 8.92 12.99 18.03 24.53
M 20 0.228 0.365 0.544 0.846 1.29 2.05 2.95 4.09 6.17 8.92 12.99 18.03 24.53
N 25 0.250 0.380 0.551 0.877 1.29 2.00 2.86 3.97 5.97 8.63 12.57 17.51 23.97
O 30 0.280 0.413 0.581 0.879 1.29 1.98 2.83 3.91 5.86 8.47 12.36 17.24 23.58
P 50 0.250 0.363 0.503 0.789 1.17 1.71 2.49 3.45 5.20 7.61 11.23 15.87 22.00
Q 75 0.228 0.330 0.467 0.720 1.07 1.60 2.29 3.20 4.87 7.15 10.63 15.13 21.11

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent
Defective, MIL-STD-414, U.S. Government Printing Office, Washington, DC, 1957, p. 46.
All AQL and table values are in percent defective.
⇩ Use first sampling plan below arrow, that is, both sample size as well as M value. When sample size equals or exceeds lot size,
every item in the lot must be inspected.
Appendix
TABLE T12.7
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

0 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00 50.00
.10 47.24 46.67 46.44 46.26 46.16 46.10 46.08 46.06 46.05 46.05 46.04 46.04 46.03 46.03 46.02 46.02
.20 44.46 43.33 42.90 42.54 42.35 42.24 42.19 42.16 42.15 42.13 42.13 42.11 42.10 42.09 42.08 42.08
.30 41.63 40.00 39.37 38.87 38.60 38.44 38.37 38.33 38.31 38.29 38.28 38.27 38.25 38.24 38.22 38.22
.31 41.35 39.67 39.02 38.50 38.23 38.06 37.99 37.95 37.93 37.91 37.90 37.89 37.87 37.86 37.84 37.84
.32 41.06 39.33 38.67 38.14 37.86 37.69 37.62 37.58 37.55 37.54 37.52 37.51 37.49 37.48 37.46 37.46
.33 40.77 39.00 38.32 37.78 37.49 37.31 37.24 37.20 37.18 37.16 37.15 37.13 37.11 37.10 37.09 37.08
.34 40.49 38.67 37.97 37.42 37.12 36.94 36.87 36.83 36.80 36.78 36.77 36.75 36.73 36.72 36.71 36.71
.35 40.20 38.33 37.62 37.06 36.75 36.57 36.49 36.45 36.43 36.41 36.40 36.38 36.36 36.35 36.33 36.33
.36 39.91 38.00 37.28 36.69 36.38 36.20 36.12 36.08 36.05 36.04 36.02 36.01 35.98 35.97 35.96 35.96
.37 39.62 37.67 36.93 36.33 36.02 35.83 35.75 35.71 35.68 35.66 35.65 35.63 35.61 35.60 35.59 35.58
.38 39.33 37.33 36.58 35.98 35.65 35.46 35.38 35.34 35.31 35.29 35.28 35.26 35.24 35.23 35.22 35.21
.39 39.03 37.00 36.23 35.62 35.29 35.10 35.01 34.97 34.94 34.93 34.91 34.89 34.87 34.86 34.85 34.84
.40 38.74 36.67 35.88 35.26 34.93 34.73 34.65 34.60 34.58 34.56 34.54 34.53 34.50 34.49 34.48 34.47
.41 38.45 36.33 35.54 34.90 34.57 34.37 34.28 34.24 34.21 34.19 34.18 34.16 34.13 34.12 34.11 34.10
.42 38.15 36.00 35.19 34.55 34.21 34.00 33.92 33.87 33.85 33.83 33.81 33.79 33.77 33.76 33.74 33.74
.43 37.85 35.67 34.85 34.19 33.85 33.64 33.56 33.51 33.48 33.46 33.45 33.43 33.40 33.39 33.38 33.37
.44 37.56 35.33 34.50 33.84 33.49 33.28 33.20 33.15 33.12 33.10 33.09 33.07 33.04 33.03 33.02 33.01
.45 37.26 35.00 34.16 33.49 33.13 32.92 32.84 32.79 32.76 32.74 32.73 32.71 32.68 32.67 32.66 32.65
.46 36.96 34.67 33.81 33.13 32.78 32.57 32.48 32.43 32.40 32.38 32.37 32.35 32.32 32.31 32.30 32.29
.47 36.66 34.33 33.47 32.78 32.42 32.21 32.12 32.07 32.04 32.02 32.01 31.99 31.96 31.95 31.94 31.93
.48 36.35 34.00 33.12 32.43 32.07 31.85 31.77 31.72 31.69 31.67 31.65 31.63 31.61 31.60 31.58 31.58
.49 36.05 33.67 32.78 32.08 31.72 31.50 31.41 31.36 31.33 31.31 31.30 31.28 31.25 31.24 31.23 31.22
.50 35.75 33.33 32.44 31.74 31.37 31.15 31.06 31.01 30.98 30.96 30.95 30.93 30.90 30.89 30.87 30.87
.51 35.44 33.00 32.10 31.39 31.02 30.80 30.71 30.66 30.63 30.61 30.60 30.57 30.55 30.54 30.52 30.52
.52 35.13 32.67 31.76 31.04 30.67 30.45 30.36 30.31 30.28 30.26 30.25 30.23 30.20 30.19 30.17 30.17
(Continued)
735
TABLE T12.7 (Continued)
736

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

.53 34.82 32.33 31.42 30.70 30.32 30.10 30.01 29.96 29.93 29.91 29.90 29.88 29.85 29.84 29.83 29.82
.54 34.51 32.00 31.08 30.36 29.98 29.76 29.67 29.62 29.59 29.97 29.55 29.53 29.51 29.49 29.48 29.48
.55 34.20 31.67 30.74 30.01 29.64 29.41 29.32 29.27 29.24 29.22 29.21 29.19 29.16 29.15 29.14 29.13
.56 33.88 31.33 30.40 29.67 29.29 29.07 28.98 28.93 28.90 28.88 28.87 28.85 28.82 28.81 28.79 28.79
.57 33.57 31.00 30.06 29.33 28.95 28.73 28.64 28.59 28.56 28.54 28.53 28.51 28.48 28.47 28.45 28.45
.58 33.25 30.67 29.73 28.99 28.61 28.39 28.30 28.25 28.22 28.20 28.19 28.17 28.14 28.13 28.12 28.11
.59 32.93 30.33 29.39 28.66 28.28 28.05 27.96 27.92 27.89 27.87 27.85 27.83 27.81 27.79 27.78 27.77
.60 32.61 30.00 29.05 28.32 27.94 27.72 27.63 27.58 27.55 27.53 27.52 27.50 27.47 27.46 27.45 27.44
.61 32.28 29.67 28.72 27.98 27.60 27.39 27.30 27.25 27.22 27.20 27.18 27.16 27.14 27.13 27.11 27.11
.62 31.96 29.33 28.39 27.65 27.27 27.05 26.96 26.92 26.89 26.87 26.85 26.83 26.81 26.80 26.78 26.78
.63 31.63 29.00 28.05 27.32 26.94 26.72 26.63 26.59 26.54 26.54 26.52 26.50 26.48 26.47 26.45 26.45
.64 31.30 28.67 27.72 26.99 26.61 26.39 26.31 26.26 26.23 26.21 26.20 26.18 26.15 26.14 26.13 26.12
.65 30.97 28.33 27.39 26.66 26.28 26.07 25.98 25.93 25.90 25.88 25.87 25.85 25.83 25.82 25.80 25.80
.66 30.63 28.00 27.06 26.33 25.96 25.74 25.66 25.61 25.58 25.56 25.55 25.53 25.51 25.49 25.48 25.48
.67 30.30 27.67 26.73 26.00 25.63 25.42 25.33 25.29 25.26 25.24 25.23 25.21 25.19 25.17 25.16 25.16
.68 29.96 27.33 26.40 25.68 25.31 25.10 25.01 24.97 24.94 24.92 24.91 24.89 24.87 24.86 24.84 24.84
.69 29.61 27.00 26.07 25.35 24.99 24.78 24.70 24.65 24.62 24.60 24.59 24.57 24.55 24.54 24.53 24.52
.70 29.27 26.67 25.74 25.03 24.67 24.46 24.38 24.33 24.31 24.29 24.28 24.26 24.24 24.23 24.21 24.21
.71 28.92 26.33 25.41 24.71 24.35 24.15 24.06 24.02 23.99 23.98 23.96 23.95 23.92 23.91 23.90 23.90
.72 28.57 26.00 25.09 24.39 24.03 23.83 23.75 23.71 23.68 23.67 23.65 23.64 23.61 23.60 23.59 23.59
.73 28.22 25.67 24.76 24.07 23.72 23.52 23.44 23.40 23.37 23.36 23.34 23.33 23.31 23.30 23.29 23.28
.74 27.86 25.33 24.44 23.75 23.41 23.21 23.13 23.09 23.07 23.05 23.04 23.02 23.00 22.99 22.98 22.98
.75 27.50 25.00 24.11 23.44 23.10 22.90 22.83 22.79 22.76 22.75 22.73 22.72 22.70 22.69 22.68 22.67
.76 27.13 24.67 23.79 23.12 22.79 22.60 22.52 22.48 22.46 22.44 22.43 22.42 22.40 22.39 22.38 22.37
.77 26.77 24.33 23.47 22.81 22.48 22.30 22.22 22.18 22.16 22.14 22.13 22.12 22.10 22.09 22.08 22.08
.78 26.39 24.00 23.15 22.50 22.18 21.99 21.92 21.89 21.86 21.85 21.84 21.82 21.80 21.79 21.78 21.78
.79 26.02 23.67 22.83 22.19 21.87 21.70 21.63 21.59 21.57 21.55 21.54 21.53 21.51 21.50 21.49 21.49
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

.80 25.64 23.33 22.51 21.88 21.57 21.40 21.33 21.29 21.27 21.26 21.25 21.23 21.22 21.21 21.20 21.20
.81 25.25 23.00 22.19 21.58 21.27 21.10 21.04 21.00 20.98 20.97 20.96 20.94 20.93 20.92 20.91 20.91
.82 24.86 22.67 21.87 21.27 20.98 20.81 20.75 20.71 20.69 20.68 20.67 20.65 20.64 20.63 20.62 20.62
.83 24.47 22.33 21.56 20.97 20.68 20.52 20.46 20.42 20.40 20.39 20.38 20.37 20.35 20.35 20.34 20.34
.84 24.07 22.00 21.24 20.67 20.39 20.23 20.17 20.14 20.12 20.11 20.10 20.09 20.07 20.06 20.06 20.05
.85 23.67 21.67 20.93 20.37 20.10 19.94 19.89 19.86 19.84 19.82 19.82 19.80 19.79 19.78 19.78 19.77
.86 23.26 21.33 20.62 20.07 19.81 19.66 19.60 19.57 19.56 19.54 19.54 19.53 19.51 19.51 19.50 19.50
.87 22.84 21.00 20.31 19.78 19.52 19.38 19.32 19.30 19.28 19.27 19.26 19.25 19.24 19.23 19.22 19.22
.88 22.42 20.67 20.00 19.48 19.23 19.10 19.04 19.02 19.00 18.99 18.98 18.98 18.96 18.96 18.95 18.95
.89 21.99 20.33 19.69 19.19 18.95 18.82 18.77 18.74 18.73 18.72 18.71 18.70 18.69 18.69 18.68 18.68
.90 21.55 20.00 19.36 18.90 18.67 18.54 18.50 18.47 18.46 18.45 18.44 18.43 18.42 18.42 18.41 18.41
.91 21.11 19.67 19.07 18.61 18.39 18.27 18.22 18.20 18.19 18.18 18.17 18.17 18.16 18.15 18.15 18.15
.92 20.66 19.33 18.77 18.33 18.11 18.00 17.96 17.94 17.92 17.92 17.91 17.90 17.89 17.89 17.88 17.88
.93 20.20 19.00 18.46 18.04 17.84 17.73 17.69 17.67 17.66 17.65 17.65 17.64 17.63 17.63 17.62 17.62
.94 19.74 18.67 18.16 17.76 17.57 17.46 17.43 17.41 17.40 17.39 17.39 17.38 17.37 17.37 17.36 17.36
.95 19.25 18.33 17.86 17.48 17.29 17.20 17.17 17.15 17.14 17.13 17.13 17.12 17.12 17.11 17.11 17.11
.96 18.76 18.00 17.56 17.20 17.03 16.94 16.91 16.89 16.88 16.88 16.87 16.87 16.86 16.86 16.86 16.85
.97 18.25 17.67 17.25 16.92 16.76 16.68 16.65 16.63 16.63 16.62 16.62 16.61 16.61 16.61 16.60 16.60
.98 17.74 17.33 16.96 16.65 16.49 16.42 16.39 16.38 16.37 16.37 16.37 16.36 16.36 16.36 16.36 16.36
.99 17.21 17.00 16.66 16.37 16.23 16.16 16.14 16.13 16.12 16.12 16.12 16.12 16.11 16.11 16.11 16.11
1.00 16.67 16.67 16.36 16.10 15.97 15.91 15.89 15.88 15.88 15.87 15.87 15.87 15.87 15.87 15.87 15.87
1.01 16.11 16.33 16.07 15.83 15.72 15.66 15.64 15.63 15.63 15.63 15.63 15.63 15.62 15.62 15.62 15.62
1.02 15.53 16.00 15.78 15.56 15.46 15.41 15.40 15.39 15.39 15.39 15.39 15.38 15.38 15.38 15.38 15.38
1.03 14.93 15.67 15.48 15.30 15.21 15.17 15.15 15.15 15.15 15.15 15.15 15.15 15.15 15.15 15.15 15.15
1.04 14.31 15.33 15.19 15.03 14.96 14.92 14.91 14.91 14.91 14.91 14.91 14.91 14.91 14.91 14.91 14.91
1.05 13.66 15.00 14.91 14.77 14.71 14.68 14.67 14.67 14.67 14.67 14.68 14.68 14.68 14.68 14.68 14.68
1.06 12.96 14.67 14.62 14.51 14.46 14.44 14.44 14.44 14.44 14.44 14.44 14.45 14.45 14.45 14.45 14.45
(Continued)
737
TABLE T12.7 (Continued)
738

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

1.07 12.27 14.33 14.33 14.26 14.22 14.20 14.20 14.21 14.21 14.21 14.21 14.22 14.22 14.22 14.22 14.23
1.08 11.51 14.00 14.05 14.00 13.97 13.97 13.97 13.98 13.98 13.98 13.99 13.99 13.99 14.00 14.00 14.00
1.09 10.71 13.67 13.76 13.75 13.73 13.74 13.74 13.75 13.75 13.76 13.76 13.77 13.77 13.77 13.78 13.78
1.10 9.84 13.33 13.48 13.49 13.50 13.51 13.52 13.52 13.53 13.54 13.54 13.54 13.55 13.55 13.56 13.56
1.11 8.89 13.00 13.20 13.25 13.26 13.28 13.29 13.30 13.31 13.31 13.32 13.32 13.33 13.34 13.34 13.34
1.12 7.82 12.67 12.93 13.00 13.03 13.05 13.07 13.08 13.09 13.10 13.10 13.11 13.12 13.12 13.12 13.13
1.13 6.60 12.33 12.65 12.75 12.80 12.83 12.85 12.86 12.87 12.88 12.89 12.89 12.90 12.91 12.91 12.92
1.14 5.08 12.00 12.37 12.51 12.57 12.61 12.63 12.65 12.66 12.67 12.67 12.68 12.69 12.70 12.70 12.70
1.15 0.29 11.67 12.10 12.27 12.34 12.39 12.42 12.44 12.45 12.46 12.46 12.47 12.48 12.49 12.49 12.50
1.16 0.00 11.33 11.83 12.03 12.12 12.18 12.21 12.22 12.24 12.25 12.25 12.26 12.28 12.28 12.29 12.29
1.17 0.00 11.00 11.56 11.79 11.90 11.96 12.00 12.02 12.03 12.04 12.05 12.06 12.07 12.08 12.08 12.09
1.18 0.00 10.67 11.29 11.56 11.68 11.75 11.79 11.81 11.82 11.84 11.84 11.85 11.87 11.88 11.88 11.89
1.19 0.00 10.33 11.02 11.33 11.46 11.54 11.58 11.61 11.62 11.63 11.64 11.65 11.67 11.68 11.69 11.69
1.20 0.00 10.00 10.76 11.10 11.24 11.34 11.38 11.41 11.42 11.43 11.44 11.46 11.47 11.48 11.49 11.49
1.21 0.00 9.67 10.50 10.87 11.03 11.13 11.18 11.21 11.22 11.24 11.25 11.26 11.28 11.29 11.30 11.30
1.22 0.00 9.33 10.23 10.65 10.82 10.93 10.98 11.01 11.03 11.04 11.05 11.07 11.09 11.09 11.10 11.11
1.23 0.00 9.00 9.97 10.42 10.61 10.73 10.78 10.81 10.84 10.85 10.86 10.88 10.90 10.91 10.91 10.92
1.24 0.00 8.67 9.72 10.20 10.41 10.53 10.59 10.62 10.64 10.66 10.67 10.69 10.71 10.72 10.73 10.73
1.25 0.00 8.33 9.46 9.98 10.21 10.34 10.40 10.43 10.46 10.47 10.48 10.50 10.52 10.53 10.54 10.55
1.26 0.00 8.00 9.21 9.77 10.00 10.15 10.21 10.25 10.27 10.29 10.30 10.32 10.34 10.35 10.36 10.37
1.27 0.00 7.67 8.96 9.55 9.81 9.96 10.02 10.06 10.09 10.10 10.12 10.13 10.16 10.17 10.18 10.19
1.28 0.00 7.33 8.71 9.34 9.61 9.77 9.84 9.88 9.90 9.92 9.94 9.95 9.98 9.99 10.00 10.01
1.29 0.00 7.00 8.46 9.13 9.42 9.58 9.65 9.70 9.72 9.74 9.76 9.78 9.80 9.82 9.83 9.83
1.30 0.00 6.67 8.21 8.93 9.22 9.40 9.48 9.52 9.55 9.57 9.58 9.60 9.63 9.64 9.65 9.66
1.31 0.00 6.33 7.97 8.72 9.03 9.22 9.30 9.34 9.37 9.39 9.41 9.43 9.46 9.47 9.48 9.49
1.32 0.00 6.00 7.73 8.52 8.85 9.04 9.12 9.17 9.20 9.22 9.24 9.26 9.29 9.30 9.31 9.32
1.33 0.00 5.67 7.49 8.32 8.66 8.86 8.95 9.00 9.03 9.05 9.07 9.09 9.12 9.13 9.15 9.15
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

1.34 0.00 5.33 7.25 8.12 8.48 8.69 8.78 8.83 8.86 8.88 8.90 8.92 8.95 8.97 8.98 8.99
1.35 0.00 5.00 7.02 7.92 8.30 8.52 8.61 8.66 8.69 8.72 8.74 8.76 8.79 8.81 8.82 8.83
1.36 0.00 4.67 6.79 7.73 8.12 8.35 8.44 8.50 8.53 8.55 8.57 8.60 8.63 8.65 8.66 8.67
1.37 0.00 4.33 6.56 7.54 7.95 8.18 8.28 8.33 8.37 8.39 8.41 8.44 8.47 8.49 8.50 8.51
1.38 0.00 4.00 6.33 7.35 7.77 8.01 8.12 8.17 8.21 8.24 8.25 8.29 8.31 8.33 8.35 8.35
1.39 0.00 3.67 6.10 7.17 7.60 7.85 7.96 8.01 8.05 8.08 8.10 8.12 8.16 8.18 8.19 8.20
1.40 0.00 3.33 5.88 6.98 7.44 7.69 7.80 7.86 7.90 7.92 7.94 7.97 8.01 8.02 8.04 8.05
1.41 0.00 3.00 5.66 6.80 7.27 7.53 7.64 7.70 7.74 7.77 7.79 7.82 7.86 7.87 7.89 7.90
1.42 0.00 2.67 5.44 6.62 7.10 7.37 7.49 7.55 7.59 7.62 7.64 7.67 7.71 7.73 7.74 7.75
1.43 0.00 2.33 5.23 6.45 6.94 7.22 7.34 7.40 7.44 7.47 7.50 7.52 7.56 7.58 7.60 7.61
1.44 0.00 2.00 5.01 6.27 6.78 7.07 7.19 7.26 7.30 7.33 7.35 7.38 7.42 7.44 7.46 7.47
1.45 0.00 1.67 4.81 6.10 6.63 6.92 7.04 7.11 7.15 7.18 7.21 7.24 7.28 7.30 7.31 7.33
1.46 0.00 1.33 4.60 5.93 6.47 6.77 6.90 6.97 7.01 7.04 7.07 7.10 7.14 7.16 7.18 7.19
1.47 0.00 1.00 4.39 5.77 6.32 6.63 6.75 6.83 6.87 6.90 6.93 6.96 7.00 7.02 7.04 7.05
1.48 0.00 .67 4.19 5.60 6.17 6.48 6.61 6.69 6.73 6.77 6.79 6.82 6.86 6.88 6.90 6.91
1.49 0.00 .33 3.99 5.44 6.02 6.34 6.48 6.55 6.60 6.63 6.65 6.69 6.73 6.75 6.77 6.78
1.50 0.00 0.00 3.80 5.28 5.87 6.20 6.34 6.41 6.46 6.50 6.52 6.55 6.60 6.62 6.64 6.65
1.51 0.00 0.00 3.61 5.13 5.73 6.06 6.20 6.28 6.33 6.36 6.39 6.42 6.47 6.49 6.51 6.52
1.52 0.00 0.00 3.42 4.97 5.59 5.93 6.07 6.15 6.20 6.23 6.26 6.29 6.34 6.36 6.38 6.39
1.53 0.00 0.00 3.23 4.82 5.45 5.80 5.94 6.02 6.07 6.11 6.13 6.17 6.21 6.24 6.26 6.27
1.54 0.00 0.00 3.05 4.67 5.31 5.67 5.81 5.89 5.95 5.98 6.01 6.04 6.09 6.11 6.13 6.15
1.55 0.00 0.00 2.87 4.52 5.18 5.54 5.69 5.77 5.82 5.86 5.88 5.92 9.97 5.99 6.01 6.02
1.56 0.00 0.00 2.69 4.38 5.05 5.41 5.56 5.65 5.70 5.74 5.76 5.80 5.85 5.87 5.89 5.90
1.57 0.00 0.00 2.52 4.24 4.92 5.29 5.44 5.53 5.58 5.62 5.64 5.68 5.73 5.75 5.78 5.79
1.58 0.00 0.00 2.35 4.10 4.79 5.16 5.32 5.41 5.46 5.50 5.53 5.56 5.61 5.64 5.66 5.67
1.59 0.00 0.00 2.19 3.96 4.66 5.04 5.20 5.29 5.34 5.38 5.41 5.45 5.50 5.52 5.54 5.56
1.60 0.00 0.00 2.03 3.83 4.54 4.92 5.09 5.17 5.23 5.27 5.30 5.33 5.38 5.41 5.43 5.44
(Continued)
739
TABLE T12.7 (Continued)
740

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

1.61 0.00 0.00 1.87 3.69 4.41 4.81 4.97 5.06 5.12 5.16 5.18 5.22 5.27 5.30 5.32 5.33
1.62 0.00 0.00 1.72 3.57 4.30 4.69 4.86 4.95 5.01 5.04 5.07 5.11 5.16 5.19 5.21 5.23
1.63 0.00 0.00 1.57 3.44 4.18 4.58 4.75 4.84 4.90 4.94 4.97 5.01 5.06 5.08 5.11 5.12
1.64 0.00 0.00 1.42 3.31 4.06 4.47 4.64 4.73 4.79 4.83 4.86 4.90 4.95 4.98 5.00 5.01
1.65 0.00 0.00 1.28 3.19 3.95 4.36 4.53 4.62 4.68 4.72 4.75 4.79 4.85 4.87 4.90 4.91
1.66 0.00 0.00 1.15 3.07 3.84 4.25 4.43 4.52 4.58 4.62 4.65 4.69 4.74 4.77 4.80 4.81
1.67 0.00 0.00 1.02 2.95 3.73 4.15 4.32 4.42 4.48 4.52 4.55 4.59 4.64 4.67 4.70 4.71
1.68 0.00 0.00 0.89 2.84 3.62 4.05 4.22 4.32 4.38 4.42 4.45 4.49 4.55 4.57 4.60 4.61
1.69 0.00 0.00 0.77 2.73 3.52 3.94 4.12 4.22 4.28 4.32 4.35 4.39 4.45 4.47 4.50 4.51
1.70 0.00 0.00 0.66 2.62 3.41 3.84 4.02 4.12 4.18 4.22 4.25 4.30 4.35 4.38 4.41 4.42
1.71 0.00 0.00 0.55 2.51 3.31 3.75 3.93 4.02 4.09 4.13 4.16 4.20 4.26 4.29 4.31 4.32
1.72 0.00 0.00 0.45 2.41 3.21 3.65 3.83 3.93 3.99 4.04 4.07 4.11 4.17 4.19 4.22 4.23
1.73 0.00 0.00 0.36 2.30 3.11 3.56 3.74 3.84 3.90 3.94 3.98 4.02 4.08 4.10 4.13 4.14
1.74 0.00 0.00 0.27 2.20 3.02 3.46 3.65 3.75 3.81 3.85 3.89 3.93 3.99 4.01 4.04 4.05
1.75 0.00 0.00 0.19 2.11 2.93 3.37 3.56 3.66 3.72 3.77 3.80 3.84 3.90 3.93 3.95 3.97
1.76 0.00 0.00 0.12 2.01 2.83 3.28 3.47 3.57 3.63 3.68 3.71 3.76 3.81 3.84 3.87 3.88
1.77 0.00 0.00 0.06 1.92 2.74 3.20 3.38 3.48 3.55 3.59 3.63 3.67 3.73 3.76 3.78 3.80
1.78 0.00 0.00 0.02 1.83 2.66 3.11 3.30 3.40 3.47 3.51 3.54 3.59 3.64 3.67 3.70 3.71
1.79 0.00 0.00 0.00 1.74 2.57 3.03 3.21 3.32 3.38 3.43 3.46 3.51 3.56 3.59 3.63 3.63
1.80 0.00 0.00 0.00 1.65 2.49 2.94 3.13 3.24 3.30 3.35 3.38 3.43 3.48 3.51 3.54 3.55
1.81 0.00 0.00 0.00 1.57 2.40 2.86 3.05 3.16 3.22 3.27 3.30 3.35 3.40 3.43 3.46 3.47
1.82 0.00 0.00 0.00 1.49 2.32 2.79 2.98 3.08 3.15 3.19 3.22 3.27 3.33 3.36 3.38 3.40
1.83 0.00 0.00 0.00 1.41 2.25 2.71 2.90 3.00 3.07 3.11 3.15 3.19 3.25 3.28 3.31 3.32
1.84 0.00 0.00 0.00 1.34 2.17 2.63 2.82 2.93 2.99 3.04 3.07 3.12 3.18 3.21 3.23 3.25
1.85 0.00 0.00 0.00 1.26 2.09 2.56 2.75 2.85 2.92 2.97 3.00 3.05 3.10 3.13 3.16 3.17
1.86 0.00 0.00 0.00 1.19 2.02 2.48 2.68 2.78 2.85 2.89 2.93 2.97 3.03 3.06 3.09 3.20
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

1.87 0.00 0.00 0.00 1.12 1.95 2.41 2.61 2.71 2.78 2.82 2.86 2.90 2.96 2.99 3.02 3.03
1.88 0.00 0.00 0.00 1.06 1.88 2.34 2.54 2.64 2.71 2.75 2.79 2.83 2.89 2.92 2.95 2.94
1.89 0.00 0.00 0.00 0.99 1.81 2.28 2.47 2.57 2.64 2.69 2.72 2.77 2.83 2.85 2.88 2.90
1.90 0.00 0.00 0.00 0.93 1.75 2.21 2.40 2.51 2.57 2.62 2.65 2.70 2.76 2.79 2.82 2.83
1.91 0.00 0.00 0.00 0.87 1.68 2.14 2.34 2.44 2.51 2.56 2.59 2.63 2.69 2.72 2.75 2.77
1.92 0.00 0.00 0.00 0.81 1.62 2.08 2.27 2.38 2.45 2.49 2.52 2.57 2.63 2.66 2.69 2.70
1.93 0.00 0.00 0.00 0.76 1.56 2.02 2.21 2.32 2.38 2.43 2.46 2.51 2.57 2.60 2.62 2.64
1.94 0.00 0.00 0.00 0.70 1.50 1.96 2.15 2.25 2.32 2.37 2.40 2.45 2.51 2.54 2.56 2.58
1.95 0.00 0.00 0.00 0.65 1.44 1.90 2.09 2.19 2.26 2.31 2.34 2.39 2.45 2.48 2.50 2.52
1.96 0.00 0.00 0.00 0.60 1.38 1.84 2.03 2.14 2.20 2.25 2.28 2.33 2.39 2.42 2.44 2.46
1.97 0.00 0.00 0.00 0.56 1.33 1.78 1.97 2.08 2.14 2.19 2.22 2.27 2.33 2.36 2.39 2.40
1.98 0.00 0.00 0.00 0.51 1.27 1.73 1.92 2.02 2.09 2.13 2.17 2.21 2.27 2.30 2.33 2.34
1.99 0.00 0.00 0.00 0.47 1.22 1.67 1.86 1.97 2.03 2.08 2.11 2.16 2.22 2.25 2.27 2.29
2.00 0.00 0.00 0.00 0.43 1.17 1.62 1.81 1.91 1.98 2.03 2.06 2.10 2.16 2.19 2.22 2.23
2.01 0.00 0.00 0.00 0.39 1.12 1.57 1.76 1.86 1.93 1.97 2.01 2.05 2.11 2.14 2.17 2.18
2.02 0.00 0.00 0.00 0.36 1.07 1.52 1.71 1.81 1.87 1.92 1.95 2.00 2.06 2.09 2.11 2.13
2.03 0.00 0.00 0.00 0.32 1.03 1.47 1.66 1.76 1.82 1.87 1.90 1.95 2.01 2.04 2.06 2.08
2.04 0.00 0.00 0.00 0.29 0.98 1.42 1.61 1.71 1.77 1.82 1.85 1.90 1.96 1.99 2.01 2.03
2.05 0.00 0.00 0.00 0.26 0.94 1.37 1.56 1.66 1.73 1.77 1.80 1.85 1.91 1.94 1.96 1.98
2.06 0.00 0.00 0.00 0.23 0.90 1.33 1.51 1.61 1.68 1.72 1.76 1.80 1.86 1.89 1.92 1.93
2.07 0.00 0.00 0.00 0.21 0.86 1.28 1.47 1.57 1.63 1.68 1.71 1.76 1.81 1.84 1.87 1.88
2.08 0.00 0.00 0.00 0.18 0.82 1.24 1.42 1.52 1.59 1.63 1.64 1.71 1.77 1.79 1.82 1.84
2.09 0.00 0.00 0.00 0.16 0.78 1.20 1.38 1.48 1.54 1.59 1.62 1.66 1.72 1.75 1.48 1.79
2.10 0.00 0.00 0.00 0.14 0.74 1.16 1.34 1.44 1.50 1.54 1.58 1.62 1.68 1.71 1.73 1.75
2.11 0.00 0.00 0.00 0.12 0.71 1.12 1.30 1.39 1.46 1.50 1.53 1.58 1.63 1.66 1.69 1.70
2.12 0.00 0.00 0.00 0.10 0.67 1.08 1.26 1.35 1.42 1.46 1.49 1.54 1.59 1.62 1.65 1.66
2.13 0.00 0.00 0.00 0.08 0.64 1.04 1.22 1.31 1.38 1.42 1.45 1.50 1.55 1.58 1.61 1.62
(Continued)
741
TABLE T12.7 (Continued)
742

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

2.14 0.00 0.00 0.00 0.07 0.61 1.00 1.18 1.28 1.34 1.38 1.41 1.46 1.51 1.54 1.57 1.58
2.15 0.00 0.00 0.00 0.06 0.58 0.97 1.14 1.24 1.30 1.34 1.37 1.42 1.47 1.50 1.53 1.54
2.16 0.00 0.00 0.00 0.05 0.55 0.93 1.10 1.20 1.26 1.30 1.34 1.38 1.43 1.46 `1.49 1.50
2.17 0.00 0.00 0.00 0.04 0.52 0.90 1.07 1.16 1.22 1.27 1.30 1.34 1.40 1.42 1.45 1.46
2.18 0.00 0.00 0.00 0.03 0.49 0.87 1.03 1.13 1.19 1.23 1.26 1.30 1.36 1.39 1.41 1.42
2.19 0.00 0.00 0.00 0.02 0.46 0.83 1.00 1.09 1.15 1.20 1.23 1.27 1.32 1.35 1.38 1.39
2.20 0.000 0.000 0.000 0.015 0.437 0.803 0.968 1.061 1.120 1.161 1.192 1.233 1.287 1.314 1.340 1.352
2.21 0.000 0.000 0.000 0.010 0.413 0.772 0.936 1.028 1.087 1.128 1.158 1.199 1.253 1.279 1.305 1.318
2.22 0.000 0.000 0.000 0.006 0.389 0.743 0.905 0.996 1.054 1.095 1.125 1.166 1.219 1.245 1.271 1.283
2.23 0.000 0.000 0.000 0.003 0.366 0.715 0.875 0.965 1.023 1.063 1.093 1.134 1.186 1.212 1.238 1.250
2.24 0.000 0.000 0.000 0.002 0.345 0.687 0.845 0.935 0.992 1.032 1.061 1.102 1.154 1.180 1.205 1.218
2.25 0.000 0.000 0.000 0.001 0.324 0.660 0.816 0.905 0.962 1.002 1.031 1.071 1.123 1.148 1.173 1.186
2.26 0.000 0.000 0.000 0.000 0.304 0.634 0.789 0.876 0.933 0.972 1.001 1.041 1.092 1.117 1.142 1.155
2.27 0.000 0.000 0.000 0.000 0.285 0.609 0.762 0.848 0.904 0.943 0.972 1.011 1.062 1.087 1.112 1.124
2.28 0.000 0.000 0.000 0.000 0.267 0.585 0.735 0.821 0.876 0.915 0.943 0.982 1.033 1.058 1.082 1.094
2.29 0.000 0.000 0.000 0.000 0.250 0.561 0.710 0.794 0.849 0.887 0.915 0.954 1.004 1.029 1.053 1.065
2.30 0.000 0.000 0.000 0.000 0.233 0.538 0.685 0.769 0.823 0.861 0.888 0.927 0.977 1.001 1.025 1.037
2.31 0.000 0.000 0.000 0.000 0.218 0.516 0.661 0.743 0.797 0.834 0.862 0.900 0.949 0.974 0.997 1.009
2.32 0.000 0.000 0.000 0.000 0.203 0.495 0.637 0.719 0.772 0.809 0.836 0.874 0.923 0.947 0.971 0.982
2.33 0.000 0.000 0.000 0.000 0.189 0.474 0.614 0.695 0.748 0.784 0.811 0.848 0.897 0.921 0.944 0.956
2.34 0.000 0.000 0.000 0.000 0.175 0.454 0.592 0.672 0.724 0.760 0.787 0.824 0.872 0.895 0.915 0.930
2.35 0.000 0.000 0.000 0.000 0.163 0.435 0.571 0.650 0.701 0.736 0.763 0.799 0.847 0.870 0.893 0.905
2.36 0.000 0.000 0.000 0.000 0.151 0.416 0.550 0.628 0.678 0.714 0.740 0.776 0.823 0.846 0.869 0.880
2.37 0.000 0.000 0.000 0.000 0.139 0.398 0.530 0.606 0.656 0.691 0.717 0.753 0.799 0.822 0.845 0.856
2.38 0.000 0.000 0.000 0.000 0.128 0.381 0.510 0.586 0.635 0.670 0.695 0.730 0.777 0.799 0.822 0.833
2.39 0.000 0.000 0.000 0.000 0.118 0.364 0.491 0.566 0.614 0.648 0.674 0.709 0.754 0.777 0.799 0.810
2.40 0.000 0.000 0.000 0.000 0.109 0.348 0.473 0.546 0.594 0.628 0.653 0.687 0.732 0.755 0.777 0.787
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

2.41 0.000 0.000 0.000 0.000 0.100 0.332 0.455 0.527 0.575 0.608 0.633 0.667 0.711 0.733 0.755 0.766
2.42 0.000 0.000 0.000 0.000 0.091 0.317 0.437 0.509 0.555 0.588 0.613 0.646 0.691 0.712 0.734 0.744
2.43 0.000 0.000 0.000 0.000 0.083 0.302 0.421 0.491 0.537 0.569 0.593 0.627 0.670 0.692 0.713 0.724
2.44 0.000 0.000 0.000 0.000 0.076 0.288 0.404 0.474 0.519 0.551 0.575 0.608 0.651 0.672 0.693 0.703
2.45 0.000 0.000 0.000 0.000 0.069 0.275 0.389 0.457 0.501 0.533 0.556 0.589 0.632 0.653 0.673 0.684
2.46 0.000 0.000 0.000 0.000 0.063 0.262 0.373 0.440 0.484 0.516 0.539 0.571 0.613 0.634 0.654 0.664
2.47 0.000 0.000 0.000 0.000 0.057 0.249 0.359 0.425 0.468 0.499 0.521 0.553 0.595 0.615 0.635 0.646
2.48 0.000 0.000 0.000 0.000 0.051 0.237 0.344 0.409 0.452 0.482 0.505 0.536 0.577 0.597 0.617 0.627
2.49 0.000 0.000 0.000 0.000 0.046 0.226 0.331 0.394 0.436 0.466 0.488 0.519 0.560 0.580 0.600 0.609
2.50 0.000 0.000 0.000 0.000 0.041 0.214 0.317 0.380 0.421 0.451 0.473 0.503 0.543 0.563 0.582 0.392
2.51 0.000 0.000 0.000 0.000 0.037 0.204 0.304 0.366 0.407 0.436 0.457 0.487 0.527 0.546 0.565 0.575
2.52 0.000 0.000 0.000 0.000 0.033 0.193 0.292 0.352 0.392 0.421 0.442 0.472 0.511 0.530 0.549 0.558
2.53 0.000 0.000 0.000 0.000 0.029 0.184 0.280 0.339 0.379 0.407 0.428 0.457 0.495 0.514 0.533 0.542
2.54 0.000 0.000 0.000 0.000 0.026 0.174 0.268 0.326 0.365 0.393 0.413 0.442 0.480 0.499 0.517 0.527
2.55 0.000 0.000 0.000 0.000 0.023 0.165 0.257 0.314 0.352 0.379 0.400 0.428 0.465 0.484 0.502 0.511
2.56 0.000 0.000 0.000 0.000 0.020 0.156 0.246 0.302 0.340 0.366 0.386 0.414 0.451 0.469 0.487 0.496
2.57 0.000 0.000 0.000 0.000 0.017 0.148 0.236 0.291 0.327 0.354 0.373 0.401 0.437 0.455 0.473 0.482
2.58 0.000 0.000 0.000 0.000 0.015 0.140 0.226 0.279 0.316 0.341 0.361 0.388 0.424 0.441 0.459 0.468
2.59 0.000 0.000 0.000 0.000 0.013 0.133 0.216 0.269 0.304 0.330 0.349 0.375 0.410 0.428 0.445 0.454
2.60 0.000 0.000 0.000 0.000 0.011 0.125 0.207 0.258 0.293 0.318 0.337 0.363 0.398 0.415 0.432 0.441
2.61 0.000 0.000 0.000 0.000 0.009 0.118 0.198 0.248 0.282 0.307 0.325 0.351 0.385 0.402 0.419 0.428
2.62 0.000 0.000 0.000 0.000 0.008 0.112 0.189 0.238 0.272 0.296 0.314 0.339 0.373 0.390 0.406 0.415
2.63 0.000 0.000 0.000 0.000 0.007 0.105 0.181 0.229 0.262 0.285 0.303 0.328 0.361 0.378 0.394 0.402
2.64 0.000 0.000 0.000 0.000 0.005 0.099 0.172 0.220 0.252 0.275 0.293 0.317 0.350 0.366 0.382 0.390
2.65 0.000 0.000 0.000 0.000 0.005 0.094 0.165 0.211 0.243 0.265 0.282 0.307 0.339 0.355 0.371 0.379
2.66 0.000 0.000 0.000 0.000 0.004 0.088 0.157 0.202 0.233 0.256 0.273 0.296 0.328 0.344 0.359 0.367
2.67 0.000 0.000 0.000 0.000 0.003 0.083 0.150 0.194 0.224 0.246 0.263 0.286 0.317 0.333 0.348 0.356
(Continued)
743
TABLE T12.7 (Continued)
744

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

2.68 0.000 0.000 0.000 0.000 0.002 0.078 0.143 0.186 0.216 0.237 0.254 0.277 0.307 0.322 0.338 0.345
2.69 0.000 0.000 0.000 0.000 0.002 0.073 0.136 0.179 0.208 0.229 0.245 0.267 0.297 0.312 0.327 0.335
2.70 0.000 0.000 0.000 0.000 0.001 0.069 0.130 0.171 0.200 0.220 0.236 0.258 0.288 0.302 0.317 0.325
2.71 0.000 0.000 0.000 0.000 0.001 0.064 0.124 0.164 0.192 0.212 0.227 0.249 0.278 0.293 0.307 0.315
2.72 0.000 0.000 0.000 0.000 0.000 0.060 0.118 0.157 0.184 0.204 0.219 0.241 0.269 0.283 0.298 0.305
2.73 0.000 0.000 0.000 0.000 0.000 0.057 0.112 0.151 0.177 0.197 0.211 0.232 0.260 0.274 0.288 0.296
2.74 0.000 0.000 0.000 0.000 0.000 0.053 0.107 0.144 0.170 0.189 0.204 0.224 0.252 0.266 0.279 0.286
2.75 0.000 0.000 0.000 0.000 0.000 0.049 0.102 0.138 0.163 0.182 0.196 0.216 0.243 0.247 0.271 0.277
2.76 0.000 0.000 0.000 0.000 0.000 0.046 0.097 0.132 0.157 0.175 0.189 0.209 0.235 0.249 0.262 0.269
2.77 0.000 0.000 0.000 0.000 0.000 0.043 0.092 0.126 0.151 0.168 0.182 0.201 0.227 0.241 0.254 0.260
2.78 0.000 0.000 0.000 0.000 0.000 0.040 0.087 0.121 0.145 0.162 0.175 0.194 0.220 0.233 0.246 0.252
2.79 0.000 0.000 0.000 0.000 0.000 0.037 0.083 0.115 0.139 0.156 0.169 0.187 0.212 0.225 0.238 0.244
2.80 0.000 0.000 0.000 0.000 0.000 0.035 0.079 0.110 0.133 0.150 0.162 0.181 0.205 0.218 0.230 0.237
2.81 0.000 0.000 0.000 0.000 0.000 0.032 0.075 0.105 0.128 0.144 0.156 0.174 0.198 0.211 0.223 0.229
2.82 0.000 0.000 0.000 0.000 0.000 0.030 0.071 0.101 0.122 0.138 0.150 0.168 0.192 0.204 0.216 0.222
2.83 0.000 0.000 0.000 0.000 0.000 0.028 0.067 0.096 0.117 0.133 0.145 0.162 0.185 0.197 0.209 0.215
2.84 0.000 0.000 0.000 0.000 0.000 0.026 0.064 0.092 0.112 0.128 0.139 0.156 0.179 0.190 0.202 0.208
2.85 0.000 0.000 0.000 0.000 0.000 0.024 0.060 0.088 0.108 0.122 0.134 0.150 0.173 0.184 0.195 0.201
2.86 0.000 0.000 0.000 0.000 0.000 0.022 0.057 0.084 0.103 0.118 0.129 0.145 0.167 0.178 0.189 0.195
2.87 0.000 0.000 0.000 0.000 0.000 0.020 0.054 0.080 0.099 0.113 0.124 0.139 0.161 0.172 0.183 0.188
2.88 0.000 0.000 0.000 0.000 0.000 0.019 0.051 0.076 0.094 0.108 0.119 0.134 0.155 0.166 0.177 0.182
2.89 0.000 0.000 0.000 0.000 0.000 0.017 0.048 0.073 0.090 0.104 0.114 0.129 0.150 0.160 0.171 0.176
2.90 0.000 0.000 0.000 0.000 0.000 0.016 0.046 0.069 0.087 0.100 0.110 0.125 0.145 0.155 0.165 0.171
2.91 0.000 0.000 0.000 0.000 0.000 0.015 0.043 0.066 0.083 0.096 0.106 0.120 0.140 0.150 0.160 0.165
2.92 0.000 0.000 0.000 0.000 0.000 0.013 0.041 0.063 0.079 0.092 0.101 0.115 0.135 0.145 0.155 0.160
2.93 0.000 0.000 0.000 0.000 0.000 0.012 0.038 0.060 0.076 0.088 0.097 0.111 0.130 0.140 0.149 0.154
2.94 0.000 0.000 0.000 0.000 0.000 0.011 0.036 0.057 0.072 0.084 0.093 0.107 0.125 0.135 0.144 0.149
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

2.95 0.000 0.000 0.000 0.000 0.000 0.010 0.034 0.054 0.069 0.081 0.090 0.103 0.121 0.130 0.140 0.144
2.96 0.000 0.000 0.000 0.000 0.000 0.009 0.032 0.051 0.066 0.077 0.086 0.099 0.117 0.126 0.135 0.140
2.97 0.000 0.000 0.000 0.000 0.000 0.009 0.030 0.049 0.063 0.074 0.083 0.095 0.112 0.121 0.130 0.135
2.98 0.000 0.000 0.000 0.000 0.000 0.008 0.028 0.046 0.060 0.071 0.079 0.091 0.108 0.117 0.126 0.130
2.99 0.000 0.000 0.000 0.000 0.000 0.007 0.027 0.044 0.057 0.068 0.076 0.088 0.104 0.113 0.122 0.126
3.00 0.000 0.000 0.000 0.000 0.000 0.006 0.025 0.042 0.055 0.065 0.073 0.084 0.101 0.109 0.118 0.122
3.01 0.000 0.000 0.000 0.000 0.000 0.006 0.024 0.040 0.052 0.062 0.070 0.081 0.097 0.105 0.114 0.118
3.02 0.000 0.000 0.000 0.000 0.000 0.005 0.022 0.038 0.050 0.059 0.067 0.078 0.093 0.101 0.110 0.114
3.03 0.000 0.000 0.000 0.000 0.000 0.005 0.021 0.036 0.048 0.057 0.064 0.075 0.090 0.098 0.106 0.110
3.04 0.000 0.000 0.000 0.000 0.000 0.004 0.019 0.034 0.045 0.054 0.061 0.072 0.087 0.094 0.102 0.106
3.05 0.000 0.000 0.000 0.000 0.000 0.004 0.018 0.032 0.043 0.052 0.059 0.069 0.083 0.091 0.099 0.103
3.06 0.000 0.000 0.000 0.000 0.000 0.003 0.017 0.030 0.041 0.050 0.056 0.066 0.080 0.088 0.095 0.099
3.07 0.000 0.000 0.000 0.000 0.000 0.003 0.016 0.029 0.039 0.047 0.054 0.064 0.077 0.085 0.092 0.096
3.08 0.000 0.000 0.000 0.000 0.000 0.003 0.015 0.027 0.037 0.045 0.052 0.061 0.074 0.081 0.089 0.092
3.09 0.000 0.000 0.000 0.000 0.000 0.002 0.014 0.026 0.036 0.043 0.049 0.059 0.072 0.079 0.086 0.089
3.10 0.000 0.000 0.000 0.000 0.000 0.002 0.013 0.024 0.034 0.041 0.047 0.056 0.069 0.076 0.083 0.086
3.11 0.000 0.000 0.000 0.000 0.000 0.002 0.012 0.023 0.032 0.039 0.045 0.054 0.066 0.073 0.080 0.083
3.12 0.000 0.000 0.000 0.000 0.000 0.002 0.011 0.022 0.031 0.038 0.043 0.052 0.064 0.070 0.077 0.080
3.13 0.000 0.000 0.000 0.000 0.000 0.002 0.011 0.021 0.029 0.036 0.041 0.050 0.061 0.068 0.074 0.077
3.14 0.000 0.000 0.000 0.000 0.000 0.001 0.010 0.019 0.028 0.034 0.040 0.048 0.059 0.065 0.071 0.075
3.15 0.000 0.000 0.000 0.000 0.000 0.001 0.009 0.018 0.026 0.033 0.038 0.046 0.057 0.063 0.069 0.072
3.16 0.000 0.000 0.000 0.000 0.000 0.001 0.009 0.017 0.025 0.031 0.036 0.044 0.055 0.060 0.066 0.069
3.17 0.000 0.000 0.000 0.000 0.000 0.001 0.008 0.016 0.024 0.030 0.035 0.042 0.053 0.058 0.064 0.067
3.18 0.000 0.000 0.000 0.000 0.000 0.001 0.007 0.015 0.022 0.028 0.033 0.040 0.050 0.056 0.062 0.065
3.19 0.000 0.000 0.000 0.000 0.000 0.001 0.007 0.015 0.021 0.027 0.032 0.038 0.049 0.054 0.059 0.062
3.20 0.000 0.000 0.000 0.000 0.000 0.001 0.006 0.014 0.020 0.026 0.030 0.037 0.047 0.052 0.057 0.060
3.21 0.000 0.000 0.000 0.000 0.000 0.000 0.006 0.013 0.019 0.024 0.029 0.035 0.045 0.050 0.055 0.058
(Continued)
745
TABLE T12.7 (Continued)
746

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

3.22 0.000 0.000 0.000 0.000 0.000 0.000 0.005 0.012 0.018 0.023 0.027 0.034 0.043 0.048 0.053 0.056
3.23 0.000 0.000 0.000 0.000 0.000 0.000 0.005 0.011 0.017 0.022 0.026 0.032 0.041 0.046 0.051 0.054
3.24 0.000 0.000 0.000 0.000 0.000 0.000 0.005 0.011 0.016 0.021 0.025 0.031 0.040 0.044 0.049 0.052
3.25 0.000 0.000 0.000 0.000 0.000 0.000 0.004 0.010 0.015 0.020 0.024 0.030 0.038 0.043 0.048 0.050
3.26 0.000 0.000 0.000 0.000 0.000 0.000 0.004 0.009 0.015 0.019 0.023 0.028 0.037 0.041 0.046 0.048
3.27 0.000 0.000 0.000 0.000 0.000 0.000 0.004 0.009 0.014 0.019 0.022 0.027 0.035 0.040 0.044 0.046
3.28 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.008 0.013 0.017 0.021 0.026 0.034 0.038 0.042 0.045
3.29 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.008 0.012 0.016 0.020 0.025 0.032 0.037 0.041 0.043
3.30 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.007 0.012 0.015 0.019 0.024 0.031 0.035 0.039 0.042
3.31 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.007 0.011 0.015 0.018 0.023 0.030 0.034 0.038 0.040
3.32 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.006 0.010 0.014 0.017 0.022 0.029 0.032 0.036 0.039
3.33 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.006 0.010 0.013 0.016 0.021 0.027 0.031 0.035 0.037
3.34 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.006 0.009 0.013 0.015 0.020 0.026 0.030 0.034 0.036
3.35 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.005 0.009 0.012 0.015 0.019 0.025 0.029 0.032 0.034
3.36 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.005 0.008 0.011 0.014 0.018 0.024 0.028 0.031 0.033
3.37 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.005 0.008 0.011 0.013 0.017 0.023 0.026 0.030 0.032
3.38 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.004 0.007 0.010 0.013 0.016 0.022 0.025 0.029 0.031
3.39 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.004 0.007 0.010 0.012 0.016 0.021 0.024 0.028 0.029
3.40 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.004 0.007 0.009 0.011 0.015 0.020 0.023 0.027 0.028
3.41 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.006 0.009 0.011 0.014 0.020 0.022 0.026 0.027
3.42 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.006 0.008 0.010 0.014 0.019 0.022 0.025 0.026
3.43 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.005 0.008 0.010 0.013 0.018 0.021 0.024 0.025
3.44 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.005 0.007 0.009 0.012 0.017 0.020 0.023 0.024
3.45 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.005 0.007 0.009 0.012 0.016 0.019 0.022 0.023
3.46 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.005 0.007 0.008 0.11 0.016 0.018 0.021 0.022
3.47 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.004 0.006 0.008 0.011 0.015 0.017 0.020 0.022
3.48 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.004 0.006 0.007 0.010 0.014 0.017 0.019 0.021
(Continued)
Appendix
TABLE T12.7 (Continued)
MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
Appendix

QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

3.49 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.004 0.005 0.007 0.010 0.014 0.016 0.019 0.020
3.50 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.003 0.005 0.007 0.009 0.013 0.015 0.018 0.019
3.51 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.003 0.005 0.006 0.009 0.013 0.015 0.017 0.018
3.52 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.003 0.005 0.006 0.008 0.012 0.014 0.017 0.018
3.53 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.004 0.006 0.008 0.012 0.014 0.016 0.017
3.54 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.004 0.005 0.008 0.011 0.013 0.015 0.016
3.55 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.003 0.004 0.005 0.007 0.011 0.012 0.015 0.016
3.56 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.004 0.005 0.007 0.010 0.012 0.014 0.015
3.57 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.005 0.006 0.010 0.011 0.013 0.014
3.58 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.006 0.009 0.011 0.013 0.014
3.59 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.006 0.009 0.010 0.012 0.013
3.60 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.006 0.008 0.010 0.012 0.013
3.61 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.005 0.008 0.010 0.011 0.012
3.62 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.003 0.005 0.008 0.009 0.011 0.012
3.63 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.005 0.007 0.009 0.010 0.011
3.64 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.007 0.008 0.010 0.011
3.65 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.007 0.008 0.010 0.010
3.66 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.006 0.008 0.009 0.010
3.67 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.003 0.004 0.006 0.007 0.009 0.010
3.68 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.002 0.004 0.006 0.007 0.008 0.009
3.69 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.002 0.003 0.005 0.007 0.008 0.009
3.70 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.002 0.003 0.005 0.006 0.008 0.008
3.71 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.005 0.006 0.007 0.008
3.72 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.005 0.006 0.007 0.008
3.73 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.005 0.006 0.007 0.007
3.74 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.005 0.007 0.007
3.75 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.002 0.004 0.005 0.006 0.007
(Continued)
747
 748

TABLE T12.7 (Continued)

MIL-STD-414 Table B-5—Table for Estimating the Lot Percent Defective Using Standard Deviation Methoda
Sample Size
QU or
QL 3 4 5 7 10 15 20 25 30 35 40 50 75 100 150 200

3.76 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.002 0.004 0.005 0.006 0.007
3.77 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.002 0.004 0.005 0.006 0.006
3.78 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.01 0.002 0.004 0.004 0.005 0.006
3.79 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.005 0.006
3.80 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.005 0.006
3.81 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.005 0.005
3.82 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.005 0.005
3.83 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.004 0.005
3.84 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.003 0.003 0.004 0.005
3.85 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.001 0.002 0.003 0.004 0.004
3.86 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.004
3.87 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.004
3.88 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.004 0.004
3.89 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.003 0.004
3.90 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001 0.002 0.003 0.003 0.004
a Values tabulated are read in percent.
Appendix
TABLE T12.8
MID-STD-414 Table B-1—Master Table for Normal and Tightened Inspection for Plans Based on Variability Unknown: Standard Deviation
Appendix

Method (Single-Specification Limit, Form 1)

Sample Acceptable Quality Levels (Normal Inspection)

Size .04 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
Code Sample
Letter Size k k k k k k k k k k k k k k

B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.12 .958 .765 .566 .341

C 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.45 1.34 1.17 1.01 .814 .617 .393
D 5 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.65 1.53 1.40 1.24 1.07 .874 .675 .455
E 7 ⇩ ⇩ ⇩ ⇩ 2.00 1.88 1.75 1.62 1.50 1.33 1.15 .955 .755 .536
F 10 ⇩ ⇩ ⇩ 2.24 2.11 1.98 1.84 1.72 1.58 1.41 1.23 1.03 .828 .611
G 15 2.64 2.53 2.42 2.32 2.20 2.06 1.91 1.79 1.65 1.47 1.30 1.09 .886 .664
H 20 2.69 2.58 2.47 2.36 2.24 2.11 1.96 1.82 1.69 1.51 1.33 1.12 .917 .695
I 25 2.72 2.61 2.50 2.40 2.26 2.14 1.98 1.85 1.72 1.53 1.35 1.14 .936 .712
J 30 2.73 2.61 2.51 2.41 2.28 2.15 2.00 1.86 1.73 1.55 1.36 1.15 .946 .723
K 35 2.77 2.65 2.54 2.45 2.31 2.18 2.03 1.89 1.76 1.57 1.39 1.18 .969 .745
L 40 2.77 2.66 2.55 2.44 2.31 2.18 2.03 1.89 1.76 1.58 1.39 1.18 .971 .746
M 50 2.83 2.71 2.60 2.50 2.35 2.22 2.08 1.93 1.80 1.61 1.42 1.21 1.00 .774
N 75 2.90 2.77 2.66 2.55 2.41 2.27 2.12 1.98 1.84 1.65 1.46 1.24 1.03 .804
O 100 2.92 2.80 2.69 2.58 2.43 2.29 2.14 2.00 1.86 1.67 1.48 1.26 1.05 .819
P 150 2.96 2.84 2.73 2.61 2.47 2.33 2.18 2.03 1.89 1.70 1.51 1.29 1.07 .841
Q 200 2.97 2.85 2.73 2.62 2.47 2.33 2.18 2.04 1.89 1.70 1.51 1.29 1.07 .845
.065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00
Acceptability Quality Levels (tightened inspection)

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC, 1957, p. 39.
All AQL values are in percent defective.
⇩ Use first sampling plan below arrow, that is, both sample size as well as k value. When sample size equals or exceeds lot size, every item in the lot must be
inspected.
749
 750

TABLE T12.9
MIL-STD-414 Table B-2—Master Table for Reduced Inspection for Plans Based on Variability Unknown: Standard Deviation Method
(Single-Specification Limit, Form 1)
Acceptable Quality Levels
Sample .04 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00
Size Code Sample
Letter Size k k k k k k k k k k k k k

B 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.12 .958 .765 .566 .341

C 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.12 .958 .765 .566 .341
D 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.12 .958 .765 .566 .341
E 3 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.12 .958 .765 .566 .341
F 4 ⇩ ⇩ ⇩ ⇩ ⇩ ⇩ 1.45 1.34 1.17 1.01 .814 .617 .393
G 5 ⇩ ⇩ ⇩ ⇩ ⇩ 1.65 1.53 1.40 1.24 1.07 .874 .675 .455
H 7 ⇩ ⇩ ⇩ 2.00 1.88 1.75 1.62 1.50 1.33 1.15 .955 .755 .536
I 10 ⇩ ⇩ 2.24 2.11 1.98 1.84 1.72 1.58 1.41 1.23 1.03 .828 .611
J 10 ⇩ ⇩ 2.24 2.11 1.98 1.84 1.72 1.58 1.41 1.23 1.03 .828 .611
K 15 2.53 2.42 2.32 2.20 2.06 1.91 1.79 1.65 1.47 1.30 1.09 .886 .664
L 20 2.58 2.47 2.36 2.24 2.11 1.96 1.82 1.69 1.51 1.33 1.12 .917 .695
M 20 2.58 2.47 2.36 2.24 2.11 1.96 1.82 1.69 1.51 1.33 1.12 .917 .695
N 25 2.61 2.50 2.40 2.26 2.14 1.98 1.85 1.72 1.53 1.35 1.14 .936 .712
O 30 2.61 2.51 2.41 2.28 2.15 2.00 1.86 1.73 1.55 1.36 1.15 .946 .723
P 50 2.71 2.60 2.50 2.35 2.22 2.08 1.93 1.80 1.61 1.42 1.21 1.00 .774
Q 75 2.77 2.66 2.55 2.41 2.27 2.12 1.98 1.84 1.65 1.46 1.24 1.03 .804

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC, 1957, p. 39.
All AQL values are in percent defective.
⇩ Use first sampling plan below arrow, that is, both sample size as well as k value. When sample size equals or exceeds lot size, every item in the lot must be
inspected.
Appendix
TABLE T12.10
Appendix

MIL-STD-414 TABLE B-8—Values of F for Maximum Standard Deviation (MSD)

Sample Acceptable Quality Levels (in Percent Defective)
Size Code Sample
Letter Size .04 .065 .10 .15 .25 .40 .65 1.00 1.50 2.50 4.00 6.50 10.00 15.00

B 3 .436 .453 .475 .502 .538

C 4 .339 .353 .374 .399 .432 .472 .528
D 5 .294 .308 .323 .346 .372 .408 .452 .511
E 7 .242 .253 .266 .280 .295 .318 .345 .381 .425 .485
F 10 .214 .224 .235 .248 .261 .276 .298 .324 .359 .403 .460
G 15 .182 .188 .195 .202 .211 .222 .235 .248 .262 .284 .309 .344 .386 .442
H 20 .177 .183 .190 .197 .206 .216 .229 .242 .255 .277 .302 .336 .377 .432
I 25 .174 .180 .187 .193 .203 .212 .225 .238 .251 .273 .297 .331 .372 .426
J 30 .173 .179 .185 .192 .201 .210 .223 .236 .249 .270 .295 .328 .369 .423
K 35 .170 .176 .183 .189 .198 .208 .220 .232 .245 .266 .291 .323 .364 .416
L 40 .169 .176 .182 .188 .198 .207 .219 .232 .245 .266 .290 .323 .363 .416
M 50 .166 .172 .178 .184 .194 .203 .214 .227 .241 .261 .284 .317 .356 .408
N 75 .162 .168 .174 .181 .189 .199 .211 .223 .235 .255 .279 .310 .348 .399
O 100 .160 .166 .172 .179 .187 .197 .208 .220 .233 .253 .276 .307 .345 .395
P 150 .158 .163 .170 .175 .185 .193 .206 .216 .230 .249 .271 .302 .341 .388
Q 200 .157 .163 .168 .175 .183 .193 .203 .215 .228 .248 .269 .302 .338 .386

Source: United States Department of Defense, Military Standard, Sampling Procedures and Tables for Inspection by Variables for Percent Defective, MIL-STD-414, U.S.
Government Printing Office, Washington, DC, 1957, p. 39.
Notes:  The MSD may be obtained by multiplying the factor F by the difference between the upper specification limit U and lower specification limit L. The formula is
MSD = F(U − L). The MSD serves as a guide for the magnitude of the estimate of lot standard deviation when using plans for the double specification limit
case, based on the estimate of lot standard deviation of unknown variability. The estimate of lot standard deviation, if it is less than the MSD, helps to insure,
but does not guarantee, lot acceptability.
There is a corresponding acceptability constant in Table B-1 for each value of F. For reduced inspection, find the acceptability constant of Table B-2 in Table B-1 and use
the corresponding value of F.
751
752 Appendix

TABLE T13.1
Values of Plotting Positions (pi) to be Used in Plotting on Normal Probability Paper
for the No-Calc Procedure
n p1 p2 p3 p4 p5 p6 p7 p8 p9 p10

2 18.775
3 14.020 50.000
4 10.982 38.288
5 8.940 31.271 50.000
6 7.490 26.485 42.231
7 6.416 22.979 36.620 50.000
8 5.592 20.290 32.350 44.140
9 4.942 18.159 28.979 39.537 50.000
10 4.419 16.426 26.245 35.816 45.282
11 3.988 14.990 23.980 32.740 41.392 50.000
12 3.629 13.779 22.073 30.151 38.125 46.047
13 3.326 12.746 20.444 27.941 35.339 42.682 50.000
14 3.066 11.853 19.036 26.032 32.933 39.779 46.596
15 2.841 11.075 17.807 24.365 30.834 37.248 43.631 50.000
16 2.645 10.390 16.724 22.897 28.985 35.021 41.024 47.010
17 2.473 9.783 15.764 21.595 27.345 33.045 38.712 44.361 50.000
18 2.321 9.241 14.906 20.431 25.879 31.279 36.648 41.996 47.333
19 2.185 8.7545 14.136 19.384 24.561 29.692 34.793 39.872 44.939 50.000
20 2.063 8.3158 13.439 18.438 23.370 28.258 33.110 37.962 42.779 47.589
Source: Reprinted from Chernoff, H. and Lieberman, G.J., Ind. Qual. Control, 13(7), 5, 1957. With
permission.
Note: When i > n/2 use pi = 100 – pn – i + 1. For n > 20 use pi = (2i – 1)/2n.
Appendix 753

TABLE T13.2
Values of Maximum Estimated Percentage Defective Allowing Acceptance
of the Lot (p*)
AQL
Code Sample
Letter Size 0.40 0.65 1.00 1.50 2.50 4.00 6.50 10.00 15.00

B 3 10.24 13.95 19.35 26.12 35.02

C 4 5.70 7.33 10.12 13.69 18.81 25.13 33.46
D 5 3.97 5.21 6.76 9.38 12.76 17.60 23.60 31.41
E 7 2.49 3.42 4.54 5.93 8.33 11.45 15.96 21.56 28.81
F 10 2.11 2.92 3.88 5.19 7.36 10.23 14.43 19.71 26.48
G 15 1.78 2.56 3.44 4.60 6.66 9.34 13.33 18.36 24.94
H 20 1.62 2.37 3.23 4.31 6.26 8.86 12.74 17.63 24.06
I 25 1.56 2.27 3.11 4.18 6.09 8.63 12.44 17.26 23.65
J 30 1.50 2.19 3.01 4.07 5.94 8.46 12.23 17.00 23.29
K 35 1.40 2.03 2.83 3.82 5.63 8.08 11.74 16.44 22.65
L 40 1.40 2.02 2.83 3.81 5.62 8.05 11.72 16.40 22.61
M 50 1.27 1.81 2.58 3.51 5.22 7.57 11.12 15.69 21.78
N 75 1.13 1.66 2.33 3.23 4.86 7.11 10.54 14.99 20.95
O 100 1.06 1.57 2.22 3.09 4.68 6.87 10.24 14.63 20.52
P 150 0.97 1.45 2.07 2.89 4.42 6.54 9.82 14.11 19.92
Q 200 0.96 1.43 2.05 2.87 4.38 6.49 9.76 14.04 19.84

Source: Reprinted from Chernoff, H. and Lieberman, G.J., Ind. Qual. Control, 13(7), 5, 1957.
With permission.
754 Appendix

TABLE T13.3
Matched Attributes Narrow Limit, Known (σ) and Unknown (s) Standard Deviation Variables
Plans for Values of p1 and p2 with α = .05, β = .10
Attributes NL-Gauge Variables
p1 p2 n c n c t nσ nS k

.001 .0015 40071 50 855 440 3.06 572 3180 3.02

.002 11729 17 285 135 2.91 191 1032 2.97
.0025 6114 10 160 80 2.94 107 567 2.93
.003 3888 7 110 55 2.91 74 381 2.90
.004 1976 4 67 35 2.92 45 226 2.84
.005 1319 3 49 23 2.75 33 160 2.80
.006 1099 3 38 19 2.80 26 124 2.77
.007 749 2 32 14 2.62 22 102 2.73
.008 655 2 28 14 2.75 19 87 2.71
.009 582 2 25 12 2.68 17 76 2.68
.01 524 2 22 11 2.72 15 67 2.66
.012 318 1 19 8 2.49 13 55 2.62
.015 254 1 15 7 2.58 11 44 2.57
.02 190 1 12 6 2.61 8 34 2.51
.025 152 1 10 5 2.58 7 27 2.46
.03 127 1 9 3 2.13 6 23 2.41
.035 108 1 8 4 2.52 6 20 2.37
.04 56 0 7 3 2.34 5 18 2.34
.05 44 0 6 3 2.48 5 15 2.28
.06 37 0 5 2 2.23 4 13 2.23
.0025 .004 11467 37 533 268 2.73 357 1678 2.72
.005 4689 17 240 113 2.61 161 736 2.68
.006 2743 11 148 71 2.60 99 443 2.64
.0075 1554 7 91 44 2.57 62 267 2.60
.01 789 4 56 28 2.56 38 157 2.54
.012 549 3 42 20 2.47 29 117 2.50
.015 439 3 32 16 2.49 22 85 2.45
.02 261 2 23 12 2.49 16 59 2.38
.025 209 2 18 9 2.40 12 45 2.33
.03 127 1 15 7 2.29 10 37 2.29
.035 108 1 13 6 2.25 9 31 2.25
.04 95 1 12 7 2.53 8 27 2.21
.05 76 1 10 3 1.77 7 22 2.15
.06 63 1 8 4 2.26 6 18 2.10
.005 .0075 8011 50 622 314 2.51 417 1714 2.50
.01 2343 17 206 111 2.54 138 547 2.44
.012 1370 11 128 62 2.37 85 327 2.40
.015 776 7 78 41 2.43 53 196 2.35
.02 394 4 47 22 2.23 32 114 2.28
.025 263 3 34 17 2.27 23 79 2.23
.03 219 3 27 14 2.28 18 61 2.19
.035 149 2 22 10 2.09 15 49 2.15
(Continued)
Appendix 755

TABLE T13.3 (Continued)

Matched Attributes Narrow Limit, Known (σ) and Unknown (s) Standard Deviation Variables
Plans for Values of p1 and p2 with α = .05, β = .10
Attributes NL-Gauge Variables
p1 p2 n c n c t nσ nS k

.04 130 2 19 9 2.11 13 41 2.11

.05 104 2 15 8 2.22 10 31 2.05
.06 63 1 12 6 2.11 9 25 2.00
.07 54 1 11 4 1.73 8 21 1.96
.0075 .01 11158 98 1137 571 2.38 763 2909 2.37
.012 3820 37 420 212 2.35 279 1040 2.33
.015 1561 17 186 98 2.36 125 450 2.29
.02 703 9 90 43 2.18 60 208 2.22
.025 416 6 58 29 2.19 39 129 2.17
.03 262 4 43 23 2.24 29 92 2.12
.035 187 3 33 15 2.01 23 71 2.08
.04 164 3 28 12 1.92 19 58 2.05
.05 104 2 21 10 1.99 14 42 1.99
.06 86 2 17 8 1.94 12 33 1.94
.07 74 2 14 6 1.81 10 27 1.90
.08 47 1 12 6 1.96 9 23 1.86
.01 .015 4003 50 525 258 2.22 351 1231 2.24
.02 1170 17 173 85 2.16 116 388 2.17
.025 609 10 96 52 2.24 64 208 2.12
.03 387 7 65 31 2.04 44 137 2.08
.035 261 5 49 24 2.04 33 100 2.04
.04 196 4 39 20 2.07 26 78 2.00
.045 174 4 32 17 2.09 22 64 1.97
.05 131 3 28 14 1.99 19 54 1.94
.06 109 3 22 11 1.95 15 41 1.89
.07 74 2 18 8 1.78 12 33 1.85
.08 64 2 15 7 1.81 11 27 1.81
.09 57 2 13 6 1.78 9 23 1.77
.10 37 1 12 6 1.84 8 20 1.74
.015 .02 5576 98 945 470 2.10 633 2036 2.11
.025 1566 31 290 143 2.04 195 603 2.05
.03 779 17 154 79 2.05 103 309 2.01
.035 468 11 101 52 2.02 67 197 1.97
.04 320 8 73 37 1.97 49 140 1.93
.045 257 7 57 29 1.95 39 107 1.90
.05 207 6 47 24 1.93 32 86 1.88
.06 130 4 34 16 1.79 23 61 1.82
.07 93 3 27 13 1.78 18 46 1.78
.08 81 3 22 11 1.80 15 37 1.74
.09 57 2 19 8 1.57 13 31 1.70
.10 51 2 16 8 1.75 11 26 1.67
.11 47 2 14 7 1.73 10 23 1.64
(Continued)
756 Appendix

TABLE T13.3 (Continued)

Matched Attributes Narrow Limit, Known (σ) and Unknown (s) Standard Deviation Variables
Plans for Values of p1 and p2 with α = .05, β = .10
Attributes NL-Gauge Variables
p1 p2 n c n c t nσ nS k

.12 43 2 13 6 1.61 9 20 1.61

.13 39 2 12 5 1.48 8 18 1.58
.14 36 2 11 4 1.33 8 16 1.56
.15 25 1 10 4 1.41 7 15 1.53
.02 .03 1963 49 429 223 2.01 287 835 1.96
.035 958 26 219 110 1.93 147 416 1.92
.04 584 17 140 67 1.84 94 259 1.88
.045 390 12 100 49 1.84 67 182 1.85
.05 304 10 77 40 1.89 52 137 1.82
.06 193 7 52 26 1.80 35 89 1.77
.07 130 5 39 19 1.73 26 64 1.73
.08 97 4 31 15 1.69 21 50 1.69
.09 86 4 26 13 1.70 17 40 1.65
.10 65 3 22 12 1.79 15 34 1.62
.11 59 3 19 9 1.59 13 29 1.59
.12 43 2 17 8 1.56 12 25 1.56
.13 39 2 15 7 1.53 10 22 1.53
.15 34 2 12 5 1.38 9 18 1.48
.17 30 2 11 4 1.21 8 15 1.44
.20 18 1 9 3 1.09 6 12 1.37
.03 .04 2732 96 756 378 1.81 506 1333 1.81
.045 1283 48 372 177 1.72 250 643 1.78
.05 781 31 230 110 1.70 154 389 1.75
.06 388 17 121 61 1.72 81 197 1.70
.07 233 11 78 39 1.67 53 124 1.65
.08 159 8 57 29 1.66 38 88 1.61
.09 128 7 44 21 1.55 30 66 1.58
.10 90 5 36 18 1.58 24 53 1.54
.11 82 5 30 16 1.64 20 43 1.51
.12 64 4 26 12 1.44 18 37 1.48
.13 59 4 23 10 1.35 16 32 1.46
.15 43 3 18 9 1.48 13 24 1.41
.20 25 2 12 6 1.40 8 15 1.30
.25 20 2 9 5 1.48 6 11 1.20
.30 12 1 7 4 1.48 5 8 1.12
.04 .06 961 48 334 165 1.63 224 524 1.64
.07 462 25 170 84 1.59 114 258 1.60
.08 276 16 108 54 1.57 72 159 1.56
.09 194 12 77 37 1.49 51 110 1.52
.10 139 9 58 29 1.51 39 82 1.49
.11 115 8 47 23 1.46 32 65 1.46
.12 85 6 39 19 1.43 26 53 1.43
(Continued)
Appendix 757

TABLE T13.3 (Continued)

Matched Attributes Narrow Limit, Known (σ) and Unknown (s) Standard Deviation Variables
Plans for Values of p1 and p2 with α = .05, β = .10
Attributes NL-Gauge Variables
p1 p2 n c n c t nσ nS k

.13 78 6 33 17 1.48 22 44 1.40

.14 64 5 28 14 1.42 20 38 1.37
.15 51 4 25 12 1.35 17 33 1.35
.17 45 4 20 10 1.37 14 25 1.30
.20 32 3 16 7 1.16 11 19 1.24
.25 20 2 11 5 1.15 8 13 1.15
.30 16 2 9 3 .78 6 9 1.06
.35 14 2 7 4 1.34 5 7 .98
.40 9 1 6 2 .69 4 6 .91
.05 .07 1131 68 448 227 1.57 300 660 1.55
.08 542 35 222 106 1.46 149 319 1.51
.09 333 23 138 71 1.52 93 194 1.47
.10 220 16 97 46 1.39 65 133 1.44
.11 158 12 73 36 1.41 49 98 1.41
.12 125 10 58 28 1.36 39 76 1.38
.13 97 8 48 23 1.33 32 62 1.35
.14 81 7 40 20 1.36 27 51 1.33
.15 68 6 35 19 1.45 24 43 1.30
.16 63 6 30 14 1.24 21 37 1.28
.17 52 5 27 13 1.26 18 33 1.26
.20 38 4 20 10 1.26 14 23 1.19
.25 25 3 14 8 1.37 10 15 1.10
.30 16 2 10 5 1.14 7 11 1.01
.35 14 2 8 4 1.09 6 8 .94
.40 12 2 7 2 .51 5 7 .86

Source: Reprinted from Schilling, E.G. and Sommers, D.J., J. Qual. Technol., 13(2), 84, 1981, Table 1. With
permission.
 758

TABLE T13.4
Tightened Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Acceptable Quality Levels (Tightened Inspection)
Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

n
t
A Ac
Re
n 3
t 0.00
B Ac 0
Re 1
n 3
t 1.19
C Ac 1
Re 2
n 4 6
t 1.14 1.07
D Ac 1 3
Re 2 4
n 5 8 10
t 1.20 1.67 0.98
E Ac 1 5 5
Re 2 6 6
n 5 9 12 14
t 1.92 1.43 1.04 1.02
F Ac 2 4 5 7
Re 3 5 6 8
(Continued)
Appendix
TABLE T13.4 (Continued)
Appendix

Tightened Inspection Optimal Narrow Limit Plans for MIL-STD-105E

Acceptable Quality Levels (Tightened Inspection)
Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 6 11 15 18 22
t 1.92 1.67 1.25 1.02 0.89
G Ac 2 5 6 7 10
Re 3 6 7 8 11
n 7 13 17 21 27 34
t 1.94 2.27 1.65 1.36 1.23 0.99
H Ac 2 8 8 9 13 17
Re 3 9 9 10 14 18
n 7 14 20 25 33 43 53
t 2.48 2.17 1.93 1.71 1.41 1.25 1.01
J Ac 3 7 10 12 15 21 26
Re 4 8 11 13 16 22 27
n 8 16 22 28 38 50 64 79
t 2.79 2.34 2.12 1.96 1.81 1.52 1.34 1.14
K Ac 4 8 11 14 20 25 33 42
Re 5 9 12 15 21 26 34 43
n 9 18 26 32 45 59 76 96
t 3.08 2.37 2.39 2.15 1.98 1.81 1.49 1.26
L Ac 5 8 14 16 23 31 36 45
Re 6 9 15 17 24 32 37 46
n 10 20 29 37 51 69 89 114
t 3.33 2.54 2.33 2.36 2.16 1.93 1.73 1.58
M Ac 6 9 13 19 26 34 43 57
Re 7 10 14 20 27 35 44 58
(Continued)
759
 760

TABLE T13.4 (Continued)

Tightened Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Acceptable Quality Levels (Tightened Inspection)
Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 11 22 32 41 57 77 102 134
t 3.56 2.70 2.47 2.40 2.34 2.09 1.90 1.78
N Ac 7 10 14 19 29 37 48 66
Re 8 11 15 20 30 38 49 67
n 11 24 35 45 64 89 118 156
t 3.23 2.96 2.89 2.63 2.45 2.28 2.21 1.97
P Ac 5 12 19 22 31 43 61 76
Re 6 13 20 23 32 44 62 77
n 12 26 37 49 70 99 132 175
t 3.46 3.00 2.89 2.83 2.75 2.42 2.32 2.22
Q Ac 6 12 18 25 38 47 65 90
Re 7 13 19 26 39 48 66 91
n 14 27 41 54 78 109 147 198
t 3.20 3.28 2.91 2.86 2.83 2.75 2.50 2.30
R Ac 5 14 18 25 40 59 73 94
Re 6 15 19 26 41 60 74 95
n 30
t 3.36
S Ac 15
Re 16
Source: Reprinted from Schilling, E.G. and Sommers, D.J., J. Qual. Technol., 13(2), 86, 1981, Table 2. With permission.
Appendix
TABLE T13.5
Normal Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Acceptable Quality Levels (Tightened Inspection)
Appendix

Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

n 2
t 0.00
A Ac 0
Re 1
n 3
t 0.00
B Ac 0
Re 1
n 3 5
t 1.19 0.00
C Ac 1 1
Re 2 2
n 4 6 7
t 1.14 1.07 0.48
D Ac 1 3 3
Re 2 4 4
n 5 8 10 11
t 1.20 1.67 0.98 0.61
E Ac 1 5 5 5
Re 2 6 6 6
n 5 9 12 14 17
t 1.92 1.43 1.04 1.02 0.60
F Ac 2 4 5 7 8
Re 3 5 6 8 9
(Continued)
761
 762

TABLE T13.5 (Continued)

Normal Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Acceptable Quality Levels (Tightened Inspection)
Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 6 11 15 18 22 26
t 1.92 1.67 1.25 1.02 0.89 0.77
G Ac 2 5 6 7 10 13
Re 3 6 7 8 11 14
n 7 13 17 21 27 32 38
t 1.94 2.27 1.65 1.36 1.23 1.00 0.84
H Ac 2 8 8 9 13 15 19
Re 3 9 9 10 14 16 20
n 7 14 20 25 33 40 48 56
t 2.48 2.17 1.93 1.71 1.41 1.29 1.20 0.98
J Ac 3 7 10 12 15 19 25 29
Re 4 8 11 13 16 20 26 30
n 8 16 22 28 38 46 56 69 86
t 2.79 2.34 2.12 1.96 1.81 1.64 1.45 1.16 1.05
K Ac 4 8 11 14 20 24 29 33 46
Re 5 9 12 15 21 25 30 34 47
n 9 18 26 32 45 54 68 82 108
t 3.08 2.37 2.39 2.15 1.98 1.90 1.57 1.63 1.30
L Ac 5 8 14 16 23 29 32 46 56
Re 6 9 15 17 24 30 33 47 57
n 10 20 29 37 51 62 79 98 127
t 3.33 2.54 2.33 2.36 2.16 1.84 1.90 1.70 1.55
M Ac 6 9 13 19 26 27 41 49 66
Re 7 10 14 20 27 28 42 50 67
(Continued)
Appendix
 Appendix

TABLE T13.5 (Continued)

Normal Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Acceptable Quality Levels (Tightened Inspection)
Sample Size
Code Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 11 22 32 41 57 72 90 115 148
t 3.56 2.70 2.47 2.40 2.34 2.19 2.05 1.92 1.66
N Ac 7 10 14 19 29 36 45 58 70
Re 8 11 15 20 30 37 46 59 71
n 11 24 35 45 64 81 103 131 171
t 3.23 2.96 2.89 2.63 2.45 2.23 2.25 2.21 1.96
P Ac 5 12 19 22 31 36 52 71 87
Re 6 13 20 23 32 37 53 72 88
n 12 26 37 49 70 89 116 148 196
t 3.46 3.00 2.89 2.83 2.75 2.46 2.47 2.23 2.06
Q Ac 6 12 18 25 38 42 61 71 93
Re 7 13 19 26 39 43 62 72 94
n 27 41 54 78 99 129 163 222
t 3.28 2.91 2.86 2.83 2.78 2.62 2.40 2.25
R Ac 14 18 25 40 53 67 78 106
Re 15 19 26 41 54 68 79 107

Source: Reprinted from Schilling, E.G. and Sommers, D.J., J. Qual. Technol., 13(2), 87, 1981, Table 2. With permission.
763
 764

TABLE T13.6
Reduced Inspection Optimal Narrow Limit Plans for MIL-STD-105E
Sample Acceptable Quality Levels (Reduced Inspection)
Size Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

n 2
t 0.00
A Ac 0
Re 1
n 2
t 0.00
B Ac 0
Re 1
n 2 2
t 0.00 0.00
C Ac 0 0
Re 1 2
n 3 3 3
t 0.00 0.00 0.00
D Ac 0 0 1
Re 1 2 3
n 3 5. 5 5
t 1.19 0.00 0.00 0.00
E Ac 1 0 1 1
Re 2 2 3 4
(Continued)
Appendix
TABLE T13.6 (Continued)
Appendix

Reduced Inspection Optimal Narrow Limit Plans for MIL-STD-105E

Sample Acceptable Quality Levels (Reduced Inspection)
Size Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 4 6 7 8 8
t 1.14 1.07 0.48 0.00 0.00
F Ac 1 1 2 1 2
Re 2 4 4 4 5
n 5 8 10 11 12 13
t 1.20 1.67 0.98 0.61 0.50 0.00
G Ac 1 3 3 2 4 3
Re 2 6 6 6 7 6
n 5 9 12 14 16 17 19
t 1.92 1.43 1.25 1.02 0.83 0.60 0.45
H Ac 2 2 4 4 5 6 8
Re 3 5 7 8 9 9 11
n 6 11 15 18 20 22 26 28
t 1.91 1.90 1.58 1.02 1.14 0.89 0.77 0.49
J Ac 2 4 6 4 7 7 10 11
Re 3 7 9 8 11 11 14 14
n 7 13 17 21 24 27 32 36 41
t 1.94 1.69 1.65 1.48 1.29 1.32 1.00 0.91 0.86
K Ac 2 3 6 7 8 11 12 15 21
Re 3 6 9 11 12 15 16 19 24
(Continued)
765
TABLE T13.6 (Continued)
766

Reduced Inspection Optimal Narrow Limit Plans for MIL-STD-105E

Sample Acceptable Quality Levels (Reduced Inspection)
Size Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10
n 7 14 20 25 29 33 40 45 52
t 2.48 2.17 1.93 1.71 1.59 1.49 1.29 1.07 0.89
L Ac 3 5 8 9 11 13 16 17 20
Re 4 8 11 13 15 17 20 21 24
n 8 16 23 28 34 39 47 54 64
t 2.80 2.34 2.06 1.87 1.84 1.71 1.61 1.41 1.34
M Ac 4 6 9 10 14 16 21 23 30
Re 5 9 12 14 18 20 25 27 34
n 9 18 26 33 39 45 55 64 77
t 2.80 2.51 2.20 2.19 2.01 1.98 1.92 1.73 1.60
N Ac 4 7 10 14 16 20 27 30 37
Re 5 10 13 18 20 24 31 34 41
n 10 20 29 37 44 51 64 75 90
t 2.57 2.54 2.42 2.36 2.28 2.21 2.12 1.88 1.77
P Ac 3 7 12 15 20 24 32 34 42
Re 4 10 15 20 24 28 36 38 46
n 10 22 32 41 49 58 73 86 104
t 3.22 2.70 2.47 2.34 2.32 2.36 2.21 2.18 1.95
Q Ac 5 8 12 14 19 27 34 43 48
Re 6 11 15 19 24 31 38 47 52
n 24 35 45 55 64 82 97 119
t 2.86 2.89 2.63 2.54 2.45 2.31 2.29 2.22
R Ac 9 17 18 23 27 36 46 59
Re 12 20 23 28 32 40 50 63

Source: Reprinted from Schilling, E.G. and Sommers, D.J., J. Qual. Technol., 13(2), 88, 1981, Table 2. With permission.
Appendix
TABLE T13.7
Appendix

MIL-STD-105E Scheme Probability of Acceptance (Pa) and Average Sample Number (ASN) at AQL Using Narrow Limit Plans
(Limit Numbers for Switching to Reduced Inspection Not Used)
Sample Acceptable Quality Levels (Normal Inspection)
Size
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

A Pa .863
ASN 2.21
B Pa .885
ASN 2.68
C Pa .901 .921
ASN 2.73 4.48
D Pa .901 .908 .987
ASN 3.67 5.69 4.06
E Pa .899 .908 .978 .981
ASN 4.18 7.55 6.68 8.02
F Pa .896 .913 .980 .978 .996
ASN 4.75 8.64 8.66 11.20 10.60
G Pa .899 .922 .978 .977 .994 .996
ASN 5.74 10.56 11.71 14.76 15.45 16.48
H Pa .903 .912 .983 ..979 .994 .995 .999
ASN 6.15 12.31 13.54 17.63 19.55 21.34 21.02
J Pa .909 .903 .977 .982 .994 .995 .997 .998
ASN 6.68 13.74 16.68 21.48 24.29 27.24 29.22 31.12
K Pa .903 .911 .975 .976 .996 .995 .998 .998 .999
ASN 7.71 15.66 18.72 24.63 28.13 32.35 35.21 40.59 44.98
L Pa .897 .912 .977 .973 .993 .996 .997 .997 .997
ASN 8.35 17.44 22.01 28.78 34.39 38.42 43.94 50.14 59.43
(Continued)
767
 768

TABLE T13.7 (Continued)

MIL-STD-105E Scheme Probability of Acceptance (Pa) and Average Sample Number (ASN) at AQL Using Narrow Limit Plans
(Limit Numbers for Switching to Reduced Inspection Not Used)
Sample Acceptable Quality Levels (Normal Inspection)
Size
Code
Letter 0.010 0.015 0.025 0.040 0.065 0.10 0.15 0.25 0.40 0.65 1.0 1.5 2.5 4.0 6.5 10

M Pa .902 .924 .978 .976 .992 .994 .999 .997 .998

ASN 9.29 19.29 24.93 32.69 39.94 45.67 50.52 59.67 71.01
N Pa .903 .911 .982 .977 .993 .993 .997 .998 .998
ASN 10.15 21.45 27.81 37.13 45.12 53.65 60.16 69.42 85.28
P Pa .910 .903 .977 .981 .993 .994 .996 .996 .999
ASN 10.67 23.54 30.91 40.90 50.64 59.98 70.47 82.94 96.77
Q Pa .903 .911 .975 .977 .995 .994 .997 .995 .997
ASN 11.43 25.31 33.60 45.07 55.33 67.07 79.46 96.36 115.26
R Pa .912 .976 .972 .993 .996 .997 .996 .996
ASN 26.79 36.88 49.74 62.84 73.26 89.09 107.25 135.31

Source: Reprinted from Schilling, E.G. and Sommers, D.J., J. Qual. Technol., 13(2), 89, 1981, Table 5. With permission.
Appendix
Appendix 769

TABLE T13.8
Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20
n = 5, i = 0
−2.50 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.45 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.40 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.35 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.30 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.25 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.20 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.15 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.10 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.05 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−2.00 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−1.95 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−1.90 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−1.85 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−1.80 .9752 .9510 .9039 .7738 .5905 .4437 .3277
−1.75 .9752 .9509 .9039 .7737 .5904 .4437 .3276
−1.70 .9752 .9509 .9038 .7737 .5904 .4436 .3276
−1.65 .9751 .9509 .9038 .7737 .5904 .4436 .3276
−1.60 .9751 .9508 .9037 .7737 .5903 .4435 .3275
−1.55 .9750 .9507 .9037 .7735 .5902 .4434 .3274
−1.50 .9749 .9506 .9035 .7734 .5901 .4433 .3273
−1.45 .9747 .9504 .9033 .7732 .5899 .4431 .3271
−1.40 .9744 .9501 .9030 .7729 .5896 .4428 .3268
−1.35 .9740 .9497 .9027 .7725 .5892 .4425 .3264
−1.30 .9734 .9492 .9021 .7720 .5887 .4419 .3259
−1.25 .9727 .9484 .9013 .7712 .5879 .4412 .3252
−1.20 .9716 .9473 .9003 .7701 .5869 .4401 .3242
−1.15 .9702 .9459 .8989 .7687 .5855 .4388 .3228
−1.10 .9683 .9440 .8970 .7669 .5836 .4370 .3211
−1.05 .9658 .9416 .8945 .7644 .5812 .4346 .3188
−1.00 .9626 .9383 .8913 .7612 .5780 .4315 .3159
−0.95 .9584 .9342 .8871 .7571 .5740 .4276 .3121
−0.90 .9532 .9289 .8819 .7518 .5689 .4227 .3075
−0.85 .9466 .9223 .8753 .7453 .5626 .4167 .3018
−0.80 .9384 .9142 .8672 .7373 .5548 .4093 .2949
−0.75 .9285 .9043 .8573 .7275 .5454 .4004 .2867
−0.70 .9165 .8923 .8453 .7158 .5342 .3899 .2771
−0.65 .9022 .8780 .8311 .7018 .5209 .3776 .2660
−0.60 .8854 .8613 .8144 .6855 .5055 .3634 .2533
−0.55 .8659 .8418 .7951 .6666 .4878 .3473 .2391
−0.50 .8436 .8195 .7729 .6451 .4678 .3294 .2235
(Continued)
770 Appendix

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20

−0.45 .8182 .7942 .7478 .6208 .4456 .3096 .2067

−0.40 .7899 .7660 .7198 .5939 .4211 .2883 .1889
−0.35 .7586 .7348 .6890 .5644 .3947 .2656 .1703
−0.30 .7245 .7008 .6554 .5325 .3666 .2419 .1513
−0.25 .6877 .6642 .6193 .4985 .3371 .2176 .1323
−0.20 .6486 .6254 .5811 .4628 .3067 .1930 .1137
−0.15 .6075 .5846 .5411 .4258 .2758 .1687 .0959
−0.10 .5649 .5424 .4998 .3880 .2449 .1452 .0792
−0.05 .5213 .4992 .4578 .3500 .2146 .1227 .0639
0.00 .4771 .4557 .4155 .3123 .1854 .1018 .0503
0.05 .4331 .4123 .3736 .2755 .1577 .0828 .0386
0.10 .3897 .3696 .3326 .2401 .1320 .0659 .0287
0.15 .3475 .3282 .2930 .2066 .1086 .0511 .0206
0.20 .3070 .2886 .2554 .1754 .0876 .0387 .0143
0.25 .2685 .2512 .2201 .1468 .0694 .0285 .0095
0.30 .2326 .2163 .1875 .1210 .0537 .0203 .0060
0.35 .1993 .1842 .1577 .0982 .0407 .0140 .0036
0.40 .1690 .1551 .1311 .0784 .0300 .0093 .0020
0.45 .1418 .1291 .1075 .0615 .0216 .0059 .0010
0.50 .1176 .1061 .0869 .0474 .0151 .0036 .0005
0.55 .0964 .0861 .0693 .0358 .0102 .0020 .0002
0.60 .0781 .0691 .0545 .0265 .0066 .0011 .0001
0.65 .0625 .0546 .0421 .0191 .0042 .0005 .0000
0.70 .0494 .0426 .0321 .0135 .0025 .0002 .0000
0.75 .0386 .0328 .0240 .0093 .0014 .0001 .0000
0.80 .0297 .0249 .0177 .0063 .0008 .0000 .0000
0.85 .0226 .0186 .0128 .0041 .0004 .0000 .0000
0.90 .0169 .0136 .0091 .0026 .0002 .0000 .0000
0.95 .0125 .0099 .0063 .0016 .0001 .0000 .0000
1.00 .0091 .0070 .0043 .0009 .0000 .0000 .0000
1.05 .0066 .0049 .0028 .0005 .0000 .0000 .0000
1.10 .0046 .0034 .0018 .0003 .0000 .0000 .0000
1.15 .0032 .0023 .0012 .0001 .0000 .0000 .0000
1.20 .0022 .0015 .0007 .0001 .0000 .0000 .0000
1.25 .0015 .0010 .0004 .0000 .0000 .0000 .0000
1.30 .0010 .0006 .0003 .0000 .0000 .0000 .0000
1.35 .0006 .0004 .0001 .0000 .0000 .0000 .0000
1.40 .0004 .0002 .0001 .0000 .0000 .0000 .0000
1.45 .0003 .0001 .0000 .0000 .0000 .0000 .0000
1.50 .0002 .0001 .0000 .0000 .0000 .0000 .0000
1.55 .0001 .0000 .0000 .0000 .0000 .0000 .0000
1.60 .0001 .0000 .0000 .0000 .0000 .0000 .0000
1.65 .0000 .0000 .0000 .0000 .0000 .0000 .0000
(Continued)
Appendix 771

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20

1.70 .0000 .0000 .0000 .0000 .0000 .0000 .0000

1.75 .0000 .0000 .0000 .0000 .0000 .0000 .0000
1.80 .0000 .0000 .0000 .0000 .0000 .0000 .0000
n = 5, i = 1
−2.50 .024 .048 .092 .204 .328 .391 .410
−2.45 .024 .048 .092 .204 .328 .391 .410
−2.40 .024 .048 .092 .204 .328 .391 .410
−2.35 .024 .048 .092 .204 .328 .391 .410
−2.30 .024 .048 .092 .204 .328 .391 .410
−2.25 .024 .048 .092 .204 .328 .391 .410
−2.20 .024 .048 .092 .204 .328 .391 .410
−2.15 .024 .048 .092 .204 .328 .391 .410
−2.10 .024 .048 .092 .204 .328 .391 .410
−2.05 .024 .048 .092 .204 .328 .391 .410
−2.00 .024 .048 .092 .204 .328 .391 .410
−1.95 .024 .048 .092 .204 .328 .391 .410
−1.90 .024 .048 .092 .204 .328 .391 .410
−1.85 .024 .048 .092 .204 .328 .391 .410
−1.80 .024 .048 .092 .204 .328 .391 .410
−1.75 .024 .048 .092 .204 .328 .391 .410
−1.70 .024 .048 .092 .204 .328 .391 .410
−1.65 .024 .048 .092 .204 .328 .391 .410
−1.60 .024 .048 .092 .204 .328 .391 .410
−1.55 .024 .048 .092 .204 .328 .391 .410
−1.50 .024 .048 .092 .204 .328 .391 .410
−1.45 .024 .048 .092 .204 .328 .391 .410
−1.40 .024 .048 .092 .204 .328 .391 .410
−1.35 .024 .048 .092 .204 .328 .391 .410
−1.30 .024 .048 .092 .204 .328 .391 .410
−1.25 .024 .048 .092 .204 .328 .391 .409
−1.20 .024 .048 .092 .204 .328 .391 .409
−1.15 .024 .048 .092 .204 .328 .391 .409
−1.10 .024 .048 .092 .204 .328 .391 .409
−1.05 .024 .048 .092 .204 .328 .391 .409
−1.00 .024 .048 .092 .204 .328 .391 .409
−0.95 .024 .048 .092 .203 .328 .391 .408
−0.90 .024 .048 .092 .203 .327 .390 .408
−0.85 .024 .048 .092 .203 .327 .390 .407
−0.80 .024 .048 .092 .203 .327 .389 .406
−0.75 .024 .048 .092 .203 .326 .388 .404
−0.70 .024 .048 .092 .203 .326 .387 .402
−0.65 .024 .048 .092 .202 .325 .385 .398
−0.60 .024 .048 .092 .202 .323 .382 .394
(Continued)
772 Appendix

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20
−0.55 .024 .048 .092 .201 .321 .379 .389
−0.50 .024 .048 .091 .201 .319 .374 .383
−0.45 .024 .048 .091 .199 .316 .369 .375
−0.40 .024 .047 .091 .198 .312 .362 .365
−0.35 .024 .047 .090 .196 .307 .353 .353
−0.30 .024 .047 .090 .194 .301 .343 .339
−0.25 .024 .047 .089 .191 .294 .331 .322
−0.20 .024 .046 .088 .187 .285 .317 .304
−0.15 .024 .046 .086 .183 .275 .302 .283
−0.10 .023 .045 .085 .178 .263 .284 .261
−0.05 .023 .044 .083 .172 .250 .264 .237
0.00 .023 .043 .081 .165 .236 .243 .212
0.05 .022 .042 .078 .158 .220 .221 .187
0.10 .022 .041 .075 .149 .203 .198 .161
0.15 .021 .039 .072 .140 .185 .175 .136
0.20 .020 .038 .068 .130 .167 .151 .113
0.25 .019 .036 .064 .120 .148 .129 .091
0.30 .018 .034 .060 .109 .129 .108 .072
0.35 .017 .032 .055 .098 .111 .088 .055
0.40 .016 .029 .051 .087 .094 .070 .041
0.45 .015 .027 .046 .077 .078 .055 .029
0.50 .014 .025 .041 .066 .064 .041 .020
0.55 .013 .022 .037 .057 .051 .030 .014
0.60 .011 .020 .032 .047 .039 .022 .009
0.65 .010 .018 .028 .039 .030 .015 .005
0.70 .009 .015 .024 .032 .022 .010 .003
0.75 .008 .013 .020 .025 .016 .006 .002
0.80 .007 .011 .017 .020 .011 .004 .001
0.85 .006 .009 .014 .015 .007 .002 .000
0.90 .005 .008 .011 .011 .005 .001 .000
0.95 .004 .006 .009 .008 .003 .001 .000
1.00 .003 .005 .007 .006 .002 .000 .000
1.05 .003 .004 .005 .004 .001 .000 .000
1.10 .002 .003 .004 .003 .001 .000 .000
1.15 .002 .002 .003 .002 .000 .000 .000
1.20 .001 .002 .002 .001 .000 .000 .000
1.25 .001 .001 .001 .001 .000 .000 .000
1.30 .001 .001 .001 .000 .000 .000 .000
1.35 .001 .001 .001 .000 .000 .000 .000
1.40 .000 .000 .000 .000 .000 .000 .000
1.45 .000 .000 .000 .000 .000 .000 .000
1.50 .000 .000 .000 .000 .000 .000 .000
1.55 .000 .000 .000 .000 .000 .000 .000
(Continued)
Appendix 773

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20

1.60 .000 .000 .000 .000 .000 .000 .000

1.65 .000 .000 .000 .000 .000 .000 .000
1.70 .000 .000 .000 .000 .000 .000 .000
1.75 .000 .000 .000 .000 .000 .000 .000
1.80 .000 .000 .000 .000 .000 .000 .000
n = 5, i = 2
−2.50 .000 .001 .004 .021 .073 .138 .205
−2.45 .000 .001 .004 .021 .073 .138 .205
−2.40 .000 .001 .004 .021 .073 .138 .205
−2.35 .000 .001 .004 .021 .073 .138 .205
−2.30 .000 .001 .004 .021 .073 .138 .205
−2.25 .000 .001 .004 .021 .073 .138 .205
−2.20 .000 .001 .004 .021 .073 .138 .205
−2.15 .000 .001 .004 .021 .073 .138 .205
−2.10 .000 .001 .004 .021 .073 .138 .205
−2.05 .000 .001 .004 .021 .073 .138 .205
−2.00 .000 .001 .004 .021 .073 .138 .205
−1.95 .000 .001 .004 .021 .073 .138 .205
−1.90 .000 .001 .004 .021 .073 .138 .205
−1.85 .000 .001 .004 .021 .073 .138 .205
−1.80 .000 .001 .004 .021 .073 .138 .205
−1.75 .000 .001 .004 .021 .073 .138 .205
−1.70 .000 .001 .004 .021 .073 .138 .205
−1.65 .000 .001 .004 .021 .073 .138 .205
−1.60 .000 .001 .004 .021 .073 .138 .205
−1.55 .000 .001 .004 .021 .073 .138 .205
−1.50 .000 .001 .004 .021 .073 .138 .205
−1.45 .000 .001 .004 .021 .073 .138 .205
−1.40 .000 .001 .004 .021 .073 .138 .205
−1.35 .000 .001 .004 .021 .073 .138 .205
−1.30 .000 .001 .004 .021 .073 .138 .205
−1.25 .000 .001 .004 .021 .073 .138 .205
−1.20 .000 .001 .004 .021 .073 .138 .205
−1.15 .000 .001 .004 .021 .073 .138 .205
−1.10 .000 .001 .004 .021 .073 .138 .205
−1.05 .000 .001 .004 .021 .073 .138 .205
−1.00 .000 .001 .004 .021 .073 .138 .205
−0.95 .000 .001 .004 .021 .073 .138 .205
−0.90 .000 .001 .004 .021 .073 .138 .205
−0.85 .000 .001 .004 .021 .073 .138 .205
−0.80 .000 .001 .004 .021 .073 .138 .205
−0.75 .000 .001 .004 .021 .073 .138 .205
−0.70 .000 .001 .004 .021 .073 .138 .205
(Continued)
774 Appendix

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20

−0.65 .000 .001 .004 .021 .073 .138 .205

−0.60 .000 .001 .004 .021 .073 .138 .205
−0.55 .000 .001 .004 .021 .073 .138 .204
−0.50 .000 .001 .004 .021 .073 .138 .204
−0.45 .000 .001 .004 .021 .073 .138 .204
−0.40 .000 .001 .004 .021 .073 .138 .203
−0.35 .000 .001 .004 .021 .073 .137 .202
−0.30 .000 .001 .004 .021 .073 .137 .201
−0.25 .000 .001 .004 .021 .072 .136 .199
−0.20 .000 .001 .004 .021 .072 .135 .197
−0.15 .000 .001 .004 .021 .072 .134 .194
−0.10 .000 .001 .004 .021 .072 .133 .191
−0.05 .000 .001 .004 .021 .071 .131 .186
0.00 .000 .001 .004 .021 .070 .128 .180
0.05 .000 .001 .004 .021 .069 .125 .173
0.10 .000 .001 .004 .021 .068 .121 .165
0.15 .000 .001 .004 .021 .066 .116 .155
0.20 .000 .001 .004 .020 .065 .111 .144
0.25 .000 .001 .004 .020 .062 .105 .132
0.30 .000 .001 .004 .020 .060 .097 .119
0.35 .000 .001 .004 .019 .056 .090 .105
0.40 .000 .001 .004 .018 .053 .081 .091
0.45 .000 .001 .003 .018 .049 .072 .078
0.50 .000 .001 .003 .017 .045 .063 .065
0.55 .000 .001 .003 .016 .041 .054 .052
0.60 .000 .001 .003 .015 .036 .046 .041
0.65 .000 .001 .003 .014 .031 .037 .031
0.70 .000 .001 .003 .012 .027 .030 .023
0.75 .000 .001 .003 .011 .023 .023 .017
0.80 .000 .001 .002 .010 .018 .018 .012
0.85 .000 .001 .002 .009 .015 .013 .008
0.90 .000 .001 .002 .007 .012 .009 .005
0.95 .000 .001 .002 .006 .009 .006 .003
1.00 .000 .001 .002 .005 .007 .004 .002
1.05 .000 .000 .001 .004 .005 .003 .001
1.10 .000 .000 .001 .003 .003 .002 .001
1.15 .000 .000 .001 .003 .002 .001 .000
1.20 .000 .000 .001 .002 .001 .001 .000
1.25 .000 .000 .001 .001 .001 .000 .000
1.30 .000 .000 .001 .001 .001 .000 .000
1.35 .000 .000 .000 .001 .000 .000 .000
1.40 .000 .000 .000 .000 .000 .000 .000
1.45 .000 .000 .000 .000 .000 .000 .000
(Continued)
Appendix 775

TABLE T13.8 (Continued)

Joint Probabilities for Mixed Plans
Fraction Defective, p
zA .005 .01 .02 .05 .10 .15 .20

1.50 .000 .000 .000 .000 .000 .000 .000

1.55 .000 .000 .000 .000 .000 .000 .000
1.60 .000 .000 .000 .000 .000 .000 .000
1.65 .000 .000 .000 .000 .000 .000 .000
1.70 .000 .000 .000 .000 .000 .000 .000
1.75 .000 .000 .000 .000 .000 .000 .000
1.80 .000 .000 .000 .000 .000 .000 .000

Source: Reprinted from Schilling, E.G. and Dodge, H.F., Technometrics, 11(2), 362, 1969. With
permission.
776 Appendix

TABLE T14.1
Values of x and y for Determining AOQL
Given Given Given Given
c x y c x y c x y c x y

0 1.00 0.3679 10 8.05 6.528 20 15.92 13.89 30 24.11 21.70

1 1.62 0.8400 11 8.82 7.233 21 16.73 14.66 31 24.95 22.50
2 2.27 1.371 12 9.59 7.948 22 17.54 15.43 32 25.78 23.30
3 2.95 1.942 13 10.37 8.670 23 18.35 16.20 33 26.62 24.10
4 3.64 2.544 14 11.15 9.398 24 19.17 16.98 34 27.45 24.90
5 4.35 3.168 15 11.93 10.13 25 19.99 17.76 35 28.29 25.71
6 5.07 3.812 16 12.72 10.88 26 20.81 18.54 36 29.13 26.52
7 5.80 4.472 17 13.52 11.62 27 21.63 19.33 37 29.97 27.33
8 6.55 5.146 18 14.31 12.37 28 22.46 20.12 38 30.82 28.14
9 7.30 5.831 19 15.12 13.13 29 23.29 20.91 39 31.66 28.96
10 8.05 6.528 20 15.92 13.89 30 24.11 21.70 40 32.51 29.77
Source: Dodge, H.F. and Romig, H.G., Sampling Inspection Tables, Single and Double Sampling, 2nd ed.,
John Wiley & Sons, New York, 1959. With permission.
Appendix 777

TABLE T16.1
Values of Y for Determining AOQL, for SkSP-2 Plans
n/N = 0
i
c f 4 6 8 10

1 2/3 0.8682 0.8479 0.8421 0.8405

1/2 0.8954 0.8564 0.8443 .08411
1/3 0.9443 0.8784 0.8493 0.8423
1/4 0.9861 0.8939 0.8549 0.8436
1/5 1.0219 0.9125 0.8613 0.8450
2 2/3 1.4281 1.3935 1.3794 1.3741
1/2 1.4785 1.4163 1.3884 1.3773
1/3 1.5619 1.4604 1.4081 1.3844
1/4 1.6284 1.5000 1.4291 1.3927
1/5 1.6835 1.5349 1.4501 1.4021
3 2/3 2.0294 1.9835 1.9610 1.9505
1/2 2.1023 2.0229 1.9806 1.9593
1/3 2.2177 2.0927 2.0205 1.9789
1/4 2.3067 2.1511 2.0582 2.0004
1/5 2.3971 2.2006 2.0925 2.0223
4 2/3 2.6604 2.6054 2.5754 2.5594
1/2 2.7547 2.6615 2.6076 2.5764
1/3 2.8998 2.7561 2.6683 2.6124
1/4 3.0097 2.8320 2.7217 2.6482
1/5 3.0980 2.8948 2.7681 2.6817
5 2/3 3.3140 3.2516 3.2151 3.1939
1/2 3.4286 3.3242 3.2605 3.2207
1/3 3.6018 3.4423 3.3417 3.2742
1/4 3.7312 3.5346 3.4098 3.3236
1/5 3.8344 3.6100 3.4674 3.3677
6 2/3 3.9857 3.9171 3.8751 3.8491
1/2 4.1197 4.0058 3.9338 3.8866
1/3 4.3195 4.1463 4.0347 3.9575
1/4 4.4673 4.2543 4.1168 4.0198
1/5 4.5846 4.3416 4.1850 4.0738
7 2/3 4.6726 4.5986 4.5518 4.5215
1/2 4.8250 4.7029 4.6237 4.5700
1/3 5.0502 4.8650 4.7437 4.6580
1/4 5.2156 4.9878 4.8390 4.7326
1/5 5.3462 5.0865 4.9174 4.7959
(Continued)
778 Appendix

TABLE T16.1 (Continued)

Values of Y for Determining AOQL, for SkSP-2 Plans
n/N = 0
i
c f 4 6 8 10
8 2/3 5.3722 5.2937 5.2425 5.2083
1/2 5.5424 5.4130 5.3275 5.2680
1/3 5.7919 5.5958 5.4658 5.3727
1/4 5.9740 5.7330 5.5739 5.4590
1/5 6.1174 5.8425 5.6620 5.5313
9 2/3 6.0829 6.0003 5.9454 5.9077
1/2 6.2702 6.1343 6.0431 5.9785
1/3 6.5430 6.3372 6.1993 6.0993
1/4 6.7413 6.4881 6.3197 6.1970
1/5 6.8969 6.6080 6.4171 6.2779
10 2/3 6.8033 6.7172 6.6588 6.6180
1/2 7.0070 6.8654 6.7690 6.7000
1/3 7.3024 7.0876 6.9425 6.8363
1/4 7.5162 7.2518 7.0748 6.9450
1/5 7.6836 7.3819 7.1812 7.0341
Source: Reprinted from Perry, R.L., A system of skip-lot ­sampling plans
for lot inspection, PhD dissertation, Rutgers—The State
University, New Brunswick, NJ, 1970. With permission.
Appendix 779

TABLE T16.2
Unity Values for SkSP-2 and Matched Single-Sampling Plans
Matched Single-Sampling Plan Skip-Lot Plan SkSP-2 Ratio of SkSP-2 Sample Size
to Matched Single-Sampling
c* OR n*p.95 f, i c OR np.95 Plan Sample Size

2 6.500 0.818 (1/5,8) 1 6.505 0.598 .731

3 4.890 1.366 (1/5,14) 2 4.883 1.090 .731
4 4.057 1.970 (1/2,4) 3 4.063 1.645 .830
5 3.549 2.613 (1/2,6) 4 3.522 2.270 .868
(1/2,8) 4 3.574 2.237 .856
(1/5,8) 3 3.561 1.876 .718
6 3.206 3.285 (1/2,10) 5 3.207 2.892 .880
(1/4,8) 4 3.191 2.505 .762
7 2.957 3.981 (1/2,12) 6 2.951 3.569 .894
(1/4,10) 5 2.930 3.166 .795
(1/5,14) 5 2.963 3.130 .789
(1/5,6) 4 2.982 2.681 .673
8 2.768 4.695 (2/3,4) 7 2.757 4.270 .909
(2/3,6) 7 2.777 5.238 .902
(1/2,14) 7 2.759 4.266 .908
(1/4,14) 6 2.778 3.791 .807
(1/5,8) 5 2.782 3.334 .709
9 2.618 5.425 (2/3,6) 8 2.611 4.977 .917
(2/3,8) 8 2.627 4.947 .912
(1/2,4) 7 2.627 4.482 .626
(1/3,10) 7 2.597 4.533 .835
(1/5,10) 6 2.629 4.007 .738
(1/5,4) 5 2.594 3.578 .659
10 2.497 6.169 (2/3,8) 9 2.493 5.698 .924
(2/3,10) 9 2.505 5.670 .919
(1/3,14) 8 2.507 5.184 .840
(1/5,12) 7 2.506 4.698 .761
(1/5,6) 6 2.499 4.215 .683
Source: Reprinted from Dodge, H.F. and Perry, R.L., ASQC Technical Conference Transactions, American Society
for Quality Control, Inc., Chicago, IL, 1971, p. 477. With permission.
780 Appendix

TABLE T16.3
Poisson Unity Values for Constructing ChSP-1 Plans
np1 for np2 for
AOQL
i L(p1) = 0.95 L(p2) = 0.10 p2/p1 nAOQL npM
p1

1 0.207 2.490 12.029 0.5033 2.431 1.000

2 0.162 2.325 14.352 0.4190 2.586 0.897
3 0.139 2.303 16.568 0.3889 2.798 0.902
4 0.124 2.303 18.573 0.3764 3.036 0.943
5 0.114 2.303 20.202 0.3717 3.261 0.972
6 0.106 2.303 21.726 0.3689 3.500 0.990
7 0.100 2.303 23.030 0.3683 3.483 0.994
8 0.094 2.303 24.500 0.3680 3.915 0.998
9 0.090 2.303 25.589 0.3679 4.088 0.999
10 0.087 2.303 26.471 0.3679 4.229 0.999
∞ 0.051 2.303 44.890 0.3680 7.214 1.000
Source: Reprinted from Soundararajan, V., J. Qual. Technol., 10(3), 58, 1978. With permission.
Appendix 781

TABLE T16.4
ChSP-1 Plans Indexed by AQL (p.95) and LTPD (p.10)
AQL in Percent
LTPD in Sample
Percent Size 0.10 0.15 0.25 0.40 0.65 1.00 1.50 2.50 4.00 6.50

1.0 228 2
1.5 152 4 1
2.0 114 7 2
2.5 91 3 1
3.0 76 4 2
3.5 65 2
4.0 57 3 1
4.5 50 4 2
5.0 45 5 2
5.5 41 7 3
6.0 38 9 3
6.5 35 4 1
7.0 32 5 1
7.5 30 5 1
8.0 28 6 2
8.5 26 7 2
9.0 25 2
9.5 23 3
10.0 22 3 1
11.0 20 4 2 1
12.0 18 5 2 1
13.0 17 5 2 1 1
14.0 16 2 1 1
15.0 15 3 1 1
16.0 14 3 2 1 1
17.0 13 4 2 1 1
18.0 12 5 2 1 1
19.0 11 6 3 1 1
20.0 11 6 3 1 1
21.0 10 7 3 1 1
22.0 10 7 3 1 1
23.0 9 4 1 1
24.0 9 4 1 1
25.0 8 5 2 1
30.0 7 7 2 1
35.0 6 2 1
40.0 5 4 2 1
50.0 4 7 3 1
60.0 3 6 2
70.0 2 8 4
Source: Reprinted from Soundararajan, V., J. Qual. Technol., 10(3), 101, 1978. With permission.
 782

TABLE T16.5
ChSP-1 Plans Indexed by AQL (p.95) and AOQL
AOQL in Percent
0.10 0.25 0.50 0.75 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
AQL in
Percent n i n i n i n i n i n i n i n i n i n i n i n i

0.05 504, 1 147,

0.075 149, 5
0.10 168, 2 73,
0.15 74, 3 49,
0.20 89, 2 50, 7 36,
0.25 101, 1 51, 4 37, 9
0.30 56, 2 38, 5 24,
0.35 68, 1 39, 3 25, 10
0.40 42, 2 25, 7
0.45 51, 1 25, 5 18,
0.50 51, 1 26, 4 19, 9
0.55 26, 3 19, 6 14,
0.60 28, 2 19, 5 15, 9
0.65 34, 1 19, 4 15, 8 12,
0.70 34, 1 20, 3 15, 6 13, 10
0.75 20, 3 15, 5 13, 9 10,
0.80 22, 2 16, 4 13, 7 11, 10
0.85 26, 1 16, 4 13, 6 11, 9
0.90 26, 1 16, 3 13, 5 11, 8 9,
0.95 26, 1 17, 2 13, 4 11, 7 10, 10
1.0 26, 1 17, 2 13, 4 11, 6 10, 9 8,
1.5 17, 1 15, 1 11, 2 9, 4
2.0 13, 1 12, 1
(Continued)
Appendix
TABLE T16.5 (Continued)
Appendix

ChSP-1 Plans Indexed by AQL (p.95) and AOQL

5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0
AQL in
Percent n i n i n i n i n i n i n i n i n i n i n i

1.0 7, 7, 6, 6,
1.5 8, 5 7, 7 7, 9 6, 10 5, 5, 5, 5,
2.0 9, 2 7, 3 7, 4 6, 5 6, 6 5, 7 5, 9 5, 10 5, 4, 4,
2.5 11, 1 10, 1 9, 1 7, 2 6, 3 5, 4 5, 5 5, 6 5, 7 4, 8 4, 9
3.0 9, 1 8, 1 8, 1 7, 1 6, 2 5, 3 5, 4 4, 4 4, 5
3.5 8, 1 7, 1 7, 1 6, 1 5, 2 5, 3 4, 3
4.0 7, 1 6, 1 6, 1 6, 1 5, 2
4.5 6, 1 6, 1 5, 1
5.0 5, 1

10.5 11.0 11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5
AQL in
Percent n i n i n i n i n i n i n i n i n i n i n i

2.00 4,
2.50 4, 10 3, 3, 3, 3, 3,
3.00 4, 6 4, 7 4, 8 3, 9 3, 9 3, 10 3, 3, 3, 3,
3.50 4, 4 4, 4 4, 5 3, 6 3, 6 3, 7 3, 8 3, 9 3, 10 3, 10 3,
4.00 4, 2 4, 3 4, 3 4, 4 3, 4 3, 5 3, 5 3, 6 3, 7 3, 8 3, 8
4.50 5, 1 5, 1 4, 2 4, 2 4, 3 3, 3 3, 4 3, 4 3, 5 3, 5 3, 6
5.00 5, 1 5, 1 5, 1 5, 1 4, 2 4, 2 3, 3 3, 3 3, 3 3, 4 3, 4
5.50 5, 1 5, 1 5, 1 4, 1 4, 1 3, 2 3, 2 3, 2 3, 3 3, 3
6.00 5, 1 4, 1 4, 1 4, 1 4, 1 4, 1 3, 2 3, 2
6.50 4, 1 4, 1 4, 1 4, 1 4, 1 4, 1
7.00 4, 1 4, 1 4, 1 4, 1
7.50 4, 1 4, 1
(Continued)
783
 784

TABLE T16.5 (Continued)

ChSP-1 Plans Indexed by AQL (p.95) and AOQL
16.00 16.50 17.00 17.50 18.00 18.50 19.00 19.50 20.00 30.00
AQL in
Percent n i n i n i n i n i n i n i n i n i n i
3.50 3, 3, 3, 2,
4.00 3, 9 3, 9 3, 10 3, 10 2, 2, 2, 2,
4.50 3, 6 3, 7 3, 8 3, 8 2, 9 2, 9 2, 10 2, 10 2,
5.00 3, 5 3, 5 3, 6 3, 6 2, 7 2, 7 2, 8 2, 8 2, 9
5.50 3, 3 3, 4 3, 4 3, 5 2, 5 2, 6 2, 6 2, 6 2, 7
6.00 3, 3 3, 3 3, 3 3, 4 2, 4 2, 5 2, 5 2, 5 2, 6
6.50 4, 1 3, 2 3, 2 3, 3 3, 3 3, 3 2, 3 2, 4 2, 4 2,
7.00 4, 1 4, 1 3, 1 3, 2 3, 2 3, 3 2, 3 2, 3 2, 3 2, 10
7.50 4, 1 4, 1 3, 1 3, 1 3, 1 3, 2 3, 2 3, 2 2, 2 2, 9
8.00 4, 1 3, 1 3, 1 3, 1 3, 1 3, 1 3, 1 3, 2 2, 7
8.50 3, 1 3, 1 3, 1 3, 1 3, 1 3, 1 3, 1 2, 6
9.00 3, 1 3, 1 2, 1 3, 1 3, 1 2, 5
9.50 2, 1 3, 1 3, 1 2, 4
10.00 3, 1 2, 4
Source: Reprinted from Soundararajan, V., J. Qual. Technol., 10(3), 100, 1978. With permission.
Appendix
Appendix 785

TABLE T17.1
Parametric Values of Some TNT Plans
np1 for np2 for AOQL/p1
s t α = 0.05 β = 0.10 R = p2/p1 np0 npm nAOQL for α = 0.05 h0

1 2 0.05102 1.18167 23.16090 0.48121 0.47536 0.24063 4.71639 1.01922

1 3 0.05086 1.15471 22.70370 0.43098 0.41973 0.21559 4.23889 1.03486
1 4 0.05070 1.15164 22.71480 0.39880 0.40199 0.19941 3.93314 0.99261
1 5 0.05052 1.15132 22.78940 0.37799 0.42687 0.18984 3.75772 0.92596
1 6 0.05033 1.15129 22.87480 0.36478 0.46871 0.18589 3.69342 0.85779
2 3 0.05052 1.15423 22.84700 0.41667 0.42436 0.20838 4.12470 0.97923
2 4 0.05023 1.15159 22.92630 0.38825 0.42261 0.19470 3.87607 0.93071
2 5 0.04992 1.15132 23.06330 0.37069 0.45351 0.18786 3.76332 0.86921
2 6 0.04961 1.15129 23.20680 0.36007 0.47980 0.18529 3.73493 0.81266
2 7 0.04927 1.15129 23.36700 0.35388 0.49168 0.18441 3.74285 0.76885
3 4 0.04984 1.15158 23.10550 0.38425 0.43274 0.19322 3.87681 0.90446
3 5 0.04944 1.15132 23.28720 0.36801 0.46136 0.18731 3.78863 0.84684
3 6 0.04903 1.15129 23.48130 0.35839 0.48293 0.18512 3.77565 0.79589
3 7 0.04861 1.15129 23.68420 0.35290 0.49286 0.18436 3.79264 0.75751
3 8 0.04818 1.15129 23.89560 0.34988 0.49704 0.18409 3.82088 0.73156
4 5 0.04905 1.15132 23.47240 0.36668 0.46488 0.18707 3.81386 0.83514
4 6 0.04857 1.15129 23.70370 0.35757 0.48433 0.18504 3.80976 0.78737
4 7 0.04808 1.15129 23.94530 0.35242 0.49339 0.18433 3.83382 0.75186
4 8 0.04759 1.15129 24.19190 0.34962 0.49724 0.18408 3.86804 0.72809
4 9 0.04709 1.15129 24.44870 0.34813 0.49886 0.18399 3.90720 0.71325
5 6 0.04818 1.15129 23.89560 0.35711 0.48507 0.18500 3.83977 0.78237
5 7 0.04764 1.15129 24.16650 0.35216 0.49367 0.18432 3.86902 0.74861
5 8 0.04711 1.15129 24.43830 0.34947 0.49735 0.18408 3.90745 0.72609
5 9 0.04657 1.15129 24.72170 0.34805 0.49890 0.18399 3.95083 0.71209
5 10 0.04603 1.15129 25.01170 0.34732 0.49952 0.18396 3.99652 0.70379

Source: Reprinted from Soundararajan, V. and Vijayaraghavan, R., J. Qual. Technol., 22(2), 151, 1990. With permission.
786 Appendix

TABLE T17.2
Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50

*0 0 0.051 2.303 44.891 0.693 0.693 .5000 .5000

1 0 0.308 2.528 8.213 1.146 1.230 .6822 .3178
*1 1 0.355 3.890 10.946 1.678 1.052 .5000 .5000
2 0 0.644 2.821 4.383 1.568 1.748 .7916 .2084
2 1 0.770 4.080 5.301 2.156 1.530 .6346 .3654
*2 2 0.818 5.322 6.509 2.674 1.319 .5000 .5000
3 0 1.005 3.149 3.134 1.976 2.259 .8614 .1386
3 1 1.210 4.335 3.581 2.608 2.012 .7342 .2658
3 2 1.318 5.496 4.170 3.159 1.769 .6116 .3884
*3 3 1.366 6.681 4.890 3.672 1.541 .5000 .5000
4 0 1.375 3.494 2.540 2.376 2.766 .9071 .0929
4 1 1.653 4.633 2.803 3.046 2.496 .8076 .1924
4 2 1.823 5.729 3.142 3.625 2.229 .7017 .2983
4 3 1.921 6.844 3.562 4.161 1.974 .5974 .4026
*4 4 1.970 7.994 4.057 4.671 1.735 .5000 .5000
5 0 1.750 3.849 2.199 2.771 3.271 .9374 .0626
5 1 2.091 4.958 2.371 3.473 2.982 .8612 .1388
5 2 2.320 6.006 2.588 4.077 2.695 .7730 .2270
5 3 2.472 7.062 2.857 4.634 2.418 .6798 .3202
5 4 2.564 8.150 3.179 5.162 2.156 .5875 .4125
*5 5 2.613 9.275 3.549 5.670 1.909 .5000 .5000
6 0 2.127 4.210 1.979 3.162 3.775 .9577 .0423
6 1 2.524 5.300 2.100 3.893 3.470 .9002 .0998
6 2 2.806 6.314 2.251 4.519 3.166 .8286 .1714
6 3 3.009 7.324 2.434 5.095 2.872 .7481 .2519
6 4 3.149 8.357 2.654 5.640 2.589 .6639 .3361
6 5 3.236 9.426 2.913 6.163 2.321 .5801 .4199
*6 6 3.285 10.532 3.206 6.670 2.069 .5000 .5000
7 0 2.505 4.574 1.826 3.550 4.278 .9713 .0287
7 1 2.951 5.654 1.916 4.307 3.960 .9285 .0715
7 2 3.279 6.643 2.026 4.954 3.642 .8714 .1286
7 3 3.530 7.618 2.158 5.547 3.332 .8036 .1964
7 4 3.716 8.606 2.316 6.107 3.033 .7291 .2709
7 5 3.847 9.625 2.502 6.644 2.746 .6516 .3484
7 6 3.932 10.680 2.716 7.164 2.474 .5744 .4256
*7 7 3.981 11.771 2.957 7.669 2.218 .5000 .5000
8 0 2.883 4.941 1.714 3.936 4.781 .9805 .0195
8 1 3.373 6.014 1.783 4.716 4.451 .9489 .0511
8 2 3.743 6.987 1.867 5.382 4.121 .9041 .0959
8 3 4.036 7.935 1.966 5.992 3.798 .8481 .1519
8 4 4.266 8.889 2.084 6.566 3.484 .7836 .2164
8 5 4.440 9.864 2.222 7.116 3.182 .7139 .2861
8 6 4.564 10.872 2.382 7.647 2.892 .6417 .3583
8 7 4.646 11.915 2.565 8.164 2.617 .5697 .4303
(Continued)
Appendix 787

TABLE T17.2 (Continued)

Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50

*8 8 4.695 12.995 2.768 8.669 2.357 .5000 .5000

9 0 3.261 5.310 1.628 4.320 5.283 .9867 .0133
9 1 3.790 6.380 1.683 5.123 4.943 .9635 .0365
9 2 4.197 7.342 1.749 5.806 4.602 .9288 .0712
9 3 4.529 8.270 1.826 6.430 4.268 .8833 .1167
9 4 4.798 9.196 1.917 7.018 3.941 .8287 .1713
9 5 5.013 10.136 2.022 7.580 3.625 .7672 .2328
9 6 5.177 11.102 2.144 8.122 3.321 .7013 .2987
9 7 5.296 12.102 2.285 8.649 3.030 .6335 .3665
9 8 5.376 13.137 2.443 9.164 2.752 .5658 .4342
*9 9 5.425 14.206 2.618 9.669 2.488 .5000 .5000
10 0 3.639 5.679 1.561 4.703 5.784 .9909 .0091
10 1 4.203 6.750 1.606 5.526 5.436 .9740 .0260
10 2 4.645 7.704 1.659 6.225 5.086 .9474 .0526
10 3 5.010 8.618 1.720 6.864 4.741 .9109 .0891
10 4 5.315 9.522 1.791 7.464 4.404 .8653 .1347
10 5 5.568 10.434 1.874 8.037 4.076 .8122 .1878
10 6 5.770 11.366 1.970 8.590 3.758 .7533 .2467
10 7 5.927 12.326 2.080 9.127 3.453 .6908 .3092
10 8 6.042 13.318 2.204 9.651 3.159 .6266 .3734
10 9 6.120 14.346 2.344 10.164 2.879 .5625 .4375
*10 10 6.169 15.407 2.497 10.669 2.613 .5000 .5000
11 0 4.017 6.050 1.506 5.085 6.285 .9938 .0062
11 1 4.614 7.121 1.544 5.926 5.929 .9815 .0185
11 2 5.085 8.071 1.587 6.641 5.571 .9612 .0388
11 3 5.481 8.974 1.637 7.293 5.218 .9323 .0677
11 4 5.819 9.862 1.695 7.905 4.871 .8948 .1052
11 5 6.105 10.752 1.761 8.489 4.532 .8496 .1504
11 6 6.344 11.656 1.837 9.052 4.203 .7979 .2021
11 7 6.537 12.581 1.925 9.598 3.884 .7414 .2586
11 8 6.687 13.536 2.024 10.131 3.577 .6818 .3182
11 9 6.798 14.523 2.136 10.653 3.282 .6206 .3794
11 10 6.875 15.545 2.261 11.165 3.001 .5596 .4404
*11 11 6.924 16.598 2.397 11.668 2.732 .5000 .5000
12 0 4.394 6.420 1.461 5.466 6.786 .9958 .0042
12 1 5.021 7.495 1.493 6.325 6.423 .9869 .0131
12 2 5.520 8.442 1.529 7.054 6.058 .9715 .0285
12 3 5.945 9.338 1.571 7.718 5.697 .9488 .0512
12 4 6.312 10.214 1.618 8.342 5.341 .9183 .0817
12 5 6.629 11.086 1.672 8.936 4.993 .8804 .1196
12 6 6.900 11.966 1.734 9.509 4.653 .8357 .1643
12 7 7.127 12.864 1.805 10.064 4.324 .7855 .2145
12 8 7.312 13.785 1.885 10.605 4.004 .7310 .2690
12 9 7.457 14.736 1.976 11.134 3.696 .6739 .3261
(Continued)
788 Appendix

TABLE T17.2 (Continued)

Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50
12 10 7.565 15.719 2.078 11.654 3.400 .6155 .3845
12 11 7.640 16.734 2.190 12.165 3.117 .5571 .4429
*12 12 7.690 17.782 2.312 12.668 2.846 .5000 .5000
13 0 4.771 6.792 1.423 5.845 7.287 .9971 .0029
13 1 5.425 7.870 1.451 6.721 6.918 .9907 .0093
13 2 5.951 8.816 1.482 7.464 6.546 .9792 .0208
13 3 6.401 9.707 1.517 8.141 6.178 .9615 .0385
13 4 6.794 10.573 1.556 8.775 5.814 .9369 .0631
13 5 7.139 11.432 1.601 9.379 5.458 .9054 .0946
13 6 7.440 12.294 1.652 9.960 5.109 .8673 .1327
13 7 7.699 13.168 1.710 10.524 4.769 .8233 .1767
13 8 7.917 14.061 1.776 11.073 4.439 .7745 .2255
13 9 8.095 14.979 1.850 11.610 4.119 .7219 .2781
13 10 8.235 15.926 1.934 12.137 3.810 .6670 .3330
13 11 8.340 16.905 2.027 12.655 3.513 .6109 .3891
13 12 8.415 17.916 2.129 13.165 3.228 .5549 .4451
*13 13 8.464 18.958 2.240 13.668 2.956 .5000 .5000
14 0 5.148 7.163 1.391 6.224 7.788 .9980 .0020
14 1 5.828 8.246 1.415 7.116 7.413 .9934 .0066
14 2 6.377 9.193 1.442 7.872 7.036 .9848 .0152
14 3 6.850 10.080 1.472 8.560 6.661 .9711 .0289
14 4 7.268 10.940 1.505 9.204 6.290 .9515 .0485
14 5 7.639 11.787 1.543 9.818 5.926 .9257 .0743
14 6 7.967 12.634 1.586 10.408 5.569 .8936 .1064
14 7 8.255 13.489 1.634 10.980 5.220 .8555 .1445
14 8 8.503 14.359 1.689 11.537 4.880 .8122 .1878
14 9 8.713 15.249 1.750 12.081 4.550 .7646 .2354
14 10 8.885 16.165 1.819 12.615 4.229 .7138 .2862
14 11 9.021 17.109 1.897 13.139 3.920 .6608 .3392
14 12 9.124 18.085 1.982 13.656 3.622 .6069 .3931
14 13 9.197 19.091 2.076 14.165 3.335 .5529 .4471
*14 14 9.246 20.128 2.177 14.668 3.062 .5000 .5000
15 0 5.524 7.535 1.364 6.603 8.289 .9986 .0014
15 1 6.228 8.623 1.385 7.509 7.909 .9953 .0047
15 2 6.799 9.571 1.408 8.278 7.526 .9889 .0111
15 3 7.295 10.456 1.433 8.976 7.145 .9784 .0216
15 4 7.735 11.311 1.462 9.631 6.769 .9629 .0371
15 5 8.129 12.150 1.495 10.253 6.398 .9419 .0581
15 6 8.482 12.985 1.531 10.852 6.033 .9151 .0849
15 7 8.796 13.824 1.572 11.432 5.676 .8826 .1174
15 8 9.073 14.674 1.617 11.996 5.327 .8447 .1553
15 9 9.312 15.541 1.669 12.547 4.987 .8022 .1978
15 10 9.515 16.429 1.727 13.088 4.656 .7558 .2442
15 11 9.682 17.342 1.791 13.619 4.336 .7065 .2935
(Continued)
Appendix 789

TABLE T17.2 (Continued)

Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50

15 12 9.814 18.285 1.863 14.141 4.026 .6553 .3447

15 13 9.914 19.257 1.942 14.657 3.727 .6032 .3968
15 14 9.987 20.260 2.029 15.165 3.439 .5512 .4488
*15 15 10.036 21.292 2.122 15.668 3.164 .5000 .5000
16 0 5.900 7.906 1.340 6.980 8.789 .9991 .0009
16 1 6.627 9.000 1.358 7.902 8.405 .9967 .0033
16 2 7.218 9.950 1.378 8.682 8.017 .9920 .0080
16 3 7.734 10.835 1.401 9.391 7.631 .9839 .0161
16 4 8.195 11.686 1.426 10.054 7.249 .9718 .0282
16 5 8.610 12.518 1.454 10.686 6.872 .9548 .0452
16 6 8.986 13.344 1.485 11.292 6.501 .9326 .0674
16 7 9.324 14.170 1.520 11.880 6.136 .9051 .0949
16 8 9.627 15.004 1.559 12.451 5.779 .8724 .1276
16 9 9.894 15.851 1.602 13.009 5.430 .8348 .1652
16 10 10.125 16.715 1.651 13.556 5.090 .7930 .2070
16 11 10.322 17.602 1.705 14.094 4.759 .7478 .2522
16 12 10.484 18.514 1.766 14.622 4.438 .6999 .3001
16 13 10.613 19.454 1.833 15.143 4.128 .6503 .3497
16 14 10.711 20.424 1.907 15.657 3.828 .5999 .4001
16 15 10.783 21.424 1.987 16.165 3.540 .5496 .4504
*16 16 10.832 22.452 2.073 16.668 3.263 .5000 .5000
17 0 6.276 8.278 1.319 7.358 9.290 .9994 .0006
17 1 7.024 9.378 1.335 8.292 8.901 .9977 .0023
17 2 7.635 10.330 1.353 9.084 8.509 .9942 .0058
17 3 8.169 11.215 1.373 9.803 8.118 .9881 .0119
17 4 8.649 12.064 1.395 10.476 7.731 .9786 .0214
17 5 9.084 12.891 1.419 11.115 7.348 .9650 .0350
17 6 9.481 13.710 1.446 11.730 6.971 .9468 .0532
17 7 9.841 14.526 1.476 12.325 6.600 .9237 .0763
17 8 10.167 15.346 1.509 12.903 6.235 .8957 .1043
17 9 10.460 16.176 1.547 13.468 5.879 .8629 .1371
17 10 10.718 17.020 1.588 14.021 5.530 .8257 .1743
17 11 10.943 17.883 1.634 14.564 5.190 .7847 .2153
17 12 11.133 18.768 1.686 15.098 4.859 .7405 .2595
17 13 11.291 19.679 1.743 15.625 4.538 .6939 .3061
17 14 11.417 20.618 1.806 16.145 4.227 .6458 .3542
17 15 11.514 21.585 1.875 16.658 3.926 .5969 .4031
17 16 11.585 22.582 1.949 17.165 3.637 .5481 .4519
*17 17 11.634 23.606 2.029 17.668 3.359 .5000 .5000
18 0 6.651 8.650 1.300 7.734 9.790 .9996 .0004
18 1 7.419 9.755 1.315 8.682 9.397 .9984 .0016
18 2 8.048 10.711 1.331 9.485 9.001 .9958 .0042
18 3 8.600 11.597 1.348 10.214 8.606 .9912 .0088
18 4 9.098 12.444 1.368 10.895 8.214 .9838 .0162
(Continued)
790 Appendix

TABLE T17.2 (Continued)

Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50
18 5 9.552 13.268 1.389 11.543 7.826 .9730 .0270
18 6 9.967 14.081 1.413 12.165 7.443 .9582 .0418
18 7 10.348 14.888 1.439 12.766 7.066 .9390 .0610
18 8 10.696 15.698 1.468 13.351 6.695 .9153 .0847
18 9 11.012 16.514 1.500 13.923 6.331 .8869 .1131
18 10 11.295 17.341 1.535 14.482 5.975 .8541 .1459
18 11 11.546 18.183 1.575 15.031 5.626 .8173 .1827
18 12 11.764 19.045 1.619 15.571 5.286 .7770 .2230
18 13 11.950 19.929 1.668 16.103 4.956 .7338 .2662
18 14 12.104 20.839 1.722 16.627 4.634 .6884 .3116
18 15 12.227 21.776 1.781 17.146 4.323 .6416 .3584
18 16 12.322 22.742 1.846 17.658 4.022 .5942 .4058
18 17 12.393 23.736 1.915 18.165 3.731 .5468 .4532
*18 18 12.442 24.756 1.990 18.668 3.453 .5000 .5000
19 0 7.027 9.022 1.284 8.111 10.291 .9997 .0003
19 1 7.813 10.133 1.297 9.071 9.894 .9988 .0012
19 2 8.459 11.092 1.311 9.884 9.494 .9970 .0030
19 3 9.028 11.980 1.327 10.622 9.095 .9935 .0065
19 4 9.542 12.826 1.344 11.312 8.699 .9878 .0122
19 5 10.013 13.648 1.363 11.967 8.306 .9792 .0208
19 6 10.447 14.456 1.384 12.597 7.918 .9673 .0327
19 7 10.847 15.257 1.407 13.205 7.535 .9515 .0485
19 8 11.215 16.057 1.432 13.797 7.159 .9315 .0685
19 9 11.551 16.862 1.460 14.374 6.788 .9072 .0928
19 10 11.857 17.674 1.491 14.939 6.425 .8786 .1214
19 11 12.132 18.500 1.525 15.494 6.068 .8459 .1541
19 12 12.376 19.341 1.563 16.039 5.720 .8095 .1905
19 13 12.589 20.202 1.605 16.577 5.380 .7699 .2301
19 14 12.770 21.085 1.651 17.107 5.049 .7276 .2724
19 15 12.921 21.994 1.702 17.630 4.728 .6834 .3166
19 16 13.042 22.930 1.758 18.147 4.416 .6378 .3622
19 17 13.136 23.894 1.819 18.659 4.114 .5917 .4083
19 18 13.205 24.885 1.884 19.166 3.823 .5455 .4545
*19 19 13.255 25.903 1.954 19.668 3.544 .5000 .5000
20 0 7.402 9.394 1.269 8.486 10.791 .9998 .0002
20 1 8.206 10.511 1.281 9.459 10.391 .9992 .0008
20 2 8.869 11.474 1.294 10.282 9.988 .9978 .0022
20 3 9.453 12.363 1.308 11.029 9.585 .9952 .0048
20 4 9.983 13.210 1.323 11.727 9.184 .9908 .0092
20 5 10.470 14.030 1.340 12.390 8.788 .9841 .0159
20 6 10.920 14.835 1.358 13.027 8.395 .9745 .0255
20 7 11.337 15.631 1.379 13.642 8.007 .9616 .0384
20 8 11.724 16.423 1.401 14.240 7.625 .9449 .0551
20 9 12.080 17.218 1.425 14.823 7.248 .9242 .0758
(Continued)
Appendix 791

TABLE T17.2 (Continued)

Unity Values for the QSS System
cN cT np.95 np.10 OR np.50 h0 PN at np.50 PT at np.50
20 10 12.407 18.019 1.452 15.394 6.878 .8995 .1005
20 11 12.705 18.829 1.482 15.954 6.515 .8707 .1293
20 12 12.973 19.653 1.515 16.505 6.159 .8382 .1618
20 13 13.210 20.494 1.551 17.047 5.811 .8022 .1978
20 14 13.418 21.354 1.591 17.582 5.472 .7633 .2367
20 15 13.595 22.237 1.636 18.110 5.141 .7219 .2781
20 16 13.742 23.145 1.684 18.632 4.819 .6787 .3213
20 17 13.861 24.080 1.737 19.148 4.507 .6343 .3657
20 18 13.953 25.041 1.795 19.659 4.205 .5893 .4107
20 19 14.023 26.030 1.856 20.166 3.913 .5444 .4556
*20 20 14.072 27.045 1.922 20.668 3.632 .5000 .5000
Source: Reprinted from Romboski, L.D., An investigation of quick switching acceptance sampling systems,
PhD dissertation, Rutgers—The State University, New Brunswick, NJ, 1969, pp. 90–95. With
permission.
* Indicates values for single-sampling plans.
792 Appendix

TABLE T17.3
Unity Values for the QSS-2(n; cN, cT) System
cN cT np.95 np.50 np.10 OR
1 0 0.2753 0.9481 2.3315 8.4689
2 1 0.7323 1.9424 3.9122 5.3423
2 0 0.5468 1.2177 2.3911 4.3729
3 2 1.2784 2.9391 5.3422 4.1788
3 1 1.1093 2.2266 3.9548 3.5651
5 4 2.5226 4.9352 8.0110 3.1757
4 2 1.7209 3.2264 5.3776 3.1249
3 0 0.8294 1.4943 2.4925 3.0052
6 5 3.1942 5.9339 9.2913 2.9088
5 3 2.3691 4.2244 6.7304 2.8409
4 1 1.4745 2.5212 4.0257 2.7302
8 7 4.6031 7.9320 11.7866 2.5606
5 2 2.1410 3.5267 5.4346 2.5383
7 5 3.7440 6.2200 9.3179 2.4888
9 8 5.3330 8.9312 13.0100 2.4395
6 3 2.8298 4.5265 6.7794 2.3957
4 0 1.1153 1.7747 2.6397 2.3668
7 4 3.5378 5.5246 8.0832 2.2848
6 2 2.5420 3.8346 5.5199 2.1715
9 6 5.0050 7.5193 10.6103 2.1218
7 3 3.2667 4.8379 6.8517 2.0974
8 4 4.0037 5.8374 8.1468 2.0348
5 0 1.4024 2.0574 2.8270 2.0158
9 5 4.7522 6.8353 9.4151 1.9812
7 2 2.9275 4.1469 5.6385 1.9260
8 3 3.6845 5.1551 6.9524 1.8869
9 4 4.4479 6.1572 8.2349 1.8514
6 0 1.6895 2.3413 3.0439 1.8017
7 1 2.5101 3.4287 4.4591 1.7765
8 2 3.3009 4.4617 5.7919 1.7546
9 3 4.0868 5.4759 7.0848 1.7336
7 0 1.9764 2.6259 3.2813 1.6602
8 1 2.8411 3.7336 4.6703 1.6438
8 0 2.2630 2.9109 3.5326 1.5610
9 1 3.1673 4.0385 4.9040 1.5483
9 0 2.5493 3.1960 3.7930 1.4879
10 0 2.8351 3.4812 4.0598 1.4320
11 0 3.1206 3.7662 4.3310 1.3879
12 0 3.4058 4.0511 4.6052 1.3522
Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl. Stat.,
17(1), 94, 1990. With permission.
Appendix 793

TABLE T17.4
Unity Values for the QSS-3(n; cN, cT) System
cN cT np.95 np.50 np.10 OR

1 0 0.2511 0.8435 2.3057 9.1824

2 1 0.7021 1.8283 3.8921 5.5435
2 0 0.4832 1.0343 2.3132 4.7972
3 2 1.2454 2.8150 5.3244 4.2753
3 1 1.0357 1.8283 3.8970 3.7627
4 2 1.6428 3.0088 5.3283 3.2434
5 4 2.4866 4.8147 7.9954 3.2154
5 3 2.2883 4.0040 6.6861 2.9219
4 1 1.3531 2.2390 3.9067 2.8872
7 6 3.8516 6.8111 10.5337 2.7349
6 4 2.9628 4.9938 7.9984 2.6996
5 2 2.0135 3.2256 5.3355 2.6499
8 7 4.5651 7.8098 11.7725 2.5788
7 5 3.6602 5.9886 9.2792 2.5352
4 0 0.9581 1.4676 2.3698 2.4734
9 8 5.2944 8.8088 12.9963 2.4547
8 6 4.3762 6.9843 10.5364 2.4077
5 1 1.6578 2.4741 3.9259 2.3681
6 2 2.3637 3.4623 5.3485 2.2628
7 3 3.0833 4.4481 6.7021 2.1735
8 4 3.8171 5.4344 8.0119 2.0989
5 0 1.1970 1.6953 2.4456 2.0431
6 1 1.9529 2.7186 3.9627 2.0291
7 2 2.6985 3.7118 5.3720 1.9907
8 3 3.4493 4.6981 6.7193 1.9480
9 4 4.3082 5.6830 8.0257 1.9072
7 1 2.2407 2.9685 4.0284 1.7978
9 3 3.8000 4.9582 6.7486 1.7759
7 0 1.6742 2.1593 2.7218 1.6257
9 1 2.7998 3.4753 4.2720 1.5258
10 0 2.3872 2.8634 3.3128 1.3877
11 0 2.2624 3.0988 3.5296 1.3450
12 0 2.8609 3.3343 3.7512 1.3112

Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl. Stat.,

17(1), 96, 1990. With permission.
794 Appendix

TABLE T17.5
Unity Values for the QSS-4(n; cN, cT) System
cN cT np.95 np.50 np.10 OR

1 0 0.2322 0.7819 2.3029 9.9177

2 1 0.6769 1.7626 3.8899 5.7466
2 0 0.4350 0.9205 2.3037 5.2959
3 2 1.2170 2.7551 5.3225 4.3735
3 1 0.9784 1.8924 3.8905 3.9764
3 0 0.6417 1.0896 2.3057 3.5931
4 2 1.5801 2.8763 5.3229 3.3687
5 4 2.4549 4.7482 7.9938 3.2563
4 1 1.2625 2.0595 3.8915 3.0824
5 3 2.2222 3.8658 6.6813 3.0066
6 5 3.1246 5.7463 9.2748 2.9683
5 2 1.9156 3.0350 5.3237 2.7791
7 6 3.8181 6.7448 10.5322 2.7585
4 0 0.8493 1.2744 2.3110 2.7211
8 7 4.5309 7.7436 11.7711 2.5980
6 3 2.5956 4.0164 6.6819 2.5743
5 1 1.5341 2.2512 3.8937 2.5381
9 8 5.2598 8.7427 12.9949 2.4706
8 6 4.3044 6.8480 10.5325 2.4469
7 4 3.2975 5.0021 7.9946 2.4244
6 2 2.2310 3.2224 5.3251 2.3869
7 3 2.9445 4.1973 6.6830 2.2697
9 6 4.7522 6.9814 10.5329 2.2164
6 1 1.7964 2.4574 3.8985 2.1702
7 2 2.5316 3.4293 5.3279 2.1046
8 3 3.2749 4.4007 6.6849 2.0413
9 4 4.0280 5.3748 7.9970 1.9854
7 1 2.0517 2.6718 3.9097 1.9056
8 2 2.8210 3.6485 5.3335 1.8906
8 1 2.3014 2.8909 3.9349 1.7098
9 1 2.5466 3.1124 3.9869 1.5656
8 0 1.6776 2.0665 2.5574 1.5244
9 0 1.8839 2.2691 2.7046 1.4356
10 0 2.0898 2.4722 2.8705 1.3736
11 0 2.2955 2.6758 3.0480 1.3278
12 0 2.5008 2.8795 3.2330 1.2928
Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl.
Stat., 17(1), 98, 1990. With permission.
TABLE T17.6
Unity Values for the QSS-2(n; kn, c0) System
Appendix

p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

0 2 24.84 32.05 49.11 0.047 0.051 117.66 151.83 232.64 0.010 0.010
1 2 6.18 7.46 10.40 0.319 0.349 13.64 16.47 22.95 0.145 0.147
2 2 3.74 4.37 5.81 0.724 0.790 6.42 7.51 9.98 0.422 0.430
3 2 2.85 3.26 4.20 1.196 1.303 4.31 4.94 6.36 0.790 0.805
4 2 2.39 2.70 3.40 1.708 1.856 3.35 3.78 4.76 1.219 1.242
5 2 2.12 2.36 2.92 2.245 2.437 2.81 3.14 3.88 1.689 1.721
6 2 1.93 2.13 2.60 2.802 3.035 2.47 2.73 3.33 2.190 2.231
7 2 1.80 1.97 2.37 3.371 3.647 2.23 2.45 2.95 2.714 2.763
8 2 1.70 1.85 2.20 3.952 4.269 2.06 2.24 2.68 3.256 3.314
9 2 1.62 1.75 2.07 4.541 4.899 1.93 2.09 2.47 3.811 3.878
10 2 1.56 1.67 1.96 5.136 5.534 1.83 1.96 2.30 4.378 4.454
0 3 18.11 23.18 35.39 0.043 0.052 81.02 103.73 158.34 0.010 0.010
1 3 4.68 5.58 7.73 0.287 0.338 9.60 11.46 15.87 0.140 0.145
2 3 2.91 3.35 4.41 0.637 0.745 4.63 5.34 7.03 0.399 0.414
3 3 2.26 2.55 3.24 1.035 1.203 3.19 3.58 4.56 0.735 0.762
4 3 1.94 2.14 2.66 1.459 1.686 2.53 2.80 3.47 1.117 1.156
5 3 1.74 1.90 2.31 1.900 2.184 2.17 2.36 2.87 1.526 1.578
6 3 1.61 1.74 2.07 2.352 2.692 1.94 2.09 2.49 1.955 2.018
7 3 1.52 1.62 1.90 2.812 3.204 1.78 1.90 2.24 2.396 2.472
8 3 1.45 1.54 1.78 3.278 3.721 1.67 1.77 2.05 2.848 2.934
9 3 1.40 1.47 1.68 3.749 4.241 1.58 1.67 1.91 3.306 3.403
10 3 1.35 1.42 1.60 4.222 4.763 1.52 1.59 1.80 3.771 3.878
0 4 14.68 18.67 28.38 0.041 0.052 62.74 79.78 121.31 0.010 0.010
1 4 3.94 4.65 6.40 0.260 0.327 7.64 9.02 12.40 0.134 0.142
2 4 2.52 2.86 3.72 0.566 0.703 3.80 4.31 5.62 0.375 0.396
3 4 2.00 2.21 2.77 0.909 1.115 2.68 2.97 3.72 0.678 0.713
4 4 1.74 1.89 2.29 1.271 1.543 2.18 2.37 2.88 1.014 1.063
(Continued)
795
 796

TABLE T17.6 (Continued)

Unity Values for the QSS-2(n; kn, c0) System
p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

5 4 1.58 1.70 2.01 1.645 1.980 1.90 2.04 2.41 1.370 1.433
6 4 1.48 1.57 1.82 2.026 2.421 1.72 1.83 2.12 1.738 1.814
7 4 1.41 1.48 1.68 2.412 2.865 1.60 1.69 1.92 2.115 2.203
8 4 1.35 1.41 1.58 2.802 3.312 1.52 1.59 1.78 2.497 2.597
9 4 1.31 1.36 1.51 3.195 3.760 1.45 1.51 1.67 2.884 2.995
10 4 1.28 1.32 1.45 3.590 4.211 1.40 1.45 1.58 3.275 3.397

Source: Reprinted from Kuralmani, V. and Govindaraju, K., J. Appl. Stat., 19(2), 264, 1992. With permission.
Appendix
TABLE T17.7
Unity Values for the QSS-3(n; kn, c0) System
Appendix

p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

0 2 25.50 33.14 50.94 0.045 0.051 117.62 152.87 234.96 0.010 0.010
1 2 6.35 7.74 10.83 0.307 0.347 13.63 16.60 23.23 0.143 0.147
2 2 3.84 4.53 6.05 0.695 0.783 6.42 7.58 10.12 0.415 0.427
3 2 2.92 3.38 4.38 1.146 1.286 4.31 4.99 6.47 0.777 0.798
4 2 2.45 2.80 3.55 1.635 1.828 3.35 3.83 4.85 1.195 1.228
5 2 2.16 2.45 3.05 2.149 2.394 2.81 3.18 3.96 1.654 1.698
6 2 1.97 2.21 2.72 2.681 2.977 2.47 2.77 3.40 2.141 2.197
7 2 1.83 2.04 2.48 3.227 3.572 2.23 2.48 3.02 2.649 2.716
8 2 1.73 1.91 2.30 3.782 4.175 2.06 2.28 2.74 3.174 3.253
9 2 1.64 1.81 2.16 4.345 4.785 1.92 2.12 2.53 3.712 3.801
10 2 1.58 1.73 2.05 4.916 5.400 1.82 1.99 2.36 4.261 4.360
0 3 18.86 24.49 37.63 0.041 0.052 81.00 105.16 161.59 0.010 0.010
1 3 4.86 5.91 8.27 0.268 0.334 9.57 11.63 16.27 0.136 0.144
2 3 3.01 3.54 4.72 0.594 0.729 4.62 5.44 7.26 0.386 0.407
3 3 2.33 2.68 3.47 0.965 1.169 3.17 3.66 4.73 0.707 0.743
4 3 1.98 2.25 2.84 1.361 1.632 2.52 2.86 3.62 1.070 1.121
5 3 1.77 1.98 2.46 1.773 2.105 2.15 2.41 3.00 1.458 1.523
6 3 1.62 1.80 2.21 2.196 2.586 1.91 2.12 2.61 1.863 1.942
7 3 1.52 1.67 2.03 2.627 3.071 1.75 1.93 2.34 2.281 2.372
8 3 1.45 1.58 1.89 3.064 3.559 1.64 1.79 2.14 2.707 2.810
9 3 1.39 1.50 1.79 3.505 4.049 1.55 1.68 1.99 3.140 3.254
10 3 1.34 1.44 1.70 3.950 4.541 1.48 1.59 1.88 3.578 3.702
0 4 15.45 20.04 30.78 0.037 0.052 62.15 80.58 123.80 0.009 0.010
1 4 4.11 4.99 6.97 0.238 0.321 7.60 9.21 12.88 0.129 0.139
2 4 2.59 3.04 4.05 0.518 0.681 3.77 4.42 5.89 0.357 0.384
3 4 2.04 2.33 3.01 0.833 1.071 2.65 3.04 3.92 0.640 0.685
4 4 1.75 1.97 2.49 1.166 1.473 2.14 2.41 3.04 0.954 1.014
797

(Continued)
 798

TABLE T17.7 (Continued)

Unity Values for the QSS-3(n; kn, c0) System
p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

5 4 1.58 1.75 2.17 1.511 1.880 1.86 2.06 2.55 1.285 1.360
6 4 1.47 1.60 1.96 1.862 2.291 1.68 1.83 2.24 1.628 1.716
7 4 1.39 1.50 1.80 2.220 2.703 1.56 1.68 2.02 1.978 2.078
8 4 1.33 1.42 1.69 2.581 3.118 1.48 1.57 1.86 2.334 2.445
9 4 1.29 1.36 1.59 2.946 3.533 1.41 1.49 1.74 2.695 2.816
10 4 1.26 1.32 1.52 3.312 3.951 1.36 1.43 1.65 3.058 3.189

Source: Reprinted from Kuralmani, V. and Govindaraju, K., J. Appl. Stat., 19(2), 267, 1992. With permission.
Appendix
TABLE T17.8
Unity Values for the QSS-4(n; kn, c0) System
Appendix

p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

0 2 26.35 34.28 52.69 0.044 0.051 118.70 154.42 237.38 0.010 0.010
1 2 6.57 8.01 11.21 0.296 0.346 13.77 16.79 23.49 0.141 0.147
2 2 3.97 4.69 6.27 0.671 0.777 6.50 7.69 10.26 0.410 0.425
3 2 3.02 3.50 4.54 1.107 1.274 4.37 5.07 6.57 0.765 0.792
4 2 2.53 2.90 3.67 1.579 1.808 3.40 3.90 4.94 1.175 1.216
5 2 2.23 2.53 3.16 2.076 2.365 2.86 3.24 4.04 1.624 1.679
6 2 2.03 2.29 2.81 2.591 2.937 2.51 2.82 3.47 2.100 2.168
7 2 1.89 2.11 2.57 3.118 3.520 2.27 2.53 3.08 2.596 2.678
8 2 1.78 1.97 2.38 3.656 4.110 2.09 2.32 2.80 3.108 3.203
9 2 1.69 1.87 2.24 4.201 4.706 1.96 2.16 2.59 3.632 3.740
10 2 1.62 1.78 2.12 4.754 5.307 1.85 2.04 2.42 4.167 4.286
0 3 19.89 25.87 39.77 0.039 0.052 82.55 107.38 165.06 0.009 0.010
1 3 5.12 6.25 8.74 0.253 0.331 9.77 11.92 16.68 0.133 0.142
2 3 3.16 3.74 4.99 0.562 0.718 4.73 5.60 7.47 0.375 0.401
3 3 2.44 2.83 3.66 0.914 1.147 3.26 3.78 4.89 0.685 0.728
4 3 2.07 2.36 3.00 1.291 1.595 2.58 2.95 3.75 1.033 1.094
5 3 1.84 2.08 2.60 1.683 2.053 2.20 2.49 3.11 1.405 1.482
6 3 1.69 1.89 2.33 2.087 2.517 1.96 2.20 2.71 1.794 1.886
7 3 1.58 1.75 2.14 2.498 2.984 1.79 2.00 2.43 2.195 2.299
8 3 1.49 1.65 1.99 2.915 3.454 1.67 1.85 2.23 2.604 2.720
9 3 1.43 1.57 1.88 3.338 3.925 1.58 1.73 2.07 3.019 3.147
10 3 1.37 1.50 1.78 3.764 4.397 1.50 1.64 1.95 3.440 3.577
0 4 16.50 21.46 32.99 0.035 0.052 63.99 83.22 127.92 0.009 0.010
1 4 4.39 5.34 7.48 0.222 0.317 7.82 9.53 13.34 0.124 0.137
2 4 2.75 3.25 4.34 0.484 0.667 3.90 4.60 6.14 0.342 0.374
3 4 2.15 2.49 3.22 0.780 1.043 2.74 3.17 4.10 0.612 0.663
(Continued)
799
 800

TABLE T17.8 (Continued)

Unity Values for the QSS-4(n; kn, c0) System
p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
c0 k β = 0.10 β = 0.05 β = 0.01 α = 0.05 p1ASN(p1) β = 0.10 β = 0.05 β = 0.01 α = 0.01 p1ASN(p1)

4 4 1.83 2.09 2.65 1.094 1.428 2.20 2.52 3.19 0.910 0.979
5 4 1.64 1.85 2.31 1.419 1.818 1.90 2.15 2.68 1.224 1.308
6 4 1.51 1.69 2.08 1.752 2.209 1.71 1.91 2.35 1.549 1.647
7 4 1.42 1.57 1.91 2.091 2.602 1.58 1.75 2.13 1.882 1.991
8 4 1.35 1.48 1.79 2.433 2.997 1.48 1.63 1.96 2.221 2.341
9 4 1.30 1.42 1.69 2.779 3.393 1.41 1.53 1.83 2.564 2.693
10 4 1.26 1.36 1.61 3.128 3.789 1.36 1.46 1.73 2.909 3.049
Source: Reprinted from Kuralmani, V. and Govindaraju, K., J. Appl. Stat., 19(2), 270, 1992. With permission.
Appendix
Appendix 801

TABLE T17.9
Unity Values for the QSS-1(n; kn, 0) System
p2/p1 for p2/p1 for p2/p1 for np1 for p2/p1 for p2/p1 for p2/p1 for np1 for
α = 0.05, α = 0.05, α = 0.05, Pa(p1) = 0.95, α = 0.01, α = 0.01, α = 0.01, Pa(p1) = 0.99,
k β = 0.10 β = 0.05 β = 0.01 α = 0.05 β = 0.10 β = 0.05 β = 0.01 α = 0.01

1.0 44.894 58.407 89.787 0.05129 229.114 298.080 458.229 0.010050

1.1 41.426 53.636 82.116 0.05104 210.598 272.669 417.450 0.010040
1.2 38.547 49.683 75.747 0.05079 195.194 251.585 383.569 0.010030
1.3 36.118 46.353 70.380 0.05054 182.170 233.802 354.990 0.010020
1.4 34.030 43.501 65.781 0.05030 170.999 218.591 330.549 0.010010
1.5 32.221 41.039 61.812 0.05006 161.300 205.440 309.430 0.010000
1.6 30.638 38.890 58.352 0.04982 152.777 193.925 290.972 0.009991
1.7 29.236 36.991 55.297 0.04959 145.256 183.789 274.742 0.009981
1.8 27.986 35.306 52.591 0.04936 138.542 174.777 250.345 0.009971
1.9 26.867 33.800 50.177 0.04913 132.517 166.710 247.485 0.009961
2.0 25.859 32.448 48.012 0.04890 127.073 159.451 235.936 0.009951
2.1 24.940 31.218 46.050 0.04858 122.118 152.857 225.478 0.009942
2.2 24.102 30.101 44.270 0.04846 117.600 146.869 215.999 0.009932
2.3 23.333 29.076 42.638 0.04825 113.465 141.393 207.347 0.009922
2.4 22.625 28.135 41.149 0.04804 109.644 136.346 199.415 0.009913
2.5 21.974 27.272 39.780 0.04783 106.129 131.718 192.134 0.009903
2.6 21.369 26.474 38.520 0.04762 102.861 127.434 185.414 0.009893
2.7 20.810 25.735 37.355 0.04741 99.818 123.442 179.178 0.009884
2.8 20.284 25.043 36.268 0.04721 96.982 119.739 173.405 0.009874
2.9 19.796 24.401 35.258 0.04701 94.334 116.280 168.018 0.009865
3.0 19.338 23.800 34.315 0.04681 91.852 113.049 162.993 0.009855
3.2 18.497 22.701 32.602 0.04543 87.312 107.157 153.894 0.009836
3.4 17.755 21.733 31.092 0.04605 83.284 101.946 145.849 0.009817
3.6 17.086 20.867 29.748 0.04568 79.651 97.275 138.677 0.009799
3.8 16.485 20.086 28.544 0.04532 76.391 93.078 132.270 0.009780
4.0 15.943 19.386 27.462 0.04496 73.428 89.285 126.480 0.009762
4.2 15.444 18.743 26.475 0.04462 70.728 85.836 121.246 0.009743
4.4 14.991 18.157 25.580 0.04428 68.257 82.674 116.473 0.009725
4.6 14.571 17.620 24.760 0.04395 65.973 79.777 112.105 0.009707
4.8 14.185 17.126 24.004 0.04363 63.877 77.118 108.092 0.009689
5.0 13.828 16.668 23.311 0.04331 61.927 74.646 104.395 0.009671
Source: Reprinted from Govindaraju, K., Commun. Stat. Simul. Computat., 20(1), 161, 1991. With permission.
802 Appendix

TABLE T17.10
Unity Values for the QSS-2(n; kn, 0) System
k OR for α = 0.05, β = 0.10 n1p1 for α = 0.05, β = 0.10

1.1 41.2480 0.0508

1.2 38.2286 0.0503
1.3 35.6112 0.0499
1.4 33.4393 0.0494
1.5 31.5041 0.0490
1.6 29.8742 0.0485
1.7 28.3846 0.0481
1.8 27.0650 0.0477
1.9 25.8901 0.0473
2.0 24.8358 0.0469
2.1 23.8860 0.0465
2.2 22.9762 0.0462
2.3 20.8320 0.0488
2.4 21.4846 0.0454
2.5 20.8333 0.0450
2.6 20.1902 0.0447
2.7 19.5968 0.0444
2.8 19.0476 0.0441
2.9 18.5388 0.0438
3.0 18.0667 0.0435
3.1 17.6659 0.0431
3.2 17.2523 0.0428
3.3 16.8682 0.0425
3.4 16.5071 0.0422
3.5 16.1286 0.0420
3.6 15.8106 0.0417
3.7 15.5121 0.0414
3.8 15.1942 0.0412
3.9 14.9291 0.0409
4.0 14.6798 0.0406
4.2 14.1845 0.0401
4.4 13.7399 0.0396
4.6 13.3036 0.0392
4.8 12.9406 0.0387
5.0 12.5770 0.0383
5.4 11.9733 0.0374
5.8 11.4317 0.0366
6.2 10.9443 0.0359
Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl.
Stat., 17(1), 106, 1990. With permission.
Appendix 803

TABLE T17.11
Unity Values for the QSS-3(n; kn, 0) System
k OR for α = 0.05, β = 0.10 n1p1 for α = 0.05, β = 0.10

1.1 41.3735 0.0506

1.2 38.4609 0.0499
1.3 36.0122 0.0492
1.4 33.8560 0.0486
1.5 31.9979 0.0480
1.6 30.3819 0.0474
1.7 28.9658 0.0468
1.8 27.6544 0.0463
1.9 26.5470 0.0457
2.0 25.5022 0.0452
2.1 24.5638 0.0447
2.2 23.7149 0.0442
2.3 22.9474 0.0437
2.4 22.1963 0.0433
2.5 21.5607 0.0428
2.6 20.9292 0.0424
2.7 20.3500 0.0420
2.8 19.8149 0.0416
2.9 19.3204 0.0412
3.0 18.8603 0.0408
3.1 18.4356 0.0404
3.2 18.0400 0.0400
3.3 17.6297 0.0397
3.4 17.2875 0.0393
3.5 16.9256 0.0390
3.6 16.5840 0.0387
3.7 16.3081 0.0383
3.8 16.0053 0.0380
3.9 15.7215 0.0377
4.0 15.4545 0.0374
4.2 14.9620 0.0368
4.4 14.4821 0.0363
4.6 14.0896 0.0357
4.8 13.6989 0.0352
5.0 13.3429 0.0347
5.4 12.7300 0.0337
5.8 12.1829 0.0328
6.2 11.6875 0.0320
6.6 11.2660 0.0312
7.0 10.8721 0.0305

Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl.

Stat., 17(1), 108, 1990. With permission.
804 Appendix

TABLE T17.12
Unity Values for the QSS-4(n; kn, 0) System
k OR for α = 0.05, β = 0.10 n1p1 for α = 0.05, β = 0.10

1.1 41.5337 0.0504

1.2 38.7657 0.0495
1.3 36.3717 0.0487
1.4 34.3382 0.0479
1.5 32.5924 0.0471
1.6 31.0172 0.0464
1.7 29.6411 0.0457
1.8 28.4289 0.0450
1.9 27.3589 0.0443
2.0 26.3478 0.0437
2.1 25.4432 0.0431
2.2 24.5728 0.0426
2.3 23.8405 0.0420
2.4 23.1229 0.0415
2.5 22.4683 0.0410
2.6 21.8716 0.0405
2.7 21.3784 0.0399
2.8 20.8253 0.0395
2.9 20.3120 0.0391
3.0 19.8886 0.0386
3.1 19.4503 0.0382
3.2 19.0423 0.0378
3.3 18.0829 0.0386
3.4 17.9233 0.0378
3.5 17.9809 0.0366
3.6 17.6740 0.0362
3.7 17.3426 0.0359
3.8 17.0761 0.0355
3.9 16.7784 0.0352
4.0 16.5014 0.0349
4.2 16.0351 0.0342
4.4 15.5833 0.0336
4.6 15.1758 0.0330
4.8 14.7692 0.0325
5.0 14.4451 0.0319
5.4 13.8091 0.0309
5.8 13.2433 0.0300
6.0 12.9730 0.0296
6.6 12.2958 0.0284
7.2 11.7252 0.0273
8.5 10.76190 0.0252
Source: Reprinted from Soundararajan, V. and Arumainayagam, S.D., J. Appl.
Stat., 17(1), 108, 1990. With permission.
Appendix 805

TABLE T17.13
MIL-STD-1916 Table I—Code Letters (CL) for Entry into the Sampling Tables
Verification Levels
Lot or Production
Interval Size VII VI V IV III II I

2–170 A A A A A A A
171–288 A A A A A A B
289–544 A A A A A B C
545–960 A A A A B C D
961–1,632 A A A B C D E
1,633–3,072 A A B C D E E
3,073–5,440 A B C D E E E
5,441–9,216 B C D E E E E
9,217–17,408 C D E E E E E
17,409–30,720 D E E E E E E
30,721 and larger E E E E E E E
Source: United States Department of Defense, Department of Defense Test Method Standard, DOD Preferred
Methods for Acceptance of Product, MIL-STD-1916, U.S. Government Printing Office, Washington,
DC, 1996, p. 15.
806 Appendix

TABLE T17.14
MIL-STD-1916 Table II—Attributes Sampling Plans
Verification Levels
Code
Letter T VII VI V IV III II I R
Sample size (na)
A 3072 1280 512 192 80 32 12 5 3
B 4096 1536 640 256 96 40 16 6 3
C 5120 2048 768 320 128 48 20 8 3
D 6144 2560 1024 384 160 64 24 10 4
E 8192 3072 1280 512 192 80 32 12 5
Source: United States Department of Defense, Department of Defense Test Method Standard, DOD
Preferred Methods for Acceptance of Product, MIL-STD-1916, U.S. Government Printing Office,
Washington, DC, 1996, p. 17.
Notes: 1. When the lot size is less than or equal to the sample size, 100% attributes inspection is
required.
2. One verification level (VL) to the left/right of the specified normal VL is the respective tight-
ened/reduced plan. Tightened inspection of VL-VII is T, reduced inspection of VL-I is R.
Appendix 807

TABLE T17.15
MIL-STD-1916 Table III—Variables Sampling Plans
Verification Levels
Code Letter T VII VI V IV III II I R
Sample size (nv)
A 113 87 64 44 29 18 9 4 2
B 122 92 69 49 32 20 11 5 2
C 129 100 74 54 37 23 13 7 2
D 136 107 81 58 41 26 15 8 3
E 145 113 87 64 44 29 18 9 4
k values (one- or two-sided)
A 3.51 3.27 3.00 2.69 2.40 2.05 1.64 1.21 1.20
B 3.58 3.32 3.07 2.79 2.46 2.14 1.77 1.33 1.20
C 3.64 3.40 3.12 2.86 2.56 2.21 1.86 1.45 1.20
D 3.69 3.46 3.21 2.91 2.63 2.32 1.93 1.56 1.20
E 3.76 3.51 3.27 3.00 2.69 2.40 2.05 1.64 1.21
F values (one- or two-sided)
A .136 .145 .157 .174 .193 .222 .271 .370 .707
B .134 .143 .154 .168 .188 .214 .253 .333 .707
C .132 .140 .152 .165 .182 .208 .242 .301 .707
D .130 .138 .148 .162 .177 .199 .233 .283 .435
E .128 .136 .145 .157 .174 .193 .222 .271 .370
Source: United States Department of Defense, Department of Defense Test Method Standard, DOD Preferred Methods
for Acceptance of Product, MIL-STD-1916, U.S. Government Printing Office, Washington, DC, 1996, p. 19.
Notes: 1. When the lot size is less than or equal to the sample size, 100% attributes inspection is required.
2. One verification level (VL) to the left/right of the specified normal VL is the respective tightened/
reduced plan. Tightened inspection of VL-VII is T, reduced inspection of VL-I is R.
808 Appendix

TABLE T17.16
MIL-STD-1916 Table IV—Continuous Sampling Plans
Verification Levels
Code
Letter T VII VI V IV III II I R
Screening phase: Clearance numbers (i)
A 3867 2207 1134 527 264 125 55 27 NA
B 7061 3402 1754 842 372 180 83 36 NA
C 11337 5609 2524 1237 572 246 116 53 NA
D 16827 8411 3957 1714 815 368 155 73 NA
E 26912 11868 5709 2605 1101 513 228 96 NA
Sampling phase: Frequencies (f)
A 1/3 4/17 1/6 2/17 1/12 1/17 1/24 1/34 1/48
B 4/17 1/6 2/17 1/12 1/17 1/24 1/34 1/48 1/68
C 1/6 2/17 1/12 1/17 1/24 1/34 1/48 1/68 1/96
D 2/17 1/12 1/17 1/24 1/34 1/48 1/68 1/96 1/136
E 1/12 1/17 1/24 1/34 1/48 1/68 1/96 1/136 1/192
Source: United States Department of Defense, Department of Defense Test Method Standard, DOD Preferred
Methods for Acceptance of Product, MIL-STD-1916, U.S. Government Printing Office, Washington, DC,
1996, p. 20.
Notes: 1. Use of other i and f combinations are permitted provided they are computed in accordance with
Appendix, paragraph 30.5.
2. During the screening phase, one verification level (VL) to the left of the specified normal VL is the
tightened plan. Tightened inspection of VL-VII is T. There is no reduced plan while in the screening
phase. During the sampling phase, one verification level (VL) to the left/right of the specified nor-
mal VL is the respective tightened/reduced plan. Tightened inspection of VL-VII is T, reduced
inspection of VL-I is R.
3. Sample units shall be chosen with frequency (f) so as to give each unit of product an equal chance
of being inspected. The inspector should allow the interval between sample units to vary
somewhat rather than draw sample units according to a rigid pattern.
Appendix 809

TABLE T17.17
Hα Values for Simplified Grand Lot Sampling
k 0 1 2 3 4 5 6 7 8 9
α = .002
0 0. 0. 2.19 2.78 3.01 3.17 3.28 3.36 3.48 3.48
10 3.53 3.57 3.60 3.64 3.66 3.69 3.71 3.74 3.76 3.77
20 3.79 3.81 3.82 3.84 3.85 3.86 3.88 3.89 3.90 3.91
30 3.92 3.93 3.94 3.95 3.96 3.97 3.97 3.98 3.99 4.00
40 4.00 4.01 4.02 4.02 4.03 4.04 4.04 4.05 4.06 4.06
50 4.07 4.07 4.08 4.08 4.09 4.09 4.10 4.10 4.11 4.11
60 4.11 4.12 4.12 4.13 4.13 4.14 4.14 4.14 4.15 4.15
70 4.15 4.16 4.16 4.17 4.17 4.17 4.18 4.18 4.18 4.19
80 4.19 4.19 4.19 4.20 4.20 4.20 4.21 4.21 4.21 4.21
90 4.22 4.22 4.22 4.23 4.23 4.23 4.23 4.24 4.24 4.24
100 4.24 4.25 4.25 4.25 4.25 4.26 4.26 4.26 4.26 4.26
110 4.27 4.27 4.27 4.27 4.28 4.28 4.28 4.28 4.28 4.29
120 4.29 4.29 4.29 4.29 4.30 4.30 4.30 4.30 4.30 4.30
130 4.31 4.31 4.31 4.31 4.31 4.32 4.32 4.32 4.32 4.32
140 4.32 4.33 4.33 4.33 4.33 4.33 4.33 4.34 4.34 4.34
150 4.34 4.34 4.34 4.34 4.35 4.35 4.35 4.35 4.35 4.35
160 4.35 4.36 4.36 4.36 4.36 4.36 4.36 4.36 4.37 4.37
170 4.37 4.37 4.37 4.37 4.37 4.38 4.38 4.38 4.38 4.38
180 4.38 4.38 4.38 4.39 4.39 4.39 4.39 4.39 4.39 4.39
190 4.39 4.40 4.40 4.40 4.40 4.40 4.40 4.40 4.40 4.40
200 4.41 4.41 4.41 4.41 4.41 4.41 4.41 4.41 4.41 4.42
α = .05
0 0. 0. 1.39 1.96 2.16 2.30 2.41 2.49 2.56 2.61
10 2.66 2.71 2.74 2.78 2.81 2.84 2.86 2.89 2.91 2.93
20 2.95 2.97 2.98 3.00 3.01 3.03 3.04 3.06 3.07 3.08
30 3.09 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18
40 3.19 3.19 3.20 3.21 3.22 3.22 3.23 3.24 3.24 3.25
50 3.26 3.26 3.27 3.28 3.28 3.29 3.29 3.30 3.30 3.31
60 3.31 3.32 3.32 3.33 3.33 3.34 3.34 3.35 3.35 3.36
70 3.36 3.36 3.37 3.37 3.38 3.38 3.38 3.39 3.39 3.40
80 3.40 3.40 3.41 3.41 3.41 3.42 3.42 3.42 3.43 3.43
90 3.43 3.44 3.44 3.44 3.45 3.45 3.45 3.45 3.46 3.46
100 3.46 3.47 3.47 3.47 3.47 3.48 3.48 3.48 3.49 3.49
110 3.49 3.49 3.50 3.50 3.50 3.50 3.51 3.51 3.51 3.51
120 3.51 3.52 3.52 3.52 3.52 3.53 3.53 3.53 3.53 3.53
130 3.54 3.54 3.54 3.54 3.55 3.55 3.55 3.55 3.55 3.56
140 3.56 3.56 3.56 3.56 3.56 3.57 3.57 3.57 3.57 3.57
150 3.58 3.58 3.58 3.58 3.58 3.58 3.59 3.59 3.59 3.59
160 3.59 3.60 3.60 3.60 3.60 3.60 3.60 3.61 3.61 3.61
170 3.61 3.61 3.61 3.61 3.62 3.62 3.62 3.62 3.62 3.62
180 3.63 3.63 3.63 3.63 3.63 3.63 3.63 3.64 3.64 3.64
190 3.64 3.64 3.64 3.64 3.65 3.65 3.65 3.65 3.65 3.65
200 3.65 3.65 3.66 3.66 3.66 3.66 3.66 3.66 3.66 3.66
Source: Reprinted from Schilling, E.G., J. Qual. Technol., 11(3), 119, 1979. With permission; Computed as in
Schilling (1973b).
810 Appendix

Maximum Sample
% defective size
10.0 1000
950
9.0 900
8.0 850
800
7.0 750
700
6.0
650
5.0 600
550
4.0 500
480
3.5 460
440
420
3.0 400
Confidence 380
0.9999 360
2.5 340
0.999 320
2.0 300
280
1.8 0.99 260
1.6 240
1.4 0.95 270
1.2 200
0.90
190
180
1.0 170
0.80
0.90 160
0.70 150
0.80
140
0.70 0.60 130
0.60 0.50 120
110
0.50 100
95
90
0.40 85
80
0.35
75
0.30 70
65
0.25 60
55
0.20 50
48
0.18 46
44
0.16 42
40
0.14 38
36
0.12 34
32
0.10 30

FIGURE F17.18
Nomograph for samples having zero defectives. (From Reprinted from Nelson, L.S., J. Qual. Technol., 10(1), 43,
1978. With permission.)
TABLE T18.1
Hazard Values Corresponding to Probability Plotting Positions for Censored Data
Probability Tenth of Percent
Appendix

Probability Percent 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0. 0. 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90

1. 1.01 1.11 1.21 1.31 1.41 1.51 1.61 1.71 1.82 1.92
2. 2.02 2.12 2.22 2.33 2.43 2.53 2.63 2.74 2.84 2.94
3. 3.05 3.15 3.25 3.36 3.46 3.56 3.67 3.77 3.87 3.98
4. 4.08 4.19 4.29 4.40 4.50 4.60 4.71 4.81 4.92 5.02
5. 5.13 5.23 5.34 5.45 5.55 5.66 5.76 5.87 5.97 6.08
6. 6.19 6.29 6.40 6.51 6.61 6.72 6.83 6.94 7.04 7.15
7. 7.26 7.36 7.47 7.58 7.69 7.80 7.90 8.01 8.12 8.23
8. 8.34 8.45 8.56 8.66 8.77 8.88 8.99 9.10 9.21 9.32
9. 9.43 9.54 9.65 9.76 9.87 9.98 10.09 10.20 10.31 10.43
10. 10.54 10.65 10.76 10.87 10.98 11.09 11.20 11.32 11.43 11.54
11. 11.65 11.77 11.88 11.99 12.10 12.22 12.33 12.44 12.56 12.67
12. 12.78 12.90 13.01 13.12 13.24 13.35 13.47 13.58 13.70 13.81
13. 13.93 14.04 14.16 14.27 14.39 14.50 14.62 14.73 14.85 14.97
14. 15.08 15.20 15.32 15.43 15.55 15.67 15.78 15.90 16.02 16.13
15. 16.25 16.37 16.49 16.61 16.72 16.84 16.96 17.08 17.20 17.32
16. 17.44 17.55 17.67 17.79 17.91 18.03 18.15 18.27 18.39 18.51
17. 18.63 18.75 18.87 19.00 19.12 19.24 19.36 19.48 19.60 19.72
18. 19.85 19.97 20.09 20.21 20.33 20.46 20.58 20.70 20.83 20.95
19. 21.07 21.20 21.32 21.44 21.57 21.69 21.82 21.94 22.06 22.19
20. 22.31 22.44 22.56 22.69 22.82 22.94 23.07 23.19 23.32 23.45
21. 23.57 23.70 23.83 23.95 24.08 24.21 24.33 24.46 24.59 24.72
22. 24.85 24.97 25.10 25.23 25.36 25.49 25.62 25.75 25.88 26.01
23. 26.14 26.27 26.40 26.53 26.66 26.79 26.92 27.05 27.18 27.31
24. 27.44 27.58 27.71 27.84 27.97 28.10 28.24 28.37 28.50 28.63
25. 28.77 28.90 29.04 29.17 29.30 29.44 29.57 29.71 29.84 29.98
26. 30.11 30.25 30.38 30.52 30.65 30.79 30.92 31.06 31.20 31.33
(Continued)
811
TABLE T18.1 (Continued)
812

Hazard Values Corresponding to Probability Plotting Positions for Censored Data

Probability Tenth of Percent
Probability Percent 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

27. 31.47 31.61 31.75 31.88 32.02 32.16 32.30 32.43 32.57 32.71
28. 32.85 32.99 33.13 33.27 33.41 33.55 33.69 33.83 33.97 34.11
29 34.25 34.39 34.53 34.67 34.81 34.96 35.10 35.24 35.38 35.52
30. 35.67 35.81 35.95 36.10 36.24 36.38 36.53 36.67 36.82 36.96
31. 37.11 37.25 37.40 37.54 37.69 37.83 37.98 38.13 38.27 38.42
32. 38.57 38.71 38.86 39.01 39.16 39.30 39.45 39.60 39.75 39.90
33. 40.05 40.20 40.35 40.50 40.65 40.80 40.95 41.10 41.25 41.40
34. 41.55 41.70 41.86 42.01 42.16 42.31 42.46 42.62 42.77 42.92
35. 43.08 43.23 43.39 43.54 43.70 43.85 44.01 44.16 44.32 44.47
36. 44.63 44.79 44.94 45.10 45.26 45.41 45.57 45.73 45.89 46.04
37. 46.20 46.36 46.52 46.68 46.84 47.00 47.16 47.32 47.48 47.64
38. 47.80 47.96 48.13 48.29 48.45 48.61 48.78 48.94 49.10 49.27
39. 49.43 49.59 49.76 49.92 50.09 50.25 50.42 50.58 50.75 50.92
40. 51.08 51.25 51.42 51.58 51.75 51.92 52.09 52.26 52.42 52.59
41. 52.76 52.93 53.10 53.27 53.44 53.61 53.79 53.96 54.13 54.30
42. 54.47 54.65 54.82 54.99 55.16 55.34 55.51 55.69 55.86 56.04
43. 56.21 56.39 56.56 56.74 56.92 57.09 57.27 57.45 57.63 57.80
44. 57.98 58.16 58.34 58.52 58.70 58.88 59.06 59.24 59.42 59.60
45. 59.78 59.97 60.15 60.33 60.51 60.70 60.88 61.06 61.25 61.43
46. 61.62 61.80 61.99 62.18 62.36 62.55 62.74 62.92 63.11 63.30
47. 63.49 63.68 63.87 64.06 64.25 64.44 64.63 64.82 65.01 65.20
48. 65.39 65.59 65.78 65.97 66.16 66.36 66.55 66.75 66.94 67.14
49. 67.33 67.53 67.73 67.92 68.12 68.32 68.52 68.72 68.92 69.11
50. 69.31 69.51 69.72 69.92 70.12 70.32 70.52 70.72 70.93 71.13
51. 71.33 71.54 71.74 71.95 72.15 72.36 72.57 72.77 72.98 73.19
52. 73.40 73.61 73.81 74.02 74.23 74.44 74.65 74.87 75.08 75.29
(Continued)
Appendix
TABLE T18.1 (Continued)
Hazard Values Corresponding to Probability Plotting Positions for Censored Data
Probability Tenth of Percent
Appendix

Probability Percent 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

53. 75.50 75.72 75.93 76.14 76.36 76.57 76.79 77.00 77.22 77.44
54. 77.65 77.87 78.09 78.31 78.53 78.75 78.97 79.19 79.41 79.63
55. 79.85 80.07 80.30 80.52 80.74 80.97 81.19 81.42 81.64 81.87
56. 82.10 82.33 82.55 82.78 83.01 83.24 83.47 83.70 83.93 84.16
57. 84.40 84.63 84.86 85.10 85.33 85.57 85.80 86.04 86.27 86.51
58. 86.75 86.99 87.23 87.47 87.71 87.95 88.19 88.43 88.67 88.92
59. 89.16 89.40 89.65 89.89 90.14 90.39 90.63 90.88 91.13 91.38
60. 91.63 91.88 92.13 92.38 92.63 92.89 93.14 93.39 93.65 93.90
61. 94.16 94.42 94.67 94.93 95.19 95.45 95.71 95.97 96.23 96.50
62. 96.76 97.02 97.29 97.55 97.82 98.08 98.35 98.62 98.89 99.16
63. 99.43 99.70 99.97 100.24 100.51 100.79 101.06 101.34 101.61 101.89
64. 102.17 102.44 102.72 103.00 103.28 103.56 103.85 104.13 104.41 104.70
65. 104.98 105.27 105.56 105.84 106.13 106.42 106.71 107.00 107.29 107.59
66. 107.88 108.18 108.47 108.77 109.06 109.36 109.66 109.96 110.26 110.56
67. 110.87 111.17 111.47 111.78 112.09 112.39 112.70 113.01 113.32 113.63
68. 113.94 114.26 114.57 114.89 115.20 115.52 115.84 116.16 116.48 116.80
69. 117.12 117.44 117.77 118.09 118.42 118.74 119.07 119.40 119.73 120.06
70. 120.40 120.73 121.07 121.40 121.74 122.08 122.42 122.76 123.10 123.44
71. 123.79 124.13 124.48 124.83 125.18 125.53 125.88 126.23 126.58 126.94
72. 127.30 127.65 128.01 128.37 128.74 129.10 129.46 129.83 130.20 130.56
73. 130.93 131.30 131.68 132.05 132.43 132.80 133.18 133.56 133.94 134.32
74. 134.71 135.09 135.48 135.87 136.26 136.65 137.04 137.44 137.83 138.23
75. 138.63 139.03 139.43 139.84 140.24 140.65 141.06 141.47 141.88 142.30
76. 142.71 143.13 143.55 143.97 144.39 144.82 145.24 145.67 146.10 146.53
77. 146.97 147.40 147.84 148.28 148.72 149.17 149.61 150.06 150.51 150.96
78. 151.41 151.87 152.33 152.79 153.25 153.71 154.18 154.65 155.12 155.59
79. 156.06 156.54 157.02 157.50 157.99 158.47 158.96 159.45 159.95 160.45
(Continued)
813
 814

TABLE T18.1 (Continued)

Hazard Values Corresponding to Probability Plotting Positions for Censored Data
Probability Tenth of Percent
Probability Percent 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

80. 160.94 161.45 161.95 162.46 162.96 163.48 163.99 164.51 165.03 165.55
81. 166.07 166.60 167.13 167.66 168.20 168.74 169.28 169.83 170.37 170.93
82. 171.48 172.04 172.60 173.16 173.73 174.30 174.87 175.45 176.03 176.61
83. 177.20 177.79 178.38 178.98 179.58 180.18 180.79 181.40 182.02 182.64
84. 183.26 183.89 184.52 185.15 185.79 186.43 187.08 187.73 188.39 189.05
85. 189.71 190.38 191.05 191.73 192.41 193.10 193.79 194.49 195.19 195.90
86. 196.61 197.33 198.05 198.78 199.51 200.25 200.99 201.74 202.50 203.26
87. 204.02 204.79 205.57 206.36 207.15 207.94 208.75 209.56 210.37 211.20
88. 212.03 212.86 213.71 214.56 215.42 216.28 217.16 218.04 218.93 219.82
89. 220.73 221.64 222.56 223.49 224.43 225.38 226.34 227.30 228.28 229.26
90. 230.26 231.26 232.28 233.30 234.34 235.39 236.45 237.52 238.60 239.69
91. 240.79 241.91 243.04 244.18 245.34 246.51 247.69 248.89 250.10 251.33
92. 252.57 253.83 255.10 256.39 257.70 259.03 260.37 261.73 263.11 264.51
93. 265.93 267.36 268.82 270.31 271.81 273.34 274.89 276.46 278.06 279.69
94. 281.34 283.02 284.73 286.47 288.24 290.04 291.88 293.75 295.65 297.59
95. 299.57 301.59 303.66 305.76 307.91 310.11 312.36 314.66 317.01 319.42
96. 321.89 324.42 327.02 329.68 332.42 335.24 338.14 341.12 344.20 347.38
97. 350.66 354.05 357.56 361.19 364.97 368.89 372.97 377.23 381.67 386.32
98. 391.20 396.33 401.74 407.45 413.52 419.97 426.87 434.28 442.28 450.99
99. 460.52 471.05 482.83 496.18 511.60 529.83 552.15 580.91 621.46 690.77
Source: Sheesley, J.H., Report Number 1300–1119, General Electric Company, Cleveland, OH, 1974.
Appendix
Appendix 815

TABLE T18.2
H108 Table 2A–1—Life Test Sampling Plan Code Designation
α = 0.01 α = 0.05 α = 0.10 α = 0.25 α = 0.50
β = 0.10 β = 0.10 β = 0.10 β = 0.10 β = 0.10
Code θ1/θ0 Code θ1/θ0 Code θ1/θ0 Code θ1/θ0 Code θ1/θ0

A–1 0.004 B–1 0.022 C–1 0.046 D–1 0.125 E–1 0.301
A–2 .038 B–2 .091 C–2 .137 D–2 .247 E–2 .432
A–3 .082 B–3 .154 C–3 .207 D–3 .325 E–3 .502
A–4 .123 B–4 .205 C–4 .261 D–4 .379 E–4 .550
A–5 .160 B–5 .246 C–5 .304 D–5 .421 E–5 .584
A–6 .193 B–6 .282 C–6 .340 D–6 .455 E–6 .611
A–7 .221 B–7 .312 C–7 .370 D–7 .483 E–7 .633
A–8 .247 B–8 .338 C–8 .396 D–8 .506 E–8 .652
A–9 .270 B–9 .361 C–9 .418 D–9 .526 E–9 .667
A–10 .291 B–10 .382 C–10 .438 D–10 .544 E–10 .681
A–11 .371 B–11 .459 C–11 .512 D–11 .608 E–11 .729
A–12 .428 B–12 .512 C–12 .561 D–12 .650 E–12 .759
A–13 .470 B–13 .550 C–13 .597 D–13 .680 E–13 .781
A–14 .504 B–14 .581 C–14 .624 D–14 .703 E–14 .798
A–15 .554 B–15 .625 C–15 .666 D–15 .737 E–15 .821
A–16 .591 B–16 .658 C–16 .695 D–16 .761 E–16 .838
A–17 .653 B–17 .711 C–17 .743 D–17 .800 E–17 .865
A–18 .692 B–18 .745 C–18 .774 D–18 .824 E–18 .882

Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108),
Office of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.2.
Notes: Producer’s risk α is the probability of rejecting lots with mean life θ0.
Consumer’s risk β is the probability of accepting lots with mean life θ1.
816 Appendix

TABLE T18.3
H108 Table 2B–1—Master Table for Life Tests Terminated upon Occurrence of Preassigned
Number of Failures
Producer’s Risk (α)
0.01 0.05 0.10 0.25 0.50
r Code C/θ0 Code C/θ0 Code C/θ0 Code C/θ0 Code C/θ0

1 A–1 0.010 B–1 0.052 C–1 0.106 D–1 0.288 E–1 0.693
2 A–2 .074 B–2 .178 C–2 .266 D–2 .481 E–2 .839
3 A–3 .145 B–3 .272 C–3 .367 D–3 .576 E–3 .891
4 A–4 .206 B–4 .342 C–4 .436 D–4 .634 E–4 .918
5 A–5 .256 B–5 .394 C–5 .487 D–5 .674 E–5 .934
6 A–6 .298 B–6 .436 C–6 .525 D–6 .703 E–6 .945
7 A–7 .333 B–7 .469 C–7 .556 D–7 .726 E–7 .953
8 A–8 .363 B–8 .498 C–8 .582 D–8 .744 E–8 .959
9 A–9 .390 B–9 .522 C–9 .604 D–9 .760 E–9 .963
10 A–10 .413 B–10 .543 C–10 .622 D–10 .773 E–10 .967
15 A–11 .498 B–11 .616 C–11 .687 D–11 .816 E–11 .978
20 A–12 .554 B–12 .663 C–12 .726 D–12 .842 E–12 .983
25 A–13 .594 B–13 .695 C–13 .754 D–13 .859 E–13 .987
30 A–14 .625 B–14 .720 C–14 .774 D–14 .872 E–14 .989
40 A–15 .669 B–15 .755 C–15 .803 D–15 .889 E–15 .992
50 A–16 .701 B–16 .779 C–16 .824 D–16 .901 E–16 .993
75 A–17 .751 B–17 .818 C–17 .855 D–17 .920 E–17 .996
100 A–18 .782 B–18 .841 C–18 .874 D–18 .931 E–18 .997
Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108),
Office of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.28.
Notes: Producer’s risk α is the probability of rejecting lots with mean life θ0.
ˆ ³ q ( C/q ) .
Acceptance criterion: Accept lot if q r, n 0 0

For explanation of the code, see par. 2A3.2 and Table 2A-1.
Appendix 817

TABLE T18.4
H108 Table 2C–1 (b)—Master Table for Life Tests Terminated at Preassigned Time: Testing
without Replacement (Values of T/α0 for α = 0.05)
Sample Size
Code r 2r 3r 4r 5r 6r 7r 8r 9r 10r 20r

B–1 1 0.026 0.017 0.013 0.010 0.009 0.007 0.006 0.006 0.005 0.003
B–2 2 .104 .065 .048 .038 .031 .026 .023 .020 .018 .009
B–3 3 .168 .103 .075 .058 .048 .041 .036 .031 .028 .014
B–4 4 .217 .132 .095 .074 .061 .052 .045 .040 .036 .017
B–5 5 .254 .153 .110 .086 .071 .060 .052 .046 .041 .020
B–6 6 .284 .170 .122 .095 .078 .066 .057 .051 .045 .022
B–7 7 .309 .185 .132 .103 .084 .072 .062 .055 .049 .024
B–8 8 .330 .197 .141 .110 .090 .076 .066 .058 .052 .025
B–9 9 .348 .207 .148 .115 .094 .080 .069 .061 .055 .027
B–10 10 .363 .216 .154 .120 .098 .083 .072 .064 .057 .028
B–11 15 .417 .246 .175 .136 .112 .094 .082 .072 .065 .032
B–12 20 .451 .266 .189 .147 .120 .102 .088 .078 .070 .034
B–13 25 .475 .280 .199 .154 .126 .107 .093 .082 .073 .036
B–14 30 .493 .290 .206 .160 .131 .111 .096 .085 .076 .037
B–15 40 .519 .305 .216 .168 .137 .116 .101 .089 .079 .039
B–16 50 .536 .315 .223 .173 .142 .120 .104 .092 .082 .040
B–17 75 .564 .331 .235 .182 .149 .126 .109 .096 .086 .042
B–18 100 .581 .340 .242 .187 .153 .130 .112 .099 .089 .043

Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108), Office
of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.45.
Note: For explanation of the code, see par. 2A3.2 and Table 2A-1.
818 Appendix

TABLE T18.5
H108 Table 2C–2 (b)—Master Table for Life Tests Terminated at Preassigned Time: Testing
with Replacement (Values of T/α0 for α = 0.05)
Sample Size
Code r 2r 3r 4r 5r 6r 7r 8r 9r 10r 20r

B–1 1 0.026 0.017 0.013 0.010 0.009 0.007 0.006 0.006 0.005 0.003
B–2 2 .089 .059 .044 .036 .030 .025 .022 .020 .018 .009
B–3 3 .136 .091 .068 .055 .045 .039 .034 .030 .027 .014
B–4 4 .171 .114 .085 .068 .057 .049 .043 .038 .034 .017
B–5 5 .197 .131 .099 .079 .066 .056 .049 .044 .039 .020
B–6 6 .218 .145 .109 .087 .073 .062 .054 .048 .044 .022
B–7 7 .235 .156 .117 .094 .078 .067 .059 .052 .047 .023
B–8 8 .249 .166 .124 .100 .083 .071 .062 .055 .050 .025
B–9 9 .261 .174 .130 .104 .087 .075 .065 .058 .052 .026
B–10 10 .271 .181 .136 .109 .090 .078 .068 .060 .054 .027
B–11 15 .308 .205 .154 .123 .103 .088 .077 .068 .062 .031
B–12 20 .331 .221 .166 .133 .110 .095 .083 .074 .066 .033
B–13 25 .348 .232 .174 .139 .116 .099 .087 .077 .070 .035
B–14 30 .360 .240 .180 .144 .120 .103 .090 .080 .072 .036
B–15 40 .377 .252 .189 .151 .126 .108 .094 .084 .075 .038
B–16 50 .390 .260 .195 .156 .130 .111 .097 .087 .078 .039
B–17 75 .409 .273 .204 .164 .136 .117 .102 .091 .082 .041
B–18 100 .421 .280 .210 .168 .140 .120 .105 .093 .084 .042
Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108), Office
of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.47.
Note: For explanation of the code, see par. 2A3.2 and Table 2A-1.
Appendix 819

TABLE T18.6
H108 Table 2D–1 (b)—Master Table for Sequential Life Tests (α = 0.05)
Code r0 h0/θ0 h1/θ0 s/θ0 E0(r) Eo1(r) Es(r) Eo0(r)

B–1 3 0.0506 –0.0650 0.0859 0.8 0.8 0.4 0.0

B–2 6 .2254 –.2894 .2400 1.2 1.6 1.1 .3
B–3 9 .4098 –.5261 .3405 1.5 2.3 1.9 .6
B–4 12 .5805 –.7453 .4086 1.8 3.0 2.6 .9
B–5 15 .7345 –.9430 .4576 2.1 3.7 3.3 1.2
B–6 18 .8842 –1.1352 .4972 2.3 4.3 4.1 1.6
B–7 21 1.0209 –1.3107 .5282 2.5 5.0 4.8 1.9
B–8 24 1.1495 –1.4757 .5538 2.7 5.6 5.5 2.3
B–9 27 1.2719 –1.6329 .5756 2.8 6.3 6.3 2.7
B–10 30 1.3916 –1.7866 .5948 3.0 6.9 7.0 3.0
B–11 45 1.9101 –2.4523 .6607 3.7 10.0 10.7 5.0
B–12 60 2.3620 –3.0325 .7024 4.3 13.1 14.5 7.0
B–13 75 2.7516 –3.5327 .7307 4.8 16.1 18.2 9.1
B–14 90 3.1217 –4.0079 .7530 5.3 19.2 22.1 11.2
B–15 120 3.7522 –4.8173 .7833 6.2 25.0 29.5 15.3
B–16 150 4.3314 –5.5610 .8053 6.9 31.0 37.1 19.7
B–17 225 5.5386 –7.1109 .8391 8.5 45.6 55.9 30.5
B–18 300 6.5773 –8.4444 .8600 9.8 60.4 75.1 41.6

Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108),
Office of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.63.
Note: For explanation of the code, see par. 2A3.2 and Table 2A-1.
820 Appendix

TABLE T18.7
H108 Table 2C–5—Master Table for Proportion Failing before Specified Time: Life Test
Sampling Plans for Specified α, β, and p1/p0
Values of r (Upper Numbers) and of D (Lower Numbers)a
α = 0.01 α = 0.05 α = 0.10
p1/p0 β = 0.01 0.05 0.10 0.01 0.05 0.10 0.01 0.05 0.10

3/2 136 101 83 95 67 55 77 52 41

110.4 79.1 63.3 79.6 54.1 43.4 66.0 43.0 33.0
2 46 35 30 33 23 19 26 18 15
31.7 22.7 18.7 24.2 15.7 12.4 19.7 12.8 10.3
5/2 27 21 18 19 14 11 15 11 9
16.4 11.8 9.62 12.4 8.46 6.17 10.3 7.02 5.43
3 19 15 13 13 10 8 11 8 6
10.3 7.48 6.10 7.69 5.43 3.98 7.02 4.66 3.15
4 12 10 9 9 7 6 7 5 4
5.43 4.13 3.51 4.70 3.29 2.61 3.90 2.43 1.75
5 9 8 7 7 5 4 5 4 3
3.51 2.91 2.33 3.29 1.97 1.37 2.43 1.75 1.10
10 5 4 4 4 3 3 3 2 2
1.28 .823 .823 1.37 .818 .818 1.10 .532 .532

Source: United States Department of Defense, Quality Control and Reliability (Interim) Handbook (H-108),
Office of the Assistant Secretary of Defense (Supply and Logistics), Washington, DC, 1960, p. 2.55.
Life test sampling plans for specified α, β, and p1/p0.
Notes: Producer’s risk α is the probability of rejecting lots with acceptable proportion of lot failing before
specified time, p0.
Consumer’s risk β is the probability of accepting lots with unacceptable proportion of lot failing
before specified time, p1.
a The sample size n is obtained by taking the largest integer less than or equal to the tabled value divided by

p0, i.e., n = [D/p0].

Appendix 821

TABLE T18.8
TR3 Table 1—Table of Values for Percent Truncation, (t/μ) × 100
Shape Parameter = β
1 1 2 1 1
p′ (in %) 1 1 2 2 3 4 5
3 2 3 2 3
.010 .010 .45 1.13 2.83 7.03 11.03 17.26
.012 .012 .49 1.24 3.04 7.42 11.55 17.91
.015 .015 .57 1.38 3.32 7.94 12.21 18.72
.020 .020 .67 1.59 3.73 8.66 13.12 19.83
.025 .025 .77 1.78 4.08 9.26 13.87 20.74
.030 .030 .86 1.95 4.40 9.77 14.52 21.50
.040 .040 1.02 2.26 4.93 10.65 15.60 22.77
.050 .050 1.18 2.53 5.39 11.40 16.49 23.82
.065 .065 1.37 2.88 5.98 12.32 17.62 25.10
.080 .080 1.56 3.19 6.50 13.13 18.56 26.16
.100 .10 1.78 3.57 7.11 14.03 19.62 27.36
.12 .12 1.98 3.92 7.65 14.82 20.53 28.37
.15 .15 2.26 4.37 8.36 15.84 21.71 29.67
.20 .20 2.69 5.07 9.39 17.27 23.33 31.43
.25 .25 3.08 5.64 10.27 18.47 24.68 32.87
.30 .30 3.44 6.18 11.05 19.51 25.83 34.09
.40 .40 4.07 7.14 12.39 21.27 27.76 36.12
.50 .001 .50 4.67 7.99 13.55 22.75 29.36 37.76
.65 .002 .65 5.46 9.12 15.06 24.62 31.35 39.81
.80 .003 .80 6.19 10.11 16.36 26.21 33.03 41.50
1.00 .005 1.01 7.08 11.31 17.90 28.03 34.93 43.40
1.2 .007 1.21 7.90 12.40 19.26 29.62 36.57 45.02
1.5 .011 1.51 9.07 13.87 21.08 31.68 38.68 47.09
2.0 .020 2.02 10.77 16.03 23.67 34.56 41.59 49.90
2.5 .032 2.53 12.33 17.95 25.90 36.98 44.01 52.21
3.0 .047 3.05 13.78 19.69 27.89 39.09 46.09 54.17
4.0 .001 .083 4.08 16.42 22.79 31.35 42.69 49.59 57.45
5.0 .002 .13 5.13 18.84 25.58 34.35 45.71 52.50 60.13
6.5 .005 .23 6.72 22.15 29.25 38.28 49.57 56.18 63.46
8.0 .010 .35 8.34 25.20 32.59 41.72 52.88 59.29 66.26
10.0 .020 .56 10.54 29.01 36.63 45.82 56.73 62.85 69.44
12 .034 .82 12.78 32.58 40.34 49.50 60.11 65.96 72.18
15 .070 1.32 16.25 37.63 45.48 54.49 64.60 70.05 75.73
20 .18 2.49 22.31 45.51 53.30 61.85 71.04 75.83 80.68
25 .40 4.14 28.77 52.99 60.53 68.47 76.67 80.80 84.89
30 .76 6.36 35.37 60.29 67.39 74.62 81.79 85.26 88.62
40 2.22 13.04 51.08 74.79 80.64 86.15 91.09 93.27 95.22
50 5.55 24.02 69.31 89.82 93.95 97.33 99.82 100.67 101.21
65 19.28 55.10 104.98 115.23 115.61 114.92 113.06 111.68 109.98
80 69.48 129.52 160.94 148.91 143.14 136.34 128.53 124.27 119.79
Source: United States Department of Defense, Quality Control and Reliability Technical Report (TR3), Office of
the Assistant Secretary of Defense (Installations and Logistics), U.S. Government Printing Office,
Washington, DC, 1961, p. 26.
 822

TABLE T18.9
TR7 Table 1A—100t/μ Ratios at the Acceptable Quality Level (Normal Inspection) for the MIL–STD–105E Plans
Shape Parameter, β

Acceptable Quality 1 1 2 1 2 2 1 1 4
1 1 1 2 3
Level p′ (%) 3 2 3 3 3 2 3
0.010 17–12 50–8 75–6 .010 .11 .45 1.13 2.83 7.03 11.0
0.015 56–12 11–7 14–5 .015 .15 .57 1.38 3.32 7.94 12.2
0.025 26–11 31–7 30–5 .025 .22 .77 1.78 4.08 9.26 13.9
0.040 11–10 80–7 60–5 .040 .31 1.02 2.26 4.93 10.7 15.6
0.065 46–10 21–6 13–4 .065 .44 1.37 2.88 5.98 12.3 17.6
0.10 17–9 50–6 25–4 .10 .61 1.78 3.57 7.11 14.0 19.6
0.15 56–9 11–5 44–4 .15 .83 2.26 4.37 8.36 15.8 21.7
0.25 26–8 31–5 94–4 .25 1.22 3.08 5.64 10.3 18.5 24.7
0.40 11–7 80–5 .019 .40 1.73 4.07 7.14 12.4 21.3 27.8
0.65 46–7 21–4 .040 .65 2.50 5.46 9.12 15.1 24.6 31.4
1.0 17–6 51–4 .076 1.01 3.45 7.08 11.3 17.9 28.0 34.9
1.5 59–6 .011 .14 1.51 4.69 9.07 13.9 21.1 31.7 38.7
2.5 27–5 .032 .30 2.53 6.91 12.3 18.0 25.9 37.0 44.0
4.0 11–4 .083 .62 4.08 9.88 16.4 22.8 31.4 42.7 49.6
6.5 51–4 .23 1.31 6.72 14.4 22.2 29.3 38.3 49.6 56.2
10 .019 .56 2.57 10.5 20.1 29.0 36.6 45.8 56.7 62.9

Source: United States Department of Defense, Quality Control and Reliability Technical Report (TR7), Office of the Assistant Secretary of Defense (Installations and
Logistics), U.S. Government Printing Office, Washington, DC, 1965, p. 14.
Note: The negative figure after a ratio shows the number of decimal points to provide. Thus, 13–4 = .0013.
Appendix
Appendix 823

TABLE T18.10
TR7 Table 1B—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.10
Shape Parameter, β

Code 1 1 2 1 2 1 1
1 1 1 2 2 3 4
Letter AQL 3 2 3 3 3 2 3
A 6.5 25 68 92 120 120 118 118 116 115 115
B 4.0 7.2 29 50 77 89 95 98 100 102 103
C 2.5 1.6 10 23 46 61 70 77 82 88 91
C 10 11 40 62 89 100 102 103 105 106 106
D 1.5 .38 4.1 11.6 28 43 53 60 68 76 80
D 6.5 2.4 13 28 53 67 76 81 86 91 94
D 10 7.2 29 50 77 89 95 98 100 102 103
E 1.0 .094 1.5 5.6 17 30 39 47 56 66 71
E 4.0 .49 4.8 13 31 45 55 63 70 78 82
E 6.5 1.5 10 22 45 59 68 76 80 86 90
E 10 3.5 17 37 60 73 82 87 90 95 97
F 0.65 .026 .66 2.9 11 22 30 38 47 58 64
F 2.5 .14 2.0 6.7 20 33 42 50 58 68 72
F 4.0 .36 4.0 11 28 42 52 59 67 75 80
F 6.5 .80 6.5 16 36 51 61 68 73 81 85
F 10 2.6 14 29 54 68 77 82 87 92 95
G 0.40 62–4 .26 1.4 7.2 15 23 30 39 50 57
G 1.5 .032 .76 3.2 12 22 31 39 48 59 65
G 2.5 .086 1.4 5.3 17 29 38 47 55 65 70
G 4.0 .18 2.4 7.7 22 35 45 53 60 70 74
G 6.5 .52 5.0 13 31 46 56 63 70 78 82
G 10 1.2 8.8 20 42 57 66 73 78 85 89
H 0.25 16–4 .11 .74 4.6 11 17 24 33 44 51
H 1.0 84–4 .31 1.6 7.8 16 24 31 40 51 58
H 1.5 .021 .59 2.6 11 20 29 37 46 57 63
H 2.5 .046 .97 3.9 14 25 34 42 51 61 67
H 4.0 .12 1.8 6.5 19 32 42 49 58 67 72
H 6.5 .27 3.2 9.7 25 39 49 57 65 73 78
H 10 .68 6.0 15 34 49 58 67 73 80 85
J 0.15 40–5 .042 .37 2.9 7.5 13 19 27 38 45
J 0.65 20–4 .12 .80 4.9 11 18 24 33 45 52
J 1.0 54–4 .23 1.3 6.7 14 22 29 38 49 57
J 1.5 .010 .36 1.8 8.3 17 25 32 42 53 59
J 2.5 .030 .72 3.1 12 22 31 39 48 58 64
J 4.0 .063 1.2 4.5 15 27 36 44 53 63 68
J 6.5 .16 2.3 7.5 21 34 44 52 60 69 74
J 10 .34 3.8 11 27 41 51 59 67 75 80
K 0.10 10–5 .017 .19 1.8 5.5 10 15 23 33 40
K .40 50–5 .049 .41 3.1 8.0 14 20 28 39 46
K .65 13–4 .093 .67 4.3 10 17 23 32 43 50
(Continued)
824 Appendix

TABLE T18.10 (Continued)

TR7 Table 1B—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.10
Shape Parameter, β

Code 1 1 2 1 2 1 1
1 1 1 2 2 3 4
Letter AQL 3 2 3 3 3 2 3
K 1.0 27–4 .15 .94 5.4 12 19 26 35 46 53
K 1.5 76–4 .29 1.5 7.6 15 23 31 40 51 58
K 2.5 .015 .47 2.2 9.8 19 27 35 44 55 61
K 4.0 .039 .85 3.5 13 23 33 41 50 60 66
K 6.5 .092 1.5 5.5 17 29 39 47 56 65 70
K 10 .27 3.2 9.7 25 39 49 57 65 73 78
L 0.065 25–6 67–4 .093 1.1 3.8 7.7 12 19 29 36
L 0.25 12–5 .019 .20 1.9 5.7 10 15 23 34 41
L 0.40 33–5 .036 .33 2.7 7.2 12 18 26 37 45
L 0.65 66–5 .058 .47 3.4 8.5 14 20 29 40 47
L 1.0 18–4 .11 .79 4.8 11 18 24 33 44 52
L 1.5 40–4 .18 1.1 6.0 13 20 27 36 48 55
L 2.5 91–4 .32 1.7 8.0 16 24 32 40 52 59
L 4.0 .020 .56 2.6 10 20 29 36 45 56 62
L 6.5 .060 1.1 4.4 15 26 36 44 52 62 68
M 0.04 60–7 26–4 .047 .73 2.7 5.8 9.6 16 25 32
M 0.15 30–6 78–4 .10 1.2 4.0 8.0 12 19 30 37
M 0.25 80–6 .015 .17 1.7 5.1 9.7 14 22 33 40
M 0.40 16–5 .023 .23 2.1 6.0 11 16 24 35 42
M 0.65 45–5 .045 .39 3.0 7.8 13 19 27 38 46
M 1.0 95–5 .074 .56 3.8 9.3 15 22 30 42 49
M 1.5 22–4 .12 .85 5.0 11 18 25 34 45 52
M 2.5 51–4 .22 1.3 6.6 14 22 29 38 49 56
M 4.0 .013 .45 2.1 9.4 18 27 34 43 54 61
N 0.025 14–7 10–4 .024 .46 1.9 4.4 7.6 13 22 28
N 0.10 72–7 31–4 .052 .79 2.8 6.1 10 16 26 32
N 0.15 19–6 56–4 .082 1.0 3.5 7.3 11 18 28 35
N 0.25 40–6 92–4 .11 1.3 4.3 8.4 13 20 30 37
N 0.40 11–5 .017 .19 1.8 5.5 10 15 23 33 40
N 0.65 22–5 .028 .27 2.4 6.6 12 17 25 36 43
N 1.0 50–5 .049 .41 3.1 8.0 14 20 28 39 46
N 1.5 12–4 .083 .62 4.0 9.8 16 22 31 42 49
N 2.5 35–4 .17 1.0 5.9 13 20 27 36 47 54
P 0.015 35–8 40–5 .012 .29 1.3 3.3 6.0 11 19 25
P 0.065 17–7 12–4 .026 .49 2.0 4.6 7.8 13 22 29
P 0.10 44–7 22–4 .041 .67 2.5 5.5 9.2 15 25 31
P 0.15 92–7 34–4 .057 .84 3.0 6.3 10 17 26 33
P 0.25 25–6 68–4 .094 1.1 3.8 7.7 12 19 29 36
P 0.40 51–6 .011 .13 1.4 4.6 8.9 13 20 31 38
P 0.65 12–5 .019 .20 1.9 5.7 10 15 23 34 41
P 1.0 28–5 .033 .30 2.5 6.9 12 18 26 37 44
(Continued)
Appendix 825

TABLE T18.10 (Continued)

TR7 Table 1B—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.10
Shape Parameter, β

Code 1 1 2 1 2 1 1
1 1 1 2 2 3 4
Letter AQL 3 2 3 3 3 2 3
P 1.5 77–5 .063 .50 3.5 8.8 15 21 29 41 48
Q 0.010 90–9 16–5 63–4 .18 .96 2.5 4.9 9.0 16 23
Q 0.040 44–8 48–5 .013 .31 1.4 3.5 6.2 11 19 26
Q 0.065 11–7 90–5 .021 .43 1.8 4.2 7.4 12 21 28
Q 0.10 22–7 14–4 .029 .53 2.1 4.8 8.2 14 23 30
Q 0.15 62–7 28–4 .048 .75 2.7 5.9 9.7 16 25 32
Q 0.25 13–6 45–4 .069 .95 3.3 6.8 11 17 27 34
Q 0.40 30–6 78–4 .10 1.2 4.1 8.0 12 19 30 37
Q 0.65 70–6 .013 .15 1.6 4.9 9.4 14 22 32 39
Q 1.0 19–5 .026 .26 2.3 6.4 11 17 25 35 43
R 0.025 10–8 18–5 68–4 .19 1.0 2.6 5.0 9.3 17 23
R 0.040 26–8 35–5 .010 .26 1.2 3.2 5.8 10 18 25
R 0.065 54–8 55–5 .015 .33 1.5 3.6 6.5 11 20 26
R 0.10 15–7 11–4 .024 .47 1.9 4.5 7.7 13 22 29
R 0.15 30–7 17–4 .034 .59 2.3 5.1 8.7 14 24 30
R 0.25 70–7 30–4 .051 .78 2.8 6.0 10 16 26 33
R 0.40 17–6 52–4 .075 1.0 3.4 7.1 11 18 28 35
R 0.65 46–6 .010 .12 1.4 4.5 8.7 13 20 31 38
Source: United States Department of Defense, Quality Control and Reliability Technical Report (TR7), Office of
the Assistant Secretary of Defense (Installations and Logistics), U.S. Government Printing Office,
Washington, DC, 1965, pp. 15–17.
Note: A negative figure after a ratio shows the number of decimal points to provide. Thus, 62–4 = .0062.
826 Appendix

TABLE T18.11
TR7 Table 1C—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.05
Shape Parameter, β

1 1 2 1 2 1 1
1 1 1 2 2 3 4
Code Letter AQL 3 2 3 3 3 2 3

A 6.5 55 120 130 140 140 130 130 120 120 120
B 4.0 16 50 73 100 110 110 110 110 110 110
C 2.5 3.5 18 35 60 74 82 87 90 96 97
C 10 20 59 84 110 120 110 110 110 110 110
D 1.5 .84 6.9 17 36 52 61 69 76 82 86
D 6.5 4.3 20 37 64 77 85 90 93 97 99
D 10 13 43 65 93 100 100 100 110 100 100
E 1.0 .22 2.8 8.6 23 37 47 55 63 72 76
E 4.0 .95 7.4 18 39 53 63 70 76 83 87
E 6.5 2.5 14 28 53 67 76 82 86 92 95
E 10 5.5 24 43 69 82 89 94 97 99 100
F 0.65 .059 1.1 4.4 15 26 35 43 52 62 68
F 2.5 .25 3.1 9.3 25 38 48 56 64 73 77
F 4.0 .60 5.4 14 33 48 57 65 72 79 83
F 6.5 1.2 8.6 20 42 57 66 73 78 85 89
F 10 3.8 19 36 62 75 83 88 92 96 98
G 0.40 .013 .43 2.1 9.3 18 27 34 43 54 61
G 1.5 .059 1.1 4.4 15 26 35 43 52 62 68
G 2.5 .13 1.9 6.7 20 32 42 50 58 67 72
G 4.0 .29 3.4 10 26 30 50 57 65 74 78
G 6.5 .76 6.3 16 35 50 60 67 74 81 85
G 10 1.6 10 23 47 61 70 77 82 88 91
H 0.25 37–4 .18 1.1 5.9 13 20 27 36 47 54
H 1.0 .014 .46 2.2 9.6 18 27 34 42 55 61
H 1.5 .034 .82 3.4 12 23 32 40 49 60 66
H 2.5 .070 1.3 4.9 16 27 37 45 54 64 70
H 4.0 .18 2.5 7.9 22 35 45 53 61 71 75
H 6.5 .40 4.1 11 28 43 53 60 68 76 80
H 10 .93 7.4 18 39 54 63 70 76 83 87
J 0.15 90–5 .072 .55 3.7 9.3 15 22 30 41 49
J 0.65 37–4 .18 1.1 5.9 13 20 27 36 47 54
J 1.0 92–4 .32 1.7 8.0 16 24 32 40 52 58
J 1.5 .016 .48 2.3 9.9 19 28 35 44 57 61
J 2.5 .046 .95 3.9 14 25 34 42 51 61 67
J 4.0 .089 1.5 5.5 17 29 39 47 55 65 70
J 6.5 .18 2.5 7.9 22 35 45 53 61 71 75
J 10 .45 4.6 12 30 44 54 62 69 77 81
K 0.10 24–5 .029 .28 2.4 6.6 12 17 25 36 43
K 0.40 10–4 .076 .58 3.8 9.4 16 22 30 42 49
K 0.65 23–4 .13 .87 5.1 11 18 25 34 45 52
(Continued)
Appendix 827

TABLE T18.11 (Continued)

TR7 Table 1C—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.05
Shape Parameter, β

1 1 2 1 2 1 1
1 1 1 2 2 3 4
Code Letter AQL 3 2 3 3 3 2 3

K 1.0 46–4 .20 1.2 6.4 13 21 28 37 49 56

K 1.5 .011 .38 1.9 8.7 17 26 33 42 53 60
K 2.5 .030 .72 3.1 12 22 31 39 48 58 64
K 4.0 .059 1.1 4.4 15 26 35 43 52 62 68
K 6.5 .12 1.8 6.5 19 32 42 49 58 67 72
K 10 .34 3.8 11 27 41 51 59 67 75 80
L 0.065 56–6 .011 .14 1.5 4.7 9.0 13 21 31 38
L 0.25 24–5 .029 .28 2.4 6.6 12 17 25 36 43
L 0.40 58–5 .053 .44 3.2 8.3 14 20 28 39 47
L 0.65 11–4 .082 .60 4.0 9.7 16 22 31 42 49
L 1.0 28–4 .15 .95 5.5 12 19 26 35 46 53
L 1.5 56–4 .24 1.3 6.8 14 22 29 38 49 57
L 2.5 .012 .40 2.0 8.9 17 26 33 42 53 60
L 4.0 .027 .67 3.0 11 22 30 38 47 58 64
L 6.5 .070 1.3 4.9 16 27 37 45 54 64 70
M 0.040 13–6 46–4 .070 .96 3.3 6.8 11 17 27 34
M 0.15 56–6 .011 .14 1.5 4.7 9.0 13 21 31 38
M 0.25 13–5 .020 .21 2.0 5.8 10 16 23 34 41
M 0.40 27–5 .032 .30 2.5 6.9 12 18 25 37 44
M 0.65 64–5 .057 .46 3.3 8.5 14 20 29 40 47
M 1.0 13–4 .093 .67 4.3 10 17 23 32 43 50
M 1.5 30–4 .15 .99 5.6 12 19 26 35 46 53
M 2.5 68–4 .27 1.4 7.3 15 23 30 39 50 57
M 4.0 .017 .51 2.4 10 19 28 35 45 56 62
N 0.025 33–7 18–4 .035 .60 2.3 5.2 8.7 14 24 31
N 0.10 13–6 46–4 .070 .96 3.3 6.8 11 17 27 34
N 0.15 40–6 92–4 .11 1.3 4.3 8.4 13 20 30 37
N 0.25 68–6 .013 .15 1.6 4.9 9.3 14 22 32 39
N 0.40 16–5 .022 .23 2.1 6.0 11 16 24 35 42
N 0.65 30–5 .035 .32 2.6 7.0 12 18 26 37 45
N 1.0 70–5 .061 .48 3.4 8.7 14 21 29 40 48
N 1.5 16–4 .10 .70 4.5 10 17 23 32 44 51
N 2.5 44–4 .20 1.2 6.3 13 21 28 37 48 55
P 0.015 80–8 72–5 .018 .38 1.6 3.9 6.9 12 21 27
P 0.065 30–7 17–4 .034 .59 2.3 5.1 8.7 14 24 30
P 0.10 67–7 31–4 .053 .80 2.8 6.1 10 16 26 32
P 0.15 14–6 47–4 .072 .98 3.3 7.0 11 17 27 34
P 0.25 40–6 92–4 .11 1.3 4.3 8.4 13 20 30 37
P 0.40 68–6 .013 .15 1.6 4.9 9.3 14 22 32 39
P 0.65 16–5 .022 .23 2.1 6.0 11 16 24 35 42
(Continued)
828 Appendix

TABLE T18.11 (Continued)

TR7 Table 1C—100t/μ Ratios at the Limiting Quality Level for the MIL–STD–105E Plans:
Consumer’s Risk = 0.05

Shape Parameter, β

1 1 2 1 2 1 1
1 1 1 2 2 3 4
Code Letter AQL 3 2 3 3 3 2 3

P 1.0 33–5 .036 .33 2.7 7.2 12 18 26 37 45

P 1.5 10–4 .076 .58 3.8 9.4 16 22 30 42 49
Q 0.010 19–8 28–5 92–4 .024 1.1 3.0 5.5 10 18 24
Q 0.040 80–8 72–5 .018 .38 1.6 3.9 6.9 12 21 27
Q 0.065 18–7 12–4 .026 .50 2.0 4.6 8.0 13 22 29
Q 0.10 35–7 19–4 .036 .63 2.4 5.3 8.9 15 24 31
Q 0.15 92–7 34–4 .057 .84 3.0 6.3 10 17 26 33
Q 0.25 21–6 62–4 .087 1.1 3.7 7.5 12 18 29 35
Q 0.40 46–6 .010 .12 1.4 4.5 8.7 13 20 31 38
Q 0.65 10–5 .017 .18 1.8 5.3 10 15 22 33 40
Q 1.0 24–5 .029 .28 2.4 6.6 12 17 25 36 43
R 0.025 19–8 28–5 92–4 .024 1.1 3.0 5.5 10 18 24
R 0.040 44–8 50–5 .014 .32 1.4 3.6 6.4 11 20 26
R 0.065 88–8 76–5 .018 .39 1.7 4.0 7.1 12 21 27
R 0.10 22–7 14–4 .029 .53 2.1 4.8 8.2 14 23 30
R 0.15 44–7 22–4 .041 .67 2.5 5.5 9.2 15 25 31
R 0.25 10–6 36–4 .059 .85 3.0 6.4 10 17 26 33
R .040 21–6 62–4 .087 1.1 3.7 7.5 12 18 29 35
R 0.65 56–6 .011 .14 1.5 4.7 9.0 13 21 31 38
Source: United States Department of Defense, Quality Control and Reliability Technical Report (TR7),
Office of the Assistant Secretary of Defense (Installations and Logistics), U.S. Government
Printing Office, Washington, DC, 1965, pp. 18–20.
Note: A negative figure after a ratio shows the number of decimal points to provide. Thus, 92–4 = .0092.
Index

A AOQL, see Average outgoing quality limit

APL, see Acceptable process level
Acceptable process level (APL), 173–178
AQL, see Acceptable quality level
Acceptable quality level (AQL), 8, 25–26, 125,
Army Service Forces (ASF) tables, 8
431, 445, 586–590
ASN, see Average sample number
Acceptance control chart, 175–179
ASTM International Standard E1994, 398
Acceptance control limit (ACL), 176–177
ASTM International Standard E2234, 297
correction terms, 678
ASTM International Standard E2819, 439
factors, 676–677
ATI, see Average total inspection
Acceptance quality control
Average outgoing quality (AOQ), 27,
Army Ordnance sampling tables, 1
92–93, 371
background of, 7–8
double-sampling plan, 141–142, 155
methods, 4
ISO 3534-2 standard definition of, 104
probability, 14
multiple-sampling plan, 144–145
and process quality control
for plan N = 20, n = 10, c = 1, 94–95, 105–106
process optimization study, 6
single-sampling attributes plans, 122–123,
Shewhart control chart, 4–6
125–126
rectification scheme, 3
Average outgoing quality limit (AOQL), 7–8,
sampling plan, 1–3
371, 409, 421, 446, 586, 591
Acceptance sampling
Dodge CSP-1 plan, 411
basic principle of administration, 596
double-sampling plan, 141–143
binomial distribution, 43–46
ISO 3534-2 standard definition of, 104
choosing quality levels, 101–102
multiple sampling plan, 144, 146
economic considerations, 593–595
rectification plan, 93–94
hypergeometric distribution (see
setting of, 590
Hypergeometric distribution)
single-sampling attributes plans, 121–123
life cycle, 582
SkSP-2 plans, values of y for, 777–778
mandatory standards, 595
upper limit on, 413
plan, 277
values of x and y for, 776
quality levels, 586–593
Average run length (ARL), 86–87, 100
reasons for, 9
Average sample number (ASN), 353, 524
scheme, 277
curtailed single-sample plans, 95–96
selection and implementation, 583–586
double-sampling plan, 133, 141–142, 155
system, 277
ISO 3534-2 standard definition of, 104
Accept on zero (AoZ) plan, 27, 515–516
multiple-sampling plan, 133, 144, 149–151
Advisory Group on Reliability of Electronic
for plan n = 10, c = 1, 105–106
Equipment (AGREE), 551
selected values, measures of, 164
Altman AOQL diagram, 372
for single, double, and multiple matched
American Society for Testing and Materials
plans, 139–140
(ASTM) standard, 555
single-sampling attributes plans, 122
American Statistical Association, 8
Average total inspection (ATI), 371, 519
Anscombe rectifying inspection procedure
double-sampling plan, 141, 143, 156
measures, 394–395
ISO 3534-2 standard definition of, 104
operation, 391
multiple-sampling plan, 144, 146
selection, 392–394
for plan N = 20, n = 10, c = 1, 93–94, 105, 107
ANSI/ASQ Standard Z1.4, 297
single-sampling attributes plans, 122–123
AOQ, see Average outgoing quality

829
830 Index

B process parameter, 207

sampling and compositing,
Barnard sequential chart, 187–189
method of, 207
Barnard sequential t-test, 189–190, 679
steps of, 207–208
Bendix Aviation Corporation, 339
sampling techniques, 208
Binomial distribution, 61
statistical analysis, 208
approximations, 66, 68–69
stream of lots, 212–216
Excel, 72
unique lots, 209–211
frequency function, 43–44
variance, components of, 213
Harvard table, 629–636
Larson nomograph, 45–46
mean of, 43–44 C
Minitab, 75–76 Cameron table, unity values, 125
process proportion defective, 44 probability of acceptance, 651–652
recursion formula, 44 single-sampling plans, 649–650
standard deviation, 44 Censored data, hazard values, 811–814
Type B probability of acceptance Censored sampling, 552–554
c = 1, 88 Central limit theorem, 59–60
c/n = 1, 89 Chain sampling plans
f-Binomial distribution, 61 ChSP-1
conventional binomial distribution, 49 AQL and AOQL, 782–784
example, 50 AQL and LTPD, 781
Excel, 73 binomial distribution, 456
frequency function, 49 CRC, 456
mean and standard deviation, 49 Dodge chain sampling inspection
recursion formula, 49 procedure, 453
f-Binomial nomograph, 116–118 OC curve, 454–455
Binomial tables, 63, 65 Poisson unity values, 454, 780
British Standards Institution Standard two-stage plans, 457–458
Number, 7 Classical probability, 13–14
Bulk sampling Compliance sampling
coal, 195 LSP
construction, 196–198 applications, examples of, 482–483
developing a standard sampling method, 196 hypergeometric distribution, 484
estimation Poisson distribution, 484
composite entire sample, 207 procedure, 478–480
conservative approximation, 205 producer’s risk, 481–482
percent potassium bitartrate in, 203 protection, 479, 481
possible combinations of, 199 zero defectives, 478
precision of, 198 MIL-STD-1916
sample mean, variance of, 199 ASTM, 499
segment and increment variances, 213 implementation, 496–498
standard error of, 201 measures, 498
with unique lots, 205 operation, 494–496
of variability, 205 structure, 494
variance table, analysis of, 202 nomograph, zero defectives, 515
ISO 10725, 216–218 QSS
liquid, purity of, 208 QSS-2(n; kn, c0), 491–492
multistage sampling, 208 QSS-3(n; kn, c0), 492
objectives, 195 QSS-4(n; kn, c0), 492
plans, 207–209 QSS-1(n, kn; 0) plan, 492–493
procedure estimates, 212 QSS-2(n, kn; 0) plan, 493
sampling plans QSS-3(n, kn; 0) plan, 493
Index 831

QSS-3(n; cN, cT) plan, 491 block continuous plans

QSS-4(n, kn; 0) plan, 493–494 Girshick plan, 430–431
QSS-4(n; cN, cT) plan, 491 Wald–Wolfowitz plan, 428–430
single-sampling plan, 490 Dodge continuous plans (see Dodge
tightened inspection, 489 continuous plans)
unity values, 489 MIL-STD-1235B
simplified grand lot procedure AFI, 437–438
acceptance sampling, 500 AOQL protection, 433
approximations, 500 AQL, 431, 437
attributes, 502–504 check sequence, 433–434
continuing series of lots, 510–513 Markov chain approach, 437
control chart, 499 MIL-STD-105E plans, 431–432
maverick limits, 513 OC curves, 437–438
means procedures, 513 procedure, CSP-T plans, 435
MIL-STD-105E, 499 sampling frequency code letters, 435
mixed variables–attributes structure of, 433
procedures, 514 type and purpose of, 432
Poisson distribution, 514 values of i, CSP-T plans, 436
Simon’s approach, 500–501 values of S, CSP-T plans, 436
variables, 504–510 MLP
software applications advantages of, 427
Excel, 525–542 AFI, 426
Minitab, 543–545 AOQL, 423, 425
statgraphics, 545–547 infinite-level plan, 423–424
TNT scheme Markov chain approach, 423
MIL-STD-105E, 484 sampling frequency, 424
procedure, 485 software applications
protection, 485–486 Excel, 439–442
selection, 487–489 Snap Sampling Plans!, 442–443
zero plans tightened multilevel plans, 427–428
AoZ and AQL plans, 517, 519 Control chart, by Shewhart, see Shewhart
chain sampling alternative, 519–521 control chart
ChQSS, 521–524 Control chart factors, 616
Squeglia plans, 516–518 Control table
U.S. Department of Defense approach, acceptance and rejection, probability of, 148
515–516 format, 146–147
Compressed limit plans, 344 multiple plan, evaluation of,
Confidence intervals, 97–99 148–151
Consumer’s quality level (CQL), 26, 113, 120, for plan having acceptance numbers,
135, 137, 139–140, 586 146–148
OC curve, 99–101 Correction terms (CTs), 177–179
setting of, 593 Cost–benefit analysis, 595
Consumer’s risk (CR) CQL, see Consumer’s quality level
CQL, 99–101 Credit-based schemes, 395–397
single-sampling attributes plans, 113 Cumulative results criterion (CRC), 27, 399, 456,
hypergeometric tables, 120 462, 467–468
Poisson unity values, tables of, 115 Cumulative results plans
Continuous distributions chain sampling plans
and exponential distribution, 53–54 ChSP-1, 453–457
sampling plans, 53 two-stage plans, 457–458
standard deviation of, 40 CRC plan, 466–468
Continuous sampling plans deferred sentencing schemes, 458–461
ASTM International Standard E2819, 439 Excel, 468–474
832 Index

skip-lot sampling plans Dodge–Perry table, 451

demerit rating plan, 462–466 Dodge–Romig curves
SkSP-1, 445–447 acceptance number, 387, 390
SkSP-2, 447–452 minimum ATI, 389
Cumulative sum (CUSUM) chart, 190–193 sample size, 388
Curtailment, single-sample plans, 94–96 Dodge–Romig sampling schemes
AOQL single-and double-sampling plans, 373
binomial operating characteristic curves, 374
D
hypergeometric probabilities, LTPD, 385
Defects LTPD single-and double-sampling plans, 373
classification, 102, 463 measures, 381
vs. defective, 83 minimum ATI
definition of, 103 AOQL plan, 388–391
specifications and, 103 LTPD plan, 386–388
state-of-the-art process average, 588 operation, 374
Type A and B OC curves, 84–85 Poisson approximation, AOQL, 385
Defense Department Quality Control and selection
Reliability Technical Report, 565–575 AOQL double sampling, 376, 378
Dependent mixed plan AOQL single sampling, 375, 377
AOQ, 365 check sequence, 375–376
ASN, 363–364 lot size and process average, 374
ATI, 364 LTPD double-sampling, 380–381
formulas for measures of, 360 LTPD single-sampling, 376, 379
inspection procedure, 362 tables, structure of, 375
mixed plans, 361 variables plans, 381
probability of acceptance, 363 Dodge–Romig tables
upper specification limit, 361 known and unknown standard deviation,
Distribution function, 35 381–383
for die, 36–37 modified templates, 399
exponential distribution, 54 Double-sampling approach, 461
location/central tendency, measures of, 39 Double-sampling attributes plans
sample location and spread, measures of, 41 ASN curve, 139–140
spread of distribution, measures of binomial/Poisson probability distributions,
range, 40 141
standard deviation, 39–40 efficiency, 133
tabular form, 36 measures of, 141–143
for watch stoppage, 37–38 OC curve, 133, 139
Weibull distribution, 55 operating ratios, 137–138
Dodge continuous plans probability of acceptance, 137
CSP-1, 2, and 3 measures procedure for, 134
formulas, 415–420 software applications
sampling inspection, 415 Excel template, 151–156
Dodge CSP-1 plan, 409–412 Statgraphics, 154
Dodge–Torrey CSP-2 and CSP-3, unity values, 135–136
412–415 Double specification limits, 184–185, 187–188
stopping rules and selection of
base logarithms, 423
E
corrective action, 420
defective units, 420 Electronic Industries Association, 587
Dodge charts, 421 Empirical probability, 13–14
graph, 422 Excel, 439–442
Dodge formula, for scheme performance of alternative sampling plan, 401
MIL-STD-105E, 299 AOQL sampling plan derivation, 402
Index 833

BETAINV function, 398 for hypergeometric distribution, 41

BINOM.DIST function, 399 negative binomial distribution, 51
binomial distribution, 72 normal distribution, 57
f-binomial distribution, 73 Poisson distribution, 45
binomial probability distribution, 398 tabular form, 36
compliance sampling for watch stoppage, 37–38
LSP plans, 525 Weibull distribution, 55–56
MIL-STD-1916 plans, 537–542
QSS plans, 527, 529–536
G
TNT plans, 525–527
cumulative results plans Gasoline, 195
AOQ curve, chain sampling plan, 474 General rule
chain sampling plan, 472 of addition, 20
OC curves, 471, 473 of multiplication, 21–22
Poisson unity values, 471 Girshick plan, 430–431
reference plan, 468 Grand lot hypothesis, 501, 504–506,
reference single-sampling plan, OC curve 513–514
for, 471 Group sequential plans, 161
SkSP-2 sampling plan, 470–471, 473
SkSP-1 skip lot sampling plan, 468–469
H
Dodge–Romig plans
AOQL, 402, 404 Harvard table, binomial distribution, 629–636
LTPD, 402–403 H-108 handbook
double-sampling attributes plans, application of, 561
151–156 check sequence for selecting plan from, 562
exponential distribution, 73–74 different types of tests provided in, 556
hypergeometric distribution, 72 life test sampling plan code designation, 815
negative binomial distribution, 73 life tests terminated at preassigned time,
OC curve, 27–30 817–818
Poisson distribution, 73 life tests terminated upon occurrence, 816
probability of acceptance, 398 operation of, 556, 558–561
sequential sampling, 166–170 proportion failing before specified time, 820
single-sampling attributes plans, 123–126 selection of plan, 558
template-generated plan, 402 sequential life tests, 819
Exponential distribution, 61 structure of, 557
distribution function for, 54 Hypergeometric distribution, 61
Excel, 73–74 approximations, 66–68
failure rate, 53–54 Excel, 72
frequency (density) function, 53–54 four-card hand of aces, probability of, 42–43
mean of, 54 frequency function of, 41
Minitab, 76–77 Lieberman–Owen table, 623–628
parameter of, 53 mean, mode, and median deviation, 43
Thorndyke chart, 54 mean of, 43
values of e–x, 620 Minitab, 74–75
Exponential tables, 65 recursion formula, 41–42
standard deviation, 43
Type A probability of acceptance
F
for plan n = 10, c = 1, 86–87
Frequency function, 35 for plan N = 20, c = 1, 88
binomial distribution, 43–44 for plan N = 20, n = 10, 89
f-binomial distribution, 49 when sample size is proportionate to lot
for die, 36–37 size, 90
exponential distribution, 53–54 Hypergeometric tables, 60, 63, 65, 119–121
834 Index

I Dodge–Romig scheme
double sampling, 380
Independent and dependent mixed plans,
single sampling, 379
366–367
Type A probabilities, 373
Independent events, 19–23
fixed consumer risk, 8
Independent mixed plan, 358–360
limiting quality, 26
Indifference quality (IQ), 25, 99–100, 586
“LTPD.xla,” 398
setting of, 590
plan attributes, 128
Inspection error, 90–92
RQL, 125
ISO 2859, 297
setting of, 590–591
ISO 21247, 498
Lower limit chart, 187
Lower specification limit, 181–184
J LSP, see Lot sensitive sampling plan
LTPD, see Lot tolerance percent defective
Joint Committee for the Development
of Statistical Applications in
Development and Manufacturing, 7 M
Markov chain approach, 423, 437
K Matched single and double sampling plan,
682–685
k method, 225–226 Matched single-sampling plans
Kolmogorov–Smirnov test, 69 average sample size reduction, 255
unity values, 779
Military Standard 414 (MIL-STD-414)
L
application of, 311
Ladany nomograph, 369 AQL conversion table, 731
Larson binomial nomograph, 45–46, 116 conversion to ANSI/ASQ Z1.9, 321–322
Lieberman–Owen tables, 120, 623–628 dependent mixed plans, 365–366
Life cycle, of acceptance control application, 582 derivatives
Limiting quality (LQ), 26 ANSI/ASQ Z1.9, 322–325
in defects, 713, 715, 720–721 ISO 3951-1, 323–326
in percent defective, 712, 714 implementation
Lot plot plans, special sampling Form 1, 319
completed modified form, 340–341 Form 2, 317–318
Pearson–Tippett method, 339 range and variability known, 320
statistical acceptance sampling techniques, 339 lot percent defective
3σ values, 342–343 reduced inspection, 729–730
ULL and LLL, 341, 344 standard deviation methoda,
Lot sensitive sampling plan (LSP) 735–748
applications, examples of, 482–483 measures of plans, 316–317
hypergeometric distribution, 484 and MIL-STD-105E, 320–321
hypergeometric probability distribution, 478 MSD, 751
Poisson distribution, 484 normal and tightened inspection, 733, 749
procedure, 478–480 operation of, 311–313
producer’s risk, 481–482 reduced inspection, 734, 750
protection, 479, 481 sample size code letter, 732
zero defectives, 478 selection
Lot size, 87–90 k method, 315–316
Lots, stream of M method, 315–316
plan, application of, 214–216 range method, 314
Lot tolerance percent defective (LTPD) sequence for, 314
ASTM International Standard E1994, 398 standard deviation method, 313
ATI, 381 variability, 314
Index 835

software normal inspection, 693

Excel, 326–329 reduced inspection, 695
Minitab, 327 tightened inspection, 694
Statgraphics, 328, 330–334 TR7
structure, 310 AQL, 822
tightened inspection, 727–728 limiting quality level, 823–828
Military Standard 1916 (MIL-STD-1916) MIL-STD-105E system, 133–137, 140
ASTM, 499 application of, 285
attributes sampling plans, 806 AQLs, determination of, 283–284
code letters, 805 implementation of, 289–294
continuous sampling plans, 808 AOQ curve, 291–292
implementation, 496–498 ASN curve, 290–291
measures, 498 ATI curve, 292
operation, 494–496 scheme OC curve, 290–291
quality management system, 298 and individual sampling plans, 294–296
structure, 494 lot size, determination of, 283
variables sampling plans, 807 measures of, performance of individual
Military Standard 105E (MIL-STD-105E) plans, 286
scheme, 309, 320–321 scheme properties, 287–289
AOQL selection of plan, 283–286
for normal plans, 711 switching rules, operation of
for tightened plans, 712 normal inspection, 282–283
ASN curves, double and multiple sampling, reduced inspection, 282–283
716 tightened inspection, 282
average outgoing quality limit Minitab, 300
factors, 710, 719 acceptance sampling, attributes, 543
Code F AoZ plan, 543
average outgoing quality curves, 724 binomial distribution, 75–76
average sample number curves, 723 exponential distribution, 76–77
average total inspection curves, 725 hypergeometric distribution, 74–75
operating characteristic curves, 722 LSP plans, 544
double sampling negative binomial distribution, 75, 77
normal inspection, 696 normal distribution, 78
reduced inspection, 698 OC curve, 29–32, 108–110
tightened inspection, 697 Poisson distribution, 75–76
LQ single-sampling attributes plans, 125–127
in defects, 713, 715, 720–721 snap sampling plans!, 544–545
in percent defective, 712, 714 Squeglia plans, 545
multiple sampling Weibull distribution, 77–78
normal inspection, 699–702 Mixed plans, joint probabilities for, 769–775
reduced inspection, 707–709 Mixed variables
tightened inspection, 703–706 grand lot scheme, 514
OC curves and probability points for plans, special sampling
code letter, 717–718 advantages and disadvantages, 356–357
operating ratios, 726 attributes plan, 354–356
optimal narrow limit plans generalized mixed dependent procedure,
normal inspection, 761–763 357–358
reduced inspection, 764–766 Modified Kemp procedure, 192
tightened inspection, 758–760 Molina table, Poisson distribution, 637–648
reduced inspection, limit numbers Multilevel plan (MLP)
for, 691 advantages of, 427
sample size code letter, 692 AFI, 426
single sampling AOQL, 423, 425
836 Index

infinite-level plan, 423–424 Negative binomial distribution, 61

Markov chain approach, 423 binomial distribution function, 52
sampling frequency, 424 cumulative negative binomial, 51
Multiple-sampling attributes plans Excel, 73
ASN curve, 139–140 frequency function, mean and standard
average sample size, reduction in, 133 deviation, 51
binomial/Poisson probability distributions, Minitab, 75, 77
143 parameters of, 51
control table, evaluation, 148–151 sampling plan, 52–53
efficiency, 133 successive values of, 51
flexibility, 133 Negative binomial tables, 64–65
measures of, 143–146 NLG, see Narrow-limit gauge
MIL-STD-105E, 133 No-Calc procedure
OC curve, 133, 139 MIL-STD-414, 337
operating ratios, 136 plotting positions values, 752
probability tree, 143, 145 Nominal center line (NCL), 176–177
procedure for, 134–136 Nomograph
Statgraphics, 154 samples having zero defectives, 515
tabular form, 135 zero defectives sample, 810
unity values, 135–136 Nonconformance criteria, 464
unity values, construction and evaluation, Normal distribution, 62
653–661 APL and RPL, 178
Mutually exclusive events, 19–23 central limit theorem, 59–60
cumulative normal probability, 621–622
frequency function, 57
N
Minitab, 78
Narrow-limit gauge (NLG) parameters, 57
assumption of normality, 354 proportion of bolts, 58–59
control charts, 354 standard normal frequency (density)
distribution, 346 function, 57
go/no-go data, 344 tables, 58, 65
implementation steps, 345
known standard deviation plan, 352
O
Ladany
iterative Use of, 351 Odeh–Owen tables, 256–258, 687–690
nomograph, 348–351 Operating characteristics (OCs) curves, 374, 415,
notation, 348 450, 461
MIL-STD-105E double-sampling plans, 133, 154, 384
AQL system and switching rules, 352 Excel templates, 27–30
average sample size, 353 inspection error, 92
properties, 353 Minitab, 29–32, 108–110
OC curves for, 346 multiple sampling, 133
optimum narrow-limit plan, 351–352 one-sided (upper tail) chi-square test, 675
Poisson approximation, 347 one-sided normal test, 673
probability of acceptance, 346 for plan n = 10, c = 1, 105
quantities, 345 PQL and CQL, 26
Schilling-Sommers table, 352 PQL, CQL, PR, CR, 99–101
variables sampling plans, 344 probability distributions, 83
Narrow-limit plans, 369 probability of acceptance, proportion
advantages of, 352 defective, 23–25
probability of acceptance and ASN, AQL, probability points, 25–26
767–768 process parameter, single sampling for, 174–175
NCL, see Nominal center line sample size and lot size, 87–90
Index 837

sampling plan, 23 plot, 69

for single, double, and multiple matched empirical cumulative probability
plans, 139 distribution, 70
single-sampling attributes plan straight-line plot, 71
acceptance number, 384–385 Weibull distribution, 72
AOQL, 384 random number tables, 15
Excel template, 125 random sampling, 14–15
hypergeometric tables, 119–121 subjective definition, 14
Larson binomial nomograph, 116 Probability distribution, 565
Minitab, 125, 127 binomial distribution (see Binomial
Poisson unity values, tables of, 115 distribution)
Statgraphics, 128–129 frequency and distribution function
two–sided normal test, 673 for die, 36–37
Type A (see Type A OC curve) tabular form, 36
Type B (see Type B OC curve) for watch stoppage, 37–38
Optimum sample size, 197 hypergeometric distribution (see
Hypergeometric distribution)
OC curves, 83–84
P
tables for evaluation of
Pascal distribution, 51 binomial tables, 63, 65
Pearson–Tippett method, 339 exponential and Weibull tables, 65
Permutations, 16–17, 19 hypergeometric tables, 60, 63, 65
Poisson distribution, 61, 373 negative binomial tables, 64–65
Excel, 73 normal distribution tables, 65
frequency function, 45 Poisson tables, 64–65
mean and standard deviation, 45 Probability of acceptance, 18–19
Minitab, 75–76 double-sampling plan, 137, 141–142
Molina table, 637–648 for hypergeometric, f-binomial, binomial,
number of bubbles per piece and Poisson, 84–85
probability distribution, 47 ISO 3534-2 standard definition of, 104
process average, 46–47 multiple-sampling plan, 144, 149
standard deviation, 47 single-sampling attributes plans,
parameter in, 45 121–122
recursion formula, successive values of, 46 Probability plot, 337–339
Thorndyke chart, 47–48 Process
Poisson tables, 64–65 capability, 6
Poisson unity values, tables of, 114–115 improvement study, 6
Pólya distribution, 51 optimization study, 6
PQL, see Producer’s quality level performance check, 6
Primary units, 197 performance evaluation, 6
Probability Process parameter, variables sampling for
classical definition, 13–14 acceptance control chart, 175–179
counting possibilities APL vs. RPL, 173
combinations, 17–18 CUSUM chart, 190–193
permutations, 16–17, 19 sequential plans for
probability of acceptance, 18–19 Barnard sequential chart, 187–189
tree diagram, 19 Barnard sequential t-test, 189–190
distribution of (see Probability distribution) double specification limits, 184–185,
empirical definition, 13–14 187–188
independent events, 19–23 lower limit chart, 187
mutually exclusive events, 19–23 lower specification limit, 181–184
OC curve (see Operating characteristics upper limit chart, 185–187
(OCs) curves) upper specification limit, 179–181
838 Index

single sampling Rectification schemes, 3

OC curves, 174–175 Anscombe rectifying inspection procedure
statistical tests of hypotheses, measures, 394–395
173–174 operation, 391
specifications, 173 selection, 392–394
Process quality control credit-based schemes, 395–397
process optimization study, 6 Dodge–Romig plan derivatives, 398
Shewhart control chart, 4–6 Dodge–Romig scheme (see Dodge–Romig
Producer’s quality level (PQL), 26, 133, 135, 137, sampling schemes)
139–140, 586 Excel
OC curve, 99–101 alternative sampling plan, 401
setting of, 593 AOQL sampling plan derivation, 402
single-sampling attributes plans, 113 BETAINV function, 398
hypergeometric tables, 120 BINOM.DIST function, 399
Poisson unity values, tables of, 115 binomial probability distribution, 398
Producer’s risk (PR) D & R AOQL plans, 402, 404
and consumer risks, 99–101 D & R LTPD plans, 402–403
single-sampling attributes plans, 113 probability of acceptance, 398
hypergeometric tables, 120 template-generated plan, 402
Poisson unity values, tables of, 115 single-sampling AOQL plans, 372–373
Reduction variance, 197
Rejectable process level (RPL), 173–178
Q
Rejectable quality level (RQL), 125
Quality Control and Reliability Handbook H-108, Resnikoff–Lieberman table, 239–240
555–556
Quick switching system (QSS), 277–278
S
QSS-2(n; cN, cT), 792
QSS-2(n; kn, c0), 491–492, 795–796, 802 Sample
QSS-3(n; cN, cT), 793 construction of, 196–198
QSS-3(n; kn, c0), 492, 797–798, 803 types of, 197
QSS-4(n; cN, cT), 794 Sampling plans
QSS-4(n; kn, c0), 492, 799–800, 804 acceptance quality control, 1–3
QSS-1(n, kn; 0) plan, 492–493, 801 AOQ curve, 105–106
QSS-2(n, kn; 0) plan, 493 ASN curve, 105–106
QSS-3(n, kn; 0) plan, 493 ATI curve, 105, 107
QSS-3(n; cN, cT) plan, 491 basic measures of, 104
QSS-4(n, kn; 0) plan, 493–494 inspection error, 90–92
QSS-4(n; cN, cT) plan, 491 ISO 3534-2 standard, 103
single-sampling plan, 490 OC curves (see Operating characteristics
tightened inspection, 489 (OCs) curves)
unity values for, 786–791 principal measures of, 104–105
progression of, 582
quality levels determination, 586
R
rectification, 92–95
Random numbers, 617–619 selection and implementation of,
tables, 15 583–585
Random sampling, 14–15 Schilling–Johnson table, 369
Rational sentencing, 458 Scope of the International Standard ISO 10725,
Rayleigh distribution, 572 216–218
Rectification plan Secondary units, 197
AOQ, 92–95 Sentencing rule, 459
AOQL, determination of, 93–94 Sequential acceptance plot, 160
ATI curve, 93–94 Sequential graph, 163
Index 839

Sequential plans average total inspection, 122–123

Barnard sequential chart, 187–189 probability of acceptance, 121–122
Barnard sequential t-test, 189–190 Poisson unity values, tables of, 114–115
double specification limits, 184–185, 187–188 procedures for, 113–114
lower limit chart, 187 selection, 113–114
lower specification limit, 181–184 software applications
upper limit chart, 185–187 Excel template, 123–126
upper specification limit, 179–181 Minitab, 125–127
Sequential sampling, 8 StatAdvisor, 129–130
attributes Statgraphics, 127–130
procedure for, 160 Thorndyke chart, 118–119
common logarithms for, 672 Single-sampling variables plans
defects per unit, 165–166 known and unknown standard deviation,
measures, 163–164 682–685
operation, 159–161 matching attributes and narrow limit plans,
plans 754–757
natural logarithms, 161 Skip-lot sampling plans
selection, 161–163 demerit rating plan
procedure, for defects per unit, 165–166 check inspection, 462
software applications, 166–170 defects, classification of, 463
Shewhart control chart, 4–6 nonconformance criteria, 464
Simon’s approach, 500–501 Poisson limits, 464
Simplified grand lot procedure samples sizes, 462
acceptance sampling, 500 SQL, 465
approximations, 500 SkSP-1, 445–447
attributes, 502–504 SkSP-2
continuing series of lots, 510–513 ASN curves, 451
control chart, 499 Dodge–Perry table, 451
maverick limits, 513 lot AOQL2 values, 449
means procedures, 513 measures of, 448
MIL-STD-105E, 499 normal inspection, 448
mixed variables–attributes procedures, 514 OC curves, 450
Poisson distribution, 514 reference plan, 449
Simon’s approach, 500–501 single-sampling plans, 452
variables skipping inspection, 448
acceptability criterion, 510 unity values, 451, 779
grand lot hypothesis, 505–507 Snap Sampling Plans!, 300–301, 442–443
means chart analysis, 505, 508 Special sampling
MIL-STD-414, 509 lot plot plans
muzzle velocity, 506 completed modified form, 340–341
standard sampling plan, 506 Pearson–Tippett method, 339
standard variables acceptance procedures, statistical acceptance sampling
510 techniques, 339
Simplified grand lot sampling, 809 3σ values, 342–343
Single-sampling attributes plans, 107–108 ULL and LLL, 341, 344
f-binomial application of nomograph, 116–118 measures
curtailment of, 94–96 dependent mixed plan, 360–365
hypergeometric tables, 119–121 independent and dependent mixed plans,
Larson binomial nomograph, 116 366–367
measures of independent mixed plan, 358–360
AOQL, 121–123 MIL-STD-414 dependent mixed plans,
average outgoing quality, 122–123 365–366
average sample number, 122 various plans, comparison of, 367–368
840 Index

mixed variables Technical Report TR7

advantages and disadvantages, 356–357 check sequence for determining the
attributes plan, 354–356 procedure for, 577
generalized mixed dependent procedure, hazard rate criteria, 568
357–358 mean life criterion, 566–567
narrow-limit gauge with MIL-STD-105E system, 576
assumption of normality, 354 operation of, 571
control charts, 354 reliability criteria, 568
distribution, 346 structure of, 570
go/no-go data, 344 Thorndyke chart, 47–48, 54,
implementation steps, 345 118–119, 592
known standard deviation plan, 352 Tightened–normal–tightened (TNT)
Ladany nomograph, 348–349 plans, 279
Larson nomograph, 351 parametric values of, 785
MIL-STD-105E AQL system and TNT scheme
switching rules, 352 MIL-STD-105E, 484
OC curves for, 346 procedure, 485
optimum narrow-limit plan, 351–352 protection, 485–486
Poisson approximation, 347 selection, 487–489
probability of acceptance, 346 Tolerance intervals, 97–99
quantities, 345 Total quality management (TQM), 515
Schilling–Sommers table, 352 Tree diagram, 19
variables sampling plans, 344 Two-point single-sampling plans, 113
No-Calc plan, 337–339 Type A OC curve
Squeglia plans, 516–518 defects per 100 units, 84–85
Standard deviation probabilities of acceptance, 84–85
continuous distributions, 40 probability distributions, 83–84
exponential distribution, 54 Type A probability of acceptance
factors and degrees of freedom, 680–681 hypergeometric distribution
Weibull distribution, 55 for plan n = 10, c = 1, 86–87
Standard error, 201 for plan N = 20, c = 1, 88
Standard quality level (SQL), 465 for plan N = 20, n = 10, 89
StatAdvisor, 129–130, 303–305 when sample size is proportionate to lot
Statgraphics, 301–305, 328, 330–334 size, 90
double-, and multiple-sampling plans, 154 producer’s and consumer’s interests, 100
MIL-STD-1916 plans, 545–547 Type B OC curve
single-sampling attributes plans, 127–130 ARL values, 86–87
StatAdvisor, 547 defects per 100 units, 84–85
Statistical process control (SPC), 428, 494 for plan n = 20, c = 1, 97–98
Statistical Research Group, 8, 91 probabilities of acceptance, 84–85
selected values, 671 probability distributions, 83–84
table of sequential sampling plans, producer’s process, 84
662–670 Type B probability of acceptance
Subjective probability, 14 AOQ and ATI, calculation of, 93
Switching rules, 277 using binomial distribution
normal inspection, 282–283 for c = 1, 88
reduced inspection, 282–283 for c/n = 1, 89
tightened inspection, 282
U
T
Unlimited AOQL (UAOQL), 412
Technical Report (TR3), 566 Upper limit chart, 185–187
table of values for percent truncation, 821 Upper specification limit, 179–181
Index 841

V on sample range, 248–249

selection, 231–234
Variables plans for life testing and reliability,
sequential sampling plans, 259–260
554–555
software applications, 262–273
Variables sampling plans, 7–8, 682–685
specification limits, 222
acceptance criteria for, 228
theoretical nature of, 223
acceptance polygon, 229–230
tolerance intervals, 255–259
advantages/disadvantages, 221–222
AOQL, 231–232
application of, 225–226 W
attributes, comparison of, 222 Wald–Wolfowitz Plan, 428–430
distributions in, 223 Wallis method, polygon derived by, 231
double-sampling plans, 252–255 Weibull distribution, 62, 555, 565–566, 568
double specification limits, 226–231 distribution function, 55
acceptance criteria for, 228 failure rate, 56–57
extensive tables of, 231–232 frequency (density) function, 55–56
formulas, 232–233 gamma function, 55
Jacobson nomograph hazard rate, 56–57
application of, 234 location parameter, 55
operating characteristics curves for, mean, 55, 57
234–235 Minitab, 77–78
for plan selection, 233–234 probability plots, 72
for proportion defective, 233 scale parameter, 55
standard deviation known, 234 shape parameter, 55–56
standard deviation unknown, 234 standard deviation, 55
lot tolerance percent defective (LTPD), 231 Weibull tables, 65
measures, 234–243 Western Electric Committee on Rating of
M method, 243–248 Quality of Telephone Products, 7
double specification limits, 246–247 Wilk–Shapiro test, 69
single specification limits, 246
standard deviation known, 243–244
unknown, standard deviation, 244, 246 X
noncentral t-distribution, 686 x̅ method, 225
OC curves, comparison of, 224
operation, 225–231
Z
for percent nonconforming, 255–259
performance, measures of, 242–243 Zero plans
principal advantage of, 222 AoZ and AQL plans, 517, 519
for process parameter (see Process parameter, chain sampling alternative, 519–521
variables sampling for) ChQSS, 521–524
for proportion nonconforming, 221 Squeglia plans, 516–518
Resnikoff–Lieberman table, 239 U.S. Department of Defense approach,
sample range, 248–252 515–516