0% found this document useful (0 votes)

755 views158 pages

ASTM - STP 588 - Manual On Statistical Planning and Analysis

Uploaded by

faruk öztürk

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

755 views158 pages

ASTM - STP 588 - Manual On Statistical Planning and Analysis

Uploaded by

faruk öztürk

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 158

MANUAL ON

STATISTICAL PLANNING
AND ANALYSIS FOR
FATIGUE EXPERIMENTS

Sponsored by
Committee on Publications
AMERICAN SOCIETY FOR
TESTING AND MATERIALS

ASTM SPECIAL TECHNICAL PUBLICATION 588

R. E. Little
University of Michigan, Dearborn Campus
Assisted by E. H. Jebe
Environmental Research Institute of Michigan

List price $15.00

04-588000-30

AMERICAN SOCIETY FOR TESTING AND MATERIALS

191 6 Race Street, Philadelphia, Pa. 191 03

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
(~) BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1975
Library of Congress Catalog Card N u m b e r : 75-13060.

NOTE
The Society is not responsible, as a body,
for the statements and opinions
advanced in this publication.

Printed in Lutherville-Tlmonium, Md
November 1975

In 1949 Committee E-9 on Fatigue published ASTM STP 91, Manual

on Fatigue Testing. Seven years later E-9 published ASTM STP 91A, A
Guide for Fatigue Testing and the Statistical Analysis of Fatigue Data.
Both publications were modest efforts, presenting what was then considered
the current practice and views of Committee E-9 members. Subsequently
effort was expended intermittently in Committee E-9 for over a decade in
various attempts to revise and expand these STP's to reflect advances in
the state of the art and in statistical analyses. This activity ultimately led
to ASTM STP 566, Handbook of Fatigue Testing, and to the present
Manual on Statistical Planning and Analysis for Fatigue Experiments.
The present manual is the last of several attempts to revise ASTM
STP 91A to include more information on statistical planning of fatigue
experiments. The reader will find in the first two chapters an effort to present
and explain the basic terminology and fundamental concepts of planned
experiments, and to indicate the necessary steps in a professional statistical
analysis (using a simple randomized complete block fatigue test program
for purposes of illustration). The last two chapters cover statistical planning
of S-N and fatigue strength (response) tests, with specific recommendations
regarding specimen replication; and nonparametric comparison of two
censored life samples, with extensive computer tables.
Overall, the intent is to provide sufficient information to permit fatigue
investigators to conduct fatigue test programs with statistically planned
organizational structures.
In the acknowledgments at the end of this manual I have expressed
my appreciation to those individuals who contributed to this manual.
As Chairman of Subcommittee E09.06, I would also like to express my
sincere appreciation to all those individuals who worked on various
antecedents to the present manual.
R. E. Little, Chairman
ASTM Committee E09.06 on
Statistical Aspects of Fatigue

Handbook of Fatigue Testing, STP 566 (1974), $17.25, 04-566000-30

The Influence of State of Stress on Low-Cycle Fatigue of Structural

Materials: A Literature Survey and Interpretive Report, STP 549
(1974), $5.25, 04-549000-30

Fatigue at Elevated Temperatures, STP 520 (1973), $45.50, 04-520000-30

This publication is made possible by the authors and, also, the unheralded
efforts of the reviewers. This body of technical experts whose dedication,
sacrifice of time and effort, and collective wisdom in reviewing the papers
must be acknowledged. The quality level of ASTM publications is a direct
function of their respected opinions. On behalf of ASTM we acknowledge
with appreciation their contribution.

A S T M Committee on Publications

Jane B. Wheeler, Managing Editor

Helen M. Hoersch, Associate Editor
Charlotte E. Wilson, Senior Assistant Editor
Ellen J. McGlinchey, Assistant Editor

Introduction 1

Chapter 1--Fundamental Concepts in the Statistical Planning (Design) of

Fatigue Experiments 3
Introduction 3
Organizational Structure of Planned Experiments 6
Paired Comparison Test Program 6
Elementary Uniformity Trial Test Program 9
Completely Randomized Design (CRD) Test Program 16
Ramdomized Complete Block (RCB) Test Program 18
Elementary Split Plot Test Program 25
Summary and Perspective 28
Appendix 28

Chapter 2--Necessary Steps in Statistical Analysis for Planned Experiments 34

Introduction 34
Necessary Steps in Analysis of Variance (ANOVA) for a Paired Com-
parison Program 34
Step 1 : Write the Mathematical Model 34
Step 2: Plot the Transformed Data 38
Step 3: Explain the Data (by appropriate parameter estimation) 39
Step 4: Statistical Analysis (generally analysis of variance (ANOVA)) 41
Summary 45

Chapter 3--Planning S - N and Response Tests 46

Introduction 46
Conventional S - N Tests 46
Preliminary and Exploratory Tests 46
Replicated S - N Tests 49
Replication Guidelines for S - N Testing 51
Discussion of Guidelines 52
Fatigue Strength and Fatigue Limit Tests 52
Probit Method 52
Up-and-Down Method 54
Two-Point Method 54
vii
Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
viii CONTENTS

Chapter 4--Distribution-Free Statistical Tests for Comparing Two Censored

Life Samples (in a CRD Test Program) 61
Nomenclature 61
Introduction 61
Complete (uncensored) Data 62
Censored Data 62
Test of Hypothesis 63
Alternative Hypotheses 64
Complete (uncensored) Data 64
Mann and Whitney's U Statistic 65
Mann and Whitney's Tables 65
Censored Data 66
Run-Outs 66
Halperin's Un Statistic 71
Halperin's Tables 71
Accelerated Comparisons 72
Young's Ur Statistic 112
Young's Tables 113
Appendix I 129
Appendix II 145

References 147

Acknowledgments 149

Introduction

The Manual on Statistical Planning and Analysis for Fatigue Experiments

attempts to present and explain the basic terminology and fundamental
concepts of planned experiments, and to provide sufficient information to
permit fatigue investigators to plan the organizational structure of simple
fatigue test programs.
The major thrust of this manual is that planned fatigue experiments
involve an understanding of two basic factors: (1) the test program objec-
tive, and (2) the role of various test program constraints on the resultant
organizational structure. As these two factors differ markedly from test
to test it is impossible to present herein a detailed step by step procedure
for test program planning. Nevertheless, it is generally a relatively simple
matter to use the fundamental concepts presented herein to plan efficient
effective fatigue test programs that not only accomplish the major test
objectives but also satisfy the specific test constraints involved.
Chapter 1 introduces the concept of organizational structure in planne.d
experiments. It emphasizes the use of randomization, blocking (planned
grouping), and replication in establishing the desired organizational
structure.
Chapter 2 enumerates and briefly illustrates the necessary steps in a
statistical analysis relevant to most planned experiments: a mathematical
model pertinent to the given organizational structure, appropriate data
plots, parameter estimation using well established statistical techniques, and
subsequent statistical analysis, generally analysis of variance (ANOVA).
Chapter 3 treats the planning of S-N tests and small sample response
tests. Emphasis is placed on replication at a few selected stress or strain
amplitudes.
Chapter 4 presents statistical tables for use in comparing two life samples
(either complete or censored) when no assumption is made regarding the
parametric form of the underlying distribution, that is, whether the life
distribution is log-normal, Weibull, or whatever. The nonparametric
(distribution-free) analysis based on the tables herein is an acceptable
alternative to the analysis of variance (ANOVA) recommended in Chapter

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Copyright*1975
Downloaded/printed by by ASTMInternational www.astm.org
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
2 INTRODUCTION

2, when the life data for the two samples are generated using a completely
randomized design (CRD) test program.
Overall, the intent is to provide sufficient information to permit fatigue
investigators to conduct fatigue experiments (test programs) with statis-
tically planned organizational structures. Such planned structures are
mandatory for test programs with severe time and cost constraints. Namely,
structured experiments are efficient experiments!

Chapter 1 Fundamental Concepts in the

Statistical Planning (Design) of Fatigue
Experiments

Introduction
No manual can provide the fatigue investigator with a complete step-
by-step detailed procedure which is valid for the statistical planning of
experiments whatever the situation. In fact, only certain very simple fatigue
test programs fit precisely into the specific formats required for well-
established planned experiments, such as the completely randomized
design (CRD) and the randomized complete block (RCB) design, [1].~
Generally these simple fatigue test programs pertain to either elementary
comparative tests (for example, comparing the fatigue life of Material A
versus Material B), or to quality assurance tests (namely, the generation of
certain fatigue data under well-defined test conditions). On the other hand,
most (exploratory) research programs involve one or more (sometimes
subtle) constraints peculiar to the specific situation, that is, to the given
material processing, specimen preparation, test machine, environment, or
whatever. Such constraints often preclude elementary statistical analysis of
the resulting data and may even present difficulties to a trained statistician,
particularly if he is consulted only after the tests have been conducted.
But whatever the nature and the complexity of the given fatigue test situa-
tion, there are certain design of experiments fundamentals which must
appear in the planning and conduct of any competent experimental pro-
gram. It is the objective of this chapter of the manual to state these funda-
mentals (presented in italics in the following paragraph) and to illustrate
their application in a few example situations. For further specific references
to the design of experiments, see Refs 2, 3, and 4.
The characteristics of experimental design and the design of experiments
terminology that should be clearly in mind at the formative stage are as
follows:
First, and foremost, a well-planned experiment has a welt-defined
organizational structure which is clearly described by appropriate figures,
charts, and tables.

1 The italic numbers in brackets refer to the list of references appended to this manual.
3
Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Copyright* 1975
Downloaded/printed byby ASTM International www.astm.org
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
4 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

Second, well-planned experiments employ clearly defined experimental

units which are not only structured appropriately in blocks (planned groups),
but upon which the desired treatments are applied following the procedure
known as mechanical randomization.
Third, well-planned experiments employ treatment levels and treatment
combinations carefully selected (generally using data from preliminary
tests or other reliable information) such that the desired treatment effect
estimates are not confounded with nuisance variables, that is, such that the
associated mathematical model and the corresponding statistical analysis
of the data are not unduly complicated.
Fourth, a well-planned experiment employs sufficient replication, that is,
provides sufficient specimens to ensure reasonably precise estimates of the
desired treatment effects.

These fundamental design of experiments concepts (in italics) are dis-

cussed and illustrated in some detail subsequently. However, to provide an
introductory statement of their meaning, the following brief definitions are
provided:

Organizational Structure--The organizational structure is a descriptive

device by means of charts, figures, or tables that clearly shows the inter-
relation of all the variables in a test program, the order of testing based on
appropriate randomization procedures, and suggests the nature of the
statistical analysis that will be used to draw inferences from the resultant
data.

Experimental Units--Experimental units generally are the individual

specimens to be tested in the time order indicated by the organizational
structure. Experimental units are often grouped is a specific way according
to the variables important to the test program.

Blocks--A block is a group of experimental units that are or may be

more homogeneous than other groupings of experimental units. Blocks are
employed in planning experiments as a device to counteract (circumvent)
the effect of nuisance variables. For example, the objective of a fatigue test
program may be to evaluate the difference in fatigue response between a
conventionally machined surface and a shot peened surface. To obtain
sufficient specimens, two sheets of test material may be needed, for which it
may be expected there will be subtle differences in fatigue behavior. "The
variability in fatigue behavior among sheets is regarded as a nuisance
variable when this variability tends to cloud the test objective. Partition
of the experimental units into blocks (groups) should be accomplished
here in such a manner that sheet-to-sheet variability does not affect the
estimate of the difference in fatigue response of specimens with conven-
tional machining and shot peening.

Treatments--Treatments are the major variables in a test program,

whose effects are to be studied. For materials tests, treatments may include
evaluation of the effect of alloying, various heat treatment procedures,
surface preparation, temperature, environment, speed of testing, notches,
etc. Treatments are assigned randomly to experimental units.
Mechanical Randomization--The detailed procedure for assigning treat-
ments to experimental units, time order of testing, etc., using random
numbers. The mechanical procedure is illustrated step by step in the
Appendix at the end of this chapter.
Treatment Levels--In large-scale fatigue experiments, the effect of certain
treatments may be studied over a specific range of interest of the treatment
variable. For example, the effect of tempering temperature may be studied
by conducting fatigue tests on several groups of identical experimental
units, each group having been tempered at some specific nominal tempera-
ture level. The treatment levels are the individual tempering temperatures
imposed on the respective groups of experimental units, for example,
600, 700, 800, 900, 1000~ Or, several levels of plate thickness may be
selected for the study of the (ostensive) effect of thickness on the fatigue
response. In this case, the treatment levels may be 1,/4, 3/8, 1/2, 5,;8, and
3/4 in., respectively.
NOTE 1: Treatment levels may be either quantitative or qualitative.
For example, qualitative levels for surface preparation of specimens may
be shot peened, and not shot peened (say, conventionally machined,
mechanically polished, or even "as received").
NOTE 2: Generally the precise meaning of treatments and their levels
will be more clearly defined when the materials test engineer exerts some
direct control over the treatment selections and the detailed processing of
the associated treatment levels.
Treatment Combinations--Application of two or more treatments
(major variables) to the same experimental unit. In certain cases the time
order in which the treatments are applied is important. Such information
must be apparent in the organizational structure.
Treatment Effect Estimates--The estimate, positive or negative, of the
incremental effect on fatigue response associated with all experimental
units receiving the given treatment.
Confounding--A situation where a desired treatment effect estimate
cannot be separated (isolated) from other sources of variation, for example,
from block-to-block variability.
Nuisance Variables--All variables which may affect fatigue response
but which are of no direct interest in the given test program at hand. The

purpose of blocking (that is, planned grouping) is to isolate the treatment

effect estimates from the spurious variability associated with differences
among blocks.

Replication--A replicate is the complete collection of observations

comprising one observation on each of the treatments (major variables)
or treatment combinations. Replication, in turn, is the number of complete
replicates involved in the overall test program. ~-
No-rE: Depending on the organizational structure of a given test program,
certain treatments may be replicated more often than other treatments.
These fundamental concepts are valid in planning any test program
whether or not the data are subjected subsequently to a statistical analysis.
These concepts pertain just as much to the well-established S-N test as to
the most sophisticated structured experimental design.

Organizational Structure of Planned Experiments

Well-planned experiments are recognized easily--their organizational
structures are apparent in simple diagrams. In the paragraphs that follow,
several simple examples of planned experiments are described, each with
certain advantages and disadvantages resulting from its special organiza-
tional structure, but all being composed of three essential elements: (1)
experimental units, (2) blocks, and (3) treatments. These examples are
generally too simplified to be considered practical relative to diverse appli-
cations in fatigue testing. The primary purpose of the examples is to illus-
trate the fundamental concepts enumerated in the Introduction to this
chapter.

Paired Comparison Test Program 3

Consider the test objective of determining how shot peening affects the
fatigue life of specimens of Material M relative to conventional mechanical
polishing. The constraints placed on time for testing and the funding avail-
able may dictate a simple program to provide a quick engineering estimate
of the true effect. A test program structure that may be utilized in this case
is termed the paired comparison test program. The basic feature of this
program is that two treatments are compared under test conditions that
within pairs are as identical as practical. The two treatments in this example
are shot peening and mechanical polishing. To emphasize the idea that all
other test conditions are as identical as practical within pairs, this first
2Table I displays four complete replicates of four treatments. The first replicate is
comprised of specimens I5, 14, 11, and 3; the second replicate of specimens t, 9, 16, and 7;
etc. Tables 5a and b provide examples of four replicates of four treatment combinations.
3 Commonly termed a "Paired t Test Program," It Ls also a randomized complete block
(RCB) program with two treatments per block.

example proceeds under the assumption that the test material required
(for example, a single bar) may be considered homogeneous, that a single
machine and technician are used in testing, and that all tests are conducted
at the same nominal stress amplitude. The only nuisance variable considered
in this first example is the possibility that subtle changes in machine
performance may occur over the time period in which this test program is
conducted. Thus, experimental units are tested in pairs, back to back in
time order, one experimental unit in each pair receiving Treatment A,
shot peening, the other Treatment B, mechanical polishing. The time order
of testing within pairs is randomized as illustrated by the organizational
structure of Fig. 1.

Organizational Struct,re--Figure 1 summarizes the paired comparison

test program, providing details of how the individual tests should be con-
ducted. This vital test-conduct information is often obscure when the test
program is defined merely by providing a mathematical model or stating
certain design of experiments nomenclature. Note that Fig. 1 indicates
that 2N homogeneous specimens are separated into N pairs of two, and then
two treatments denoted A and B are randomly applied to one member of
each pair, after which the fatigue tests are conducted in random time order
within each pair (and the pairs themselves are tested in random time order).
Experimental Units--The creation of homogeneous experimental units
for this paired comparison example may be illustrated as follows: (a) start
with a single piece of bar stock to obtain homogeneous specimens, say 8 ft
long, (b) next cut the bar into 2N blanks, say 16, and number each succeed-
ing blank in order--1, 2, 3 . . . . . 16, (c) then each blank is rough machined
(to dinaensions larger than the final specimen dimensions by a suitable
allowance)--taking care even in this preliminary processing to avoid
potentially troublesome problems such as having groups of specimens
prepared by different machinists, and finally, (el), perfmm the final finishing
process as uniformly as practical, whatever that final processing may be.
This procedure creates 16 apparently homogeneous experimental units.
Treatments--The second step in setting up the paired comparison test
program is to assign randomly the required treatments to the experimental
units. In the program of Fig. 1, Treatments A and B are shot peening and
mechanical polishing, respectively. In more general context, the term
treatment is used to refer to something done to an experimental unit that
may affect the response of the unit in subsequent testing. For example, a
set of treatments may simply be two or more stress amplitudes (sinusoidal
or root mean square) used in S-N testing. But, treatments may also be
applied in combination, that is, two or more materials, or material process-
ings, each tested at two or more load and environmental test conditions
as implied in the brief definitions given in the Introduction to this chapter.

ORGANIZATIONAL STRUCTURE
OF A PAIRED COMPARISON
TEST PROGRAM

2N specimens , tested tn pairs, 5. Pairs arbitrarIly seIected from

one specimen In each pair sub- homogeneous specimens. (a)
jected to Treatment A (e.g., 4. AH tests conducted on o
shot peening) , the other to single test m a c h i n e . (~
Treatment B (e.g. , mechanical 5 . All tests conducted under
polishing). the some nominal l o a d - time
Time order of testing within history and test environment.
each pair chosen 01 random.
RESULTING DATA
(Rearranged)
[,] y = Fatigue Life
Pa r
[-~-~t(i) ~ (z) 5 YA,I(I ) YB,I(2 )

YA,2(I) YB ' 2 (2)

I~l I21 I'l

P0,r~ ~
E'I 121 /'l YA,5(2 ) YB,3(I )

I,.:.] ~i:,l ~i=21 N YA,4(i) YB,4 (Z)

N
~~ ] ~"--1 (2) (,, 2 YA,N (I) YB,N (2)

(a) Only in this elementary example . Refer later to

F i g u r e 3.

FIG. 1--Diagram of the organizational structure of a simple paired comparison test pro-
gram. Test specimens are indicated schematically by small rectangles, treatments denoted by
capital letters A and B, the random time order for testing within pairs by subscript (j), and
the time order for testhlg pairs by [k].

The primary feature of a treatment as opposed to a nuisance variable

(such as having different machinists prepare different specimens), is that a
treatment connotes an effect to be measured (as evident in the test objective),
whereas a nuisance variable is any other test variable whose effect on the
fatigue behavior is not presently of direct interest. Thus the effects of
nuisance variables should be circumvented by appropriate test planning

(namely, by structuring the blocks such that treatment effect estimates may
be isolated from nuisance effects associated with block-to-block variability).
Blocks--The "blocks" in the paired comparison example are literally
time blocks, namely, all laboratory conditions and test machine performance
are expected to be more uniform for two tests back to back in time order
than for two tests randomly or arbitrarily selected from the total 2N tests.
Accordingly, the experimental units (specimens) are grouped in blocks of
size two (pairs), the number of treatments being compared.
Since in most test situations it is a simple matter to list several miscel-
laneous variables that potentially pose nuisance effects, the first engineering
problem faced in experiment planning is to condense this list to a tractable
size, nevertheless taking care to include all truly important variables.
Generally the experience of the investigator and the information available
in the literature suffice to identify the critical sources of potential bias.
In other cases, however, this information must be gained by studying
experimentally the effect of a given potential nuisance variable, that is,
by regarding that potential nuisance variable as a "treatment" in a planned
experiment. When the purpose of a test program is to study basic varia-
bility and potential nuisance variables, the program is termed a uniformity
trial. (This program is illustrated next, following a brief paragraph out-
lining the analysis of paired comparison programs.)
Analysis--Data from this paired comparison test program may be
analyzed by standard statistical techniques to determine, in engineering
interpretation, whether significant differences in fatigue behavior can be
expected which are attributable to surface preparation. The Student+s t
statistic is applied customarily to paired comparison data [1 ], but signed
rank tests are applica+ble when the underlying life distribution is unknown
[51.~
Elementary Uniformity Trial Test Program
An elementary uniformity trial program provides the simplest organiza-
tional structure of all planned experiments. Only two structural elements
are involved: experimental units and treatments, namely, the potential
nuisance variable(s) being studied. In the uniformity trial example which
follows, the experimental units are individual fatigue specimens prepared
as described previously for the paired comparison program, and the
potential nuisance variable being studied is simply two different but nomi-
nally identical test machines. Such a study might be considered a first phase
of a subsequent test program, wherein, for timeliness, the use of two (or
more) test machines would be necessary.
4 One advantage of the paired comparison program is that the number of test blocks used
is flexible. The analysts may proceed as data accumulate, and the program can be extended
if unusual variability is encountered.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
10 MANUAL ON STATISTICAL P L A N N I N G A N D ANALYSIS

Organizational Structure, Experimental Units, and Treatments--Figure

2a summarizes the preparation of the experimental units, emphasizing the
point that for simplicity it is assumed all specimens are cut from a single
piece of bar stock, if two or more bars of stock were used, and variability
among bars is also regarded as a potential nuisance variable, then an
organizational structure with certain groupings of adjacent specimen
blanks (material blocks) would be required.
To establish the organizational structure for this elementary uniformity
trial, the 16 specimens (experimental units) shown in Fig. 2a are allocated
randomly, 8 to each test machine. In turn, each set of 8 specimens are
ordered randomly for testing. Then, during the test program, for this
simple example, all specimens are tested at the same stress level, test
frequency, and in the same laboratory environment; both sets of specimens
being tested simultaneously in their respective machines. Figure 2b diagrams
this organizational structure and summarizes in statements 1, 2, and 3 the
test conduct for this elementary uniformity trial.

Analysis--The test results from each set of fatigue life (Y) data provides
an independent estimate of the standard deviation of the population associa-
ted with the bar stock from which the single bar used in this test was
selected randomly. If it is assumed the logarithms of the respective fatigue
lives are distributed normally, the pooled estimate of the population
standard deviation provides sufficient information, based on the null
hypothesis that no difference between test machines actually exists, to test
whether the observed differences in average (log) lives can be expected
to occur more often than say once every 20 (or 100) times in repeated sam-
piing. The numerical details for this statistical analysis are illustrated in
Ref I, 6, 7, 8, and 9. If no assumption regarding the form of the life distribu-

(a) Single Piece of Bar Stock - 8' long.

t I

(b) Cut Bar into Blanks, and Number Successively (as s h o w n ) .

I I I 1 I I I I I I I l
I 2 3 ...... 14 15 16

(c) Rough Machine each Blank as uniformly as practical.

(d) Perform Material Processing (Heat Treatment) and Final

Surface Finishing as uniformly as practical for oil
specimens.

FIG. 2a--The preparation o f experimental units far the mtiJormity trial example (also
valid for the paired comparison program o f Fig. 1). This diagram or its equivalent, however
elementary, should be regarded as a necessary adjunct to the organizational structttre diagrams,
Figs. 1 and 2b.

ORGANIZATIONAL STRUCTURE
OF A UNIFORMITY TRIAL
TEST PROGRAM

I. Randomly s e l e c t e i g h t speci- 3. All tests conducted under

mens to be t e s t e d on each the same nominal load-time
machine. history and test environment
?. Test each of these two sets i.e. , conduct both sets of
of eight specimens in random tests concurrently.
time order. [ J
(J)

TEST TEST RESULTING DATA

MACHINE ONE MACHINETWO (Rearranged)
Y = Fatigue Life
Machi ne Machine
Spec. One Spec. Two
(2) --~ (7) Ci) (i}
2 YI (i) 12 YZ
-~-----I (6) (I)
(5)
[ II Y
i(z)
7 Y2
(Z)
(I) ~ ( 2 )
io y 13
I(3) Y2(3 )
(85 ~ ( 4 )
3 y 4 Y2
1(4 5 (4)
(65 (5)
9 YI (5) 5 Y2 (S)
(7) (B)
6 YI(6) 14 Y2(6 )
(4)
---'~ (3) 8 YI ( 7 ) I Y2 (7)
(~)
15 YI 16 Y2
(S) (S5

Specimen J T e s t J 4 ime Order

Number Machine

FIG. 2b--Diagram of organizational structure of the example uniformity trial discussed

in text.

tion is made, the appropriate nonparametric (distribution-free) analysis is

found in Chapter 4.
The computed test statistic for the data analysis associated with this
uniformity trial example will be such that: (a) its magnitude is so small
that one cannot reasonably reject the null hypothesis that no machine
differences actually exist, ~ (b) its magnitude is so large that there is no

5 Said in engineering terms, the text statistic magnitude is so small that one does not
hesitate to accept the equality of fatigue responses for the two machines, that is, the ma-
chines are assumed identical for practical purposes.

hesitation to reject the null hypothesis, that is, that one must conclude that
the two test machines really do give (markedly) different fatigue life re-
sponses, or (c) its magnitude is borderline (and perhaps a larger sample
size is required to make a firm decision). In situation (a) the conclusion
from this simple uniformity trial is that both machines may be used in a
larger subsequent program without undue concern about machines acting
as a nuisance variable. However, for the latter two situations, (b) and (c),
further consideration must be given to machines as a (possible) nuisance
variable, as described later.

Blocks--Suppose it is concluded from the analysis of the uniformity

trial that the two nominally identical test machines (may) indeed yield
fatigue life responses which are different. In this situation it is mandatory
subsequently either to conduct all critical comparative tests on a single
test machine or to account analytically for the apparent difference between
test machines. Moreover, even if the results of the uniformity trial are not
so clear cut, it still is desirable to consider possible machine differences in
planning future tests.
The experimental design methodology used to account for both possible
and proven nuisance variables is the deliberate grouping of experimental
units into blocks. The purpose of this planned grouping to is segregate
experimental units or test conditions or both into subgroups (blocks) that
either are, or logically should be, more homogeneous than the aggregate--
so that the treatment effects being studied can be isolated from the differ-
ences (biases) associated with nuisance variables. In word equation format,
it is clear that
observed effect = actual effect q- nuisance effects
The experimental structure desired is such that, by appropriate formation
of blocks (planned groups), the observed effect can be statistically parti-
tioned such that the estimates of the actual effects are independent of the
effects of the nuisance variables.
To illustrate by example the organizational structures where blocking
(planned grouping) is used to account for possible or proven nuisance
variable, it is again assumed that the technical question for the paired
comparison example (see Fig. 1) has been raised, namely, "How does
surface finishing Procedure A affect the fatigue behavior relative to Fro-
cedure B?"
In this case two bars of Material M are utilized (each as in Fig. 2a) to
prepare 32 specimens (experimental units). In the first part of this example
all specimens will be tested in a single machine. But, since two bars of stock
are involved, the possibility exists that slight differences in fatigue behavior
could occur from bar to bar that could influence the comparison between
the two surface finishes. Therefore, by treating bars as a possible nuisance

variable, more homogeneous test results would be logically obtained

(less variability among experimental units within each block) by blocking
(planned grouping) with regard to both bars and time order of testing.
Material blocking (grouping) is accomplished by making sure that both
specimens, comprising each pair in Fig. 1, are taken from the same bar and
preferably from adjacent blanks. Time blocking is accomplished by testing
both specimens in the given pair (a block of size two) before starting the
tests on another pair (block).
In the second part of this example, two test machines will be utilized, but
it is not necessary that half the 32 specimens be tested on each machine.
Rather, if test machines are a possible or proven nuisance variable, the
proper grouping constraint is that all specimens in each block be tested on
the same machine, either machine one or machine two. In this example
there are three nuisance variables: material bars, test machines, and time
(that is, machine behavior change or laboratory environment change with
time). The planned grouping (blocking) relevant here is to superimpose
(aggregate) the nui~sance variables in blocks such that for each block (pair),
both specimens are chosen from adjacent blanks in the same bar, both
specimens are tested on the same test machine, either machine one or
machine two, back to back in random time order. The individual blocks,
of course, are tested in random time order themselves as indicated in Fig. 1.
In the third part of this example various other nuisance variables are
considered. Suppose, to expedite testing, two machinists are used to pre-
pare the specimens, two heat treating furnaces are used in processing, etc.
Each of these additional nuisance variables may be superimposed in blocks
in whatever manner is convenient, as long as the specimens are paired
such that both specimens in each block are, or logically should be, more
homogeneous than if the blocking (grouping) had not taken place. There
is no mathematical limit on the number of nuisance variables that may be
conceivably superimposed in a given block, and individual blocks need
have little in common with one another.
Figure 3 summarizes the foregoing three part example. Treatments A
and B are assigned randomly to the individual specimens (experimental
units) in each pair (block). The individual blocks are prepared such that
all nuisance variables superimposed in a block are nominally identical for
all specimens in the block, as indicated in the columns under the heading
"Superimposed Nuisance Variables." Then, the blocks are tested in
random time order, with individual specimens within blocks tested back
to back in random time order.
As indicated in the three part example, it is not necessary to know that
a possible nuisance variable does indeed have a pronounced spurious effect
on the resulting data to consider this variable when forming blocks, but
merely if a reasonable suspicion exists, then blocking (planned grouping)

16 BLOCKS Random
Block Randomly Test Superimposed
Number Specimen Number Assign Order In Nuisance
Treat merits Blocks Variables

~ A B 2 I c c
'I2]__
2[.]__ ~ A B I 2 d d

3[,.] ~ ~ B A 2 c c

8[,2] ~ ~ A B I 2 d d

9[,] ~ l ~ B A , 2 e e

"b] ~ ~ B A , 2 e e

15 [14]' ~ _ _ [--'7] A B 2 I e e

,. [3] ~ ~ A . , 2 e o

FIG. 3--Final structure o f paired comparison program originally described in Fig. 1.

Experimental units prepared as indicated in Fig. 2a, using two bars o f stock. Time order o f
test for blocks indicated by [k]. The superimposed or aggregated (additional) nuisance variables
denoted by c may be, for example, material bar two, test machine one, technician two,
furnace one, machinist three, etc. There is no need for these additional nuisance variables to
be alike among blocks, nor, for that matter, to be different. The four blocks with super-
imposed blocking variable denoted e could each be selected from the same bar o f material,
tested on the same machine by the same technician, etc.

may be used as a precautionary device. The term "insurance" is often used

in advocating precautionary blocking.
The statistical motivation for blocking (planned grouping) with regard to
known and possible nuisance variables is quite straightforward, and is
apparent in the "Superimposed Nuisance Variables" columns of Fig. 3.
Observe that if a given collection of superimposed nuisance variables, say
d, is deleterious to the fatigue response in general, then the relevant test
outcomes for both Treatments A and B are reduced by (nominally) the
same amount; whereas, if the given collection of superimposed nuisance
variables has the opposite effect, then the relevant test outcomes for both
Treatments A and B are increased by (nominally) the same amount.
Either way, the differences in response between Treatments A and B are
isolated from the block effects. In symbolic form (compare the previous
word equation)
A - B = (A - a) - (B - a)
or
A - B = (A + a) - (B + a)

in which A and B are the expected responses for the respective treatments
and d is the effect on the response of a given collection of superimposed
nuisance variables.
The symbolic form of the blocking (planned grouping) rationalization
is exact provided the actual relationship between the nuisance variables
and the imposed treatments is simply additive. However, even if the actual
relationship is quite complex and nonlinear, the concept of planned
grouping is still valid--and the symbolic approximation in this case should
still be reasonably accurate when d is small compared to A and B. The
only truly harmful situation arises when the magnitude of ddepends mark-
edly on whether A or B is applied, namely, when the magnitude of the effect
of the nuisance variables differs markedly when A is applied from when B
is applied. When d depends on A or B, the nuisance variables are said to
interact with the treatments. In proper planned grouping it is always
assumed that nuisance variables do not interact with treatments. If, there-
fore, an interaction is even suspected, the suspected variable is not a per-
missable blocking variable, that is, it should not be superimposed with
other nuisance variables in forming blocks.

Recommendation--In fatigue testing it generally seems worthwhile to

form precautionary blocks with regard to both time and test machines.
The former blocking minimizes problems in data analysis should a machine
failure occur, the latter circumvents (isolates) a major long-range source of
variability commonly neglected by fatigue investigators, namely, machine
differences.

Confounding--Consider the alternative program structure where the

nuisance variables differ from specimen to specimen in one or more blocks.
Then, for such a block with differences in nuisances variables, the difference
in response for Treatments A and B is, say

A - B = (A - f ) - (B- g) = (A - B ) - (f- g)
or

A - B = (A % - f ) - - (B %-g) = (A - B) %- ( f - - g)

In either case, it should be noted the statistical expectation of the estimated

difference is unequal to the actual (true) difference (A -- B). The estimated
difference is thus said to be biased.
The engineering significance of bias is quite apparent when it is inter-
preted as disrupting a fair comparison. The only engineering argument in
defense of confounded nuisance variables might be that nothing was delib-
erately done to favor one or the other treatments, that is, the bias, if any,
was allegedly " r a n d o m . " This argument is unacceptable. First, if the argu-
ment were accepted on face value, the effect of confounding would be to

dilute the statistical power of the comparison, it would be an inefficient

test. Second, there is no evidence to support the claim of being " r a n d o m . "
Mechanical Randomization--Now consider the purpose of randomization
procedures in assigning treatments to experimental units, of selecting the
time order in which the individual specimens are tested, etc. Mechanical
randomization is required to assure that subsequent statistical analyses
are valid, that is, that the subsequent statistical inference made is truly
based on a random sample.
Statistical theory always reads: "Given a random sample, Yx, Y2, Y~. . . . .
YN, and . . . . " - - but, of course, in practice one is never "given" a random
sample. Rather, purposeful effort must be expended to create a random
sample by employing the process of mechanical randomization. If this
process is not used in constructing the organizational structure of the
planned experiment, then literally there can be no valid statistical analyses
or inferences from the results. Thus, mechanical randomization (illustrated
later) is a necessary and vital step in all fatigue test planning and conduct.
It cannot be overlooked or compensated for at a later time. Mechanical
randomization is as deliberate and premeditated a process as blocking on
possible nuisance variables.
Mechanical randomization is necessary to obtain statistically unbiased
estimates of the (actual) treatment differences and their respective variances.
Subsequent statistical analysis of variance (ANOVA) is based on the under-
standing that, under the null hypothesis of no treatment differences, the
statistical expectations of the estimated variances are known. These
expectations cannot be considered known quantities unless mechanical
randomization of treatments to experimental units has taken place in
experimental planning.
The planned experiment organizational structure used here to illustrate
the details of mechanical randomization is termed the completely random-
ized design (CRD). The C R D is similar to the uniformity trial program,
Fig. 2, in organizational structure, except that in general more than two
treatments are compared simultaneously relative to the null hypothesis of
equality of treatment effect on the fatigue behavior for all treatments. +

Completely Randomized Design ( CRD) Test Program

Given 16 experimental units prepared as indicated in Fig. 2a and a
single test machine. Suppose four specimens are to be tested at each o f

+ The corresponding statistical analyses differ, however, because the simple t test is valid
for the CRD only if the two treatments being compared are selected before the test is
conducted (or randomly selected after the test program has been conducted), namely, the
Student's t statistic is not valid for testing the null hypothesis associated with the largest
estimated difference. Either the F statistic can be used to test the null hypothesis with
regard to all treatments, or Tukey's t can be used to compare each (all) estimated dif-
ference(s) (T, -- T+), ll]. Generally, Tukey's t is used to determine which treatments differ
after the F statistic has been used to establish differences among (at least) some of the
treatments.

four stress amplitudes to generate an S - N curve. The stress amplitudes are

the treatments, no blocking variables are considered yet, and the individual
specimens are the experimental units.
R a n d o m i z a t i o n - - T h e first step in developing the organizational structure
of the completely randomized design program is to assign treatments ran-
domly to experimental units. Suppose the "stream" of random numbers
developed for this example (generated as outIined in detail in the appendix)
are the sequence: 1 5 - 1 - 4 - 6 - 1 4 - 9 - 1 2 - 2 - 1 1 - 1 6 - 8 - 1 3 - 3 - 7 - 1 0 - 5 . Thus stress
amplitude one was assigned to specimens 15, 1, 4, and 6; stress amplitude
two to specimens 14, 9, 12, and 2; etc. The results of this assignment are
summarized in Table la. Next the order of testing is established using a
second permutation of random numbers, say: 11-5-9-6-2-15 13-1-8-12-
14-16-3-4-10-7. The overall results of both randomizations are now sum-
marized in Table lb. Therefore, the tests proceed as follows: test 1, speci-
men 7, stress amplitude four; test 2, specimen 1, stress amplitude one; etc.
Many fatigue investigators still test all (or several) specimens at each
given stress amplitude before changing to tests at a different stress ampli-
tude. This test conduct method is not acceptable under proper interpreta-
tion of randomization requirements. In fact, it is important to reset the
stress amplitude even when two consecutive tests are assigned to the same
stress amplitude by mechanical randomization! When each of the four
specimens at a given stress amplitude are tested such that each stress
amplitude setting reflects an independent attempt to achieve the nominal
(target) value, then the four specimens are termed replicates. The CRD in
Table lb is replicated four times, that is, each of the four rows forms a
replicate of the "minimal" CRD.

TABLE 1--Example of the mechanical randomization process for a

completely randomized design (CRD) test program.
Stress Amplitudes (treatments)
(a) Random assignment of four specimens, Fig. 2a,
to each stress amplitude.

One Two Three Four

15 14 11 3
1 9 16 7
4 12 8 10
6 2 13 5

(b) Random assignment of time order of testing (i).

One Two Three Four

15(m 14(5~ 11 ~9~ 3(6~

4(8) 12(l~) 8(14) 10(16)

6(3) 2(4) 13(lo) 5(7)

The rationale for independently attempting to achieve the nominal

(target) value is that under true replication conditions one may expect the
average value of stress amplitude settings to agree closely with the nominal
value, that is, to expect a high value on one attempt to compensate for a
low value on another attempt. On the other hand, if only one setting is
made, it is either high or low, and no one knows which. Moreover, the
error associated with a single setting exceeds the error associated with the
average of several attempts.

Blocking--The completely randomized design has very limited applica-

tion in fatigue testing due to its lack of blocking on time, test machines,
material, etc.
Time blocking is patricularly important when large numbers of specimens
are to be compared with respect to several treatments. To illustrate this
point consider a recent experiment to determine whether different labora-
tories could prepare uniformly mild steel specimens with regard to the
effect of surface finish on fatigue life. Five laboratories each prepared ten
specimens, with all 50 specimens tested in random time order on a single
machine. Suppose the test machine broke down sometime in the middle of
the tests: how would the remaining tests be influenced by the machine
repair? Even worse, suppose that the machine broke down after several
specimens from one laboratory had been tested but before any specimens
from another had been tested: then what data could be salvaged?
Consider the alternative (revised) program with ten time blocks, each
containing five specimens, one from each of the five laboratories involved.
Then assuming a single test machine is still used in all 50 tests, the most
pessimistic view of the effect of a machine breakdown is simply to discard
the data for the five specimens in the time block during which the machine
failure occurred--and there still would be nine specimens properly tested
from each laboratory. The organizational structure of the revised test
program is such that two or more test machines can be easily superimposed
with the time blocks to speed the overall test. The latter blocking is accom-
plished as indicated in the "Superimposed Nuisance Variables" column of
Fig. 3. In fact, the resulting blocked test program is identical to the example
paired comparison program of Fig. 3--except that it now compares five
treatments (laboratories) rather than only two. This increase in the number
of treatments considered leads to the new terminology, randomized com-
plete block (RCB) design.

Randomized Complete Block (RCB) Test Program7

The CRD in Table I will now be restructured as a randomized complete
block (RCB) design. Since there are four treatments (stress amplitudes) and

7The statistical analysis for the RCB is illustrated in Chapter 2.

16 e x p e r i m e n t a l units (specimens) within each bar, each b a r m a y be sub-

divided to f o r m four material blocks. Accordingly adjacent specimens 1,
2, 3, and 4 f o r m material b l o c k 1 (because these four specimens are expected
to be m o r e h o m o g e n e o u s than four specimens chosen arbitrarily or at
r a n d o m f r o m the whole bar); adjacent specimens 5, 6, 7, and 8 f o r m
m a t e r i a l b l o c k 2; etc.

R a n d o m i z a t i o n - - S t r e s s a m p l i t u d e s (treatments) are assigned r a n d o m l y

to the four specimens (experimental units) in each block as illustrated in
T a b l e 2a. Then, as indicated in Table 2b, the time order of testing within
blocks is assigned r a n d o m l y . Finally, the order of testing the blocks is
then established by mechanical r a n d o m i z a t i o n .
E a c h b l o c k is of size four and forms one complete replicate of the
t r e a t m e n t s (the four stress amplitudes used in testing)"regardless of the
n u m b e r o f additional nuisance variables ultimately s u p e r i m p o s e d in various
blocks. Thus there is really nothing to gain in this p r o g r a m by using a single

TABLE 2--Example of mechanical randomization process for a randomized

complete block (RCB) test program.
Stress Amplitudes (treatments)
(a) Random assignment of specimens in each block to stress amplitudes.

One Two Three Four

Block 1 3 2 1 4
Block 2 8 6 5 7
Block 3 9 12 11 10
Block 4 14 15 16 13

(b) Random assignment of time order of test within each block (i).

One Two Three Four

Block l 3(4) 2(t) 1(2) 4(3)

Block 2 8(1) 6(a) 5(2) 7(4)
Block 3 9(4) 12(3) 11(1) 10(2)
Block 4 14(3) 15(z) 16o) 13(4)

(c) Random assignment of time order of testing of blocks [kl. (This block
order randomization is not absolutely necessary but nevertheless reflects
good practice.)

One Two Three Four

Block 1 [tl 3(4) 2(1) 1(2) 4(3)

Block 2[4] 8(1) 6(~) 5(2) 7(4)
Block 3 m 9(4) 12(~) 11(1) 10(2)
Block 4131 14(~) 15(2) 16(1) 13(4)

test machine. Rather, either two or four test machines should be used
whenever there is advantage to speeding up the overall tests.

Blocking--The distinction in randomization techniques for the test

programs summarized in Tables lb (CRD) and 2e (RCB) lies specifically in
the formation of blocks in the latter program, s In this context, the differ-
ences between various test programs result directly from the differences in
the planned groupings and the associated constraints placed on randomiza-
tion in the construction of the respective organizational structures. Thus
the RCB test program may be viewed as a special modification of the
basic CRD test program.
The literature on design of experiments contains numerous other planned
experiments which may also be viewed as special modifications of the basic
CRD test program. Each of these test programs has a distinctive organiza-
tional structure generated by the particular planned grouping deliberately
employed in its development. The influence of planned grouping on organi-
zational structure is apparent in Table 3 where three common variations
(extensions) of the basic CRD test program are displayed for comparative
purposes. The latter two planned experiments are termed the randomized
incomplete block (RIB) and the Latin Square test programs, respectively.
Programs of the sophistication of the Latin Square are said to have
structured blocking, namely, simultaneous planned grouping with regard
to more than one collection of superimposed nuisance variables. The analog
to structured blocking for nuisance variables is treatment combinations
in special arrangements for major variables. The next subsection describes
a special arrangement of wide applicability, namely, treatment combina-
tions in factorial arrangement.

Treatment Combinations in Factorial Arrangement--When two or more

treatments are assigned to the same experimental unit (specimen), it general-
ly is advantageous to apply the treatments in combinations of the type that
each treatment at each of its levels appears with all other treatments at
each of their respective levels. 9 This factorial arrangement is illustrated
easily for two treatments, each at two levels, in Table 4a.
Note that quantitative and qualitative treatments have been mixed deliber-
ately in Table 4a-- to emphasize the point that "levels" may simply indicate
the application of different types of qualitative variables. For example,
flame hardening, carburizing, nitriding, diffusion coatings, paint (and no

8 Good practice for test programs employing constrained randomization is first to estab-
lish deliberately the desired grouping, for example, each treatment appearing once and
only once in each block, after which all remaining assignments of test variables, time
orders, etc. are estabhshed randomly by randomization techniques.
9 Factorml arrangements permit simple assessment of the interactions (synergistic
effects) among treatments. Thus factorial arrangements are relevant whenever interactions
a r e physically meaningful.

TABLE 3--Three example test programs, chosen to illustrate the influence of

blocking (planned grouping) on the resulting organizational structure. Each test
program bzvolves the same four treatments, A, B, C, and D,for direct comparison.
(a) Randomized Complete Block Example
Four treatments ~ four Blocks, each of size four. Note that each treatment
appears once and only once in each block.
One Blocking Variable

Treatment

Block 1 A B C D
Block 2 A B C D
Block 3 A B C D
Block 4 A B C D

(b) Randomized Incomplete Block Example

Four treatments =:~ four Blocks, each of say size three. Note that although
each treatment cannot occur in every block, each treatment appears three
times, and exactly twice in the same block with each other treatment.
One Blocking Variable

Treatment

Block 1 A B C
Block 2 A B D
Block 3 A "C' D
Block 4 ... "B c D

(c) Latin Square Example (two blocking variables)

Four treatments =~ four blocks for each of the two blocking variables, all "8"
blocks of size 4, specifically structured to form 16 combinations. Note that
for each blocking variable each treatment appears once and only once in each
block.

Blocking Blocking Variable One

Variable
Two Block 1.1 Block 1.2 Block 1.3 Block 1.4

Block 2. i A B C D
Block 2.2 B A D C
Block 2.3 C D A B
Block 2.4 D C B A

paint), m e c h a n i c a l polishing, and e l e c t r o c h e m i c a l p o l i sh i n g are all valid

levels o f the q u a l i t a t i v e surface p r e p a r a t i o n variable. T h e r e are, o f course,
an infinite n u m b e r o f possible levels for the q u a n t i t a t i v e t e m p e r i n g t e m p e r a -
t u r e v a r i a b l e (but only a finite practical n u m b e r ) .
Special s y m b o l o g y a n d t a b u l a r f o r m a t s have e v o l v e d in the design o f
e x p e r i m e n t s l i t e r a t u r e to s u m m a r i z e the a s si g n m en t s o f t r e a t m e n t c o m -
b i n a t i o n s for N t r e a t m e n t s , each at M levels. T h e t w o m o s t c o m m o n sum-
m a r i e s o f the test p r o g r a m for three t r e a t m e n t s , each at t w o levels, are dis-

TABLE 4a--Treatment combinations applied in factorial arrangement: two treatments

(factors), two levels.

Treatment Combinations

Treatment One, Treatment Two,

Experimental Unit tempering temperature ~ surface preparation

One 300 and mechanical polish

Two 300 and shot peen
Three 700 and mechanical polish
Four 700 and shot peen

Summary Schematic

Treatment One, tempering temperature ~

Treatment Two,
surface preparation 300 700

Mechanical polish experimental unit one experimental unit three

Shot peen experimental unit two experimental unit four

TABLE 4b--Summary format for all treatment combhlations--

given three treatments, A, B, and C; each applied at two levels. Lower
case letters are used to denote the "'upper" level (an arbitrary choice
for qualitative variables), whereas lower case letters are omitted to
htdicate the "lower" level. The symbol 1 (one) is used to indicate that
all variables are simultaneously at their lower level. Note that there are
2~ treatment combhtations in this example.

Experimental Unita Treatment Combination

1 1
2 a
3 b
4 ab
5 c
6 ac
7 bc
8 abc

Treatment combinations are assigned at random to experimental

umts. The ordered numbers given in this column are intended
merely to simplify enumeration of the eight treatment combinations
in this example.

played in Tables 4b a n d e. The latter table is m u c h simpler to c o n s t r u c t

when the n u m b e r of t r e a t m e n t s b e c o m e s large ( a n d it is also c o n v e n i e n t
for c o n s t r u c t i n g fractional factorial designs): the first c o l u m n consists of
a l t e r n a t e m i n u s a n d plus signs, the second c o l u m n consists of a l t e r n a t e
sets of two m i n u s e s a n d two pluses, the third c o l u m n consists of a l t e r n a t e

T A B L E 4c--Summary format for all treatment combinations--given the same

three treatments, A, B, and C," each applied also at two levels. In this format a
plus sign btdieates the "upper" level, while a minus sign refers to the correspond-
ing "lower" level.

Treatment C o m b l n a t m n
Treatments

Experimental Unit- A B C

2 + - -
3 - + -
4 + + -
5 - - +
6 + - +
7 - + +
8 + + +

" Treatment combinations are assigned at r a n d o m to experimental units.

The ordered n u m b e r s given in this column are intended merely to simplify
enumeration of the eight treatment combinations in this example.

T A B L E 5a--CRD test program for two treatments (tempering temperature and surface
preparation) at two levels (300 and 700~ mtd mechanical pohshing and shot peening).
This test program employs a single test machine. The 16 experimental units prepared as
described in Fig. 2a provide four complete replicates o f the four treatment combinations.
See text for discussion o f heat treating in batches.

Treatment A,
Experimental tempering Treatment B, Treatment
Unit i ~ temperature ~ surface preparation Combination Notation

l(la) 300 shot peen b - +

2(0) 300 mechanical polish 1
3(i4) 700 mechamcal polish a + -
4{;~ 300 mechanical pohsh 1
5(a) 700 mechanical polish a + -
6(1) 700 shot peen ab + q-
7(8) 300 shot peen b - -+-
8(6) 700 mechanical polish a + -
9(2) 300 mechanical pohsh 1
10o6) 700 shot peen ab + +
11(12) 300 shot peen b - - +
12oo) 700 mechamcal polish a + -
13(s) 300 mechamcal polish 1
14(4) 700 shot peen ab + +
15(~t) 700 shot peen ab + +
16(1~) 300 shot peen b -- +

i = 1 .... , 16.
(j) = test o r d e r ; j = I . . . . . 16.

TABLE 5b---RCB test program for two treatments at two levels. This test program may
employ one or more test machines and bars o f stock or both. Ttle 16 experimental units form
four blocks, each o f which forms a complete replicate o f the four treatment combinations.

Experimental Units Treatment Combination Notation

Block One(~)
1(3) 300~ shot peen b - +
2(4) 700~ mechanical polish a + --
3c~) 300~ mechanical polish l
4(1) 700~ shot peen ab + +
Block Two(2)
5(3) 700~ mechanical polish a + -
6/~) 300~ shot peen b -- +
7(2) 700~ shot peen ab q- +
8(4) 300~ mechanical polish 1
Block Three.~
9(3) 300~ shot peen b - +
10(2) 700~ mechanical polish a + -
11(,o 700~ shot peen ab + +
12(4) 300~ mechanical polish 1
Block Four(~)
13(4) 300~ mechanical polish 1
14(o 300~ shot peen b - +
15(2) 700~ shot peen ab + +
16(3~ 700~ mechanical polish a + --

sets of four minuses and four pluses, the fourth column (if required)
consists of alternate sets of eight minuses and eight pluses, etc.
The C R D and RCB test programs are constructed easily for the two
t r e a t m e n t / t w o levels situation summarized in Table 4a. These two pro-
grams appear in Tables 5a and b, respectively. Note that the four treatment
combinations of Table 4a merely replace the original four treatments
denoted One, Two, Three, F o u r in Tables 1 and 2.
All experimental designs presented thus far, including the C R D and RCB
test programs with four treatment combinations, encounter certain practical
difficulties when applied to fatigue test situations. The basic problem is
that m a n y processing steps in specimen preparation, for example, heat
treating, are not legitimate treatments in the context defined h e r e i n - -
because practical considerations such as time and cost lead to the treatments
being processed in batches, that is, the 16 specimens in Tables 5a and b
will be tempered in two batches, one batch of eight specimens tempered
at 300~ another batch of eight at 700~ whereas the fundamental defini-
tions of treatment and replication require that there be 16 separate heat
treatments, eight independent tempering processes at 300~ and another
eight independent tempering processes at 700~ Sixteen separate heat

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
(o) I B,ock O.e I Block Two I B,ock Three I .,ock .our ]

Typical Biock
(b) IMain Plot JMain Plot I
~--- Exp er i me n t al Unit for the Primary Treatment.
Typical Main Plot
I
(c)
I , 1
-r
"9'~ /oi'S"~//)'~olZ:, Experimento Units for the Secondary Treatment >=

o
Z
Eight Main Plots , each with Two Split Ptots
Z
(d) I I i' z I 3 6' 4 l I 5 ,I 7 l' 8 I 9 I', l~ I 'l i' 12 I 13 i' 14 I 15 i' 16
I 2 3 4 5 6 7 8
Z
SIXTEEN SPECIMENS
r-
r3
FIG. 4--Subdivision of bar stock for split plot test program example. o
Z

1 I I
300 i 300 700 I 700 Block One M.P. S,P. I S.P. II MP Block One
'1
(4)
I'l t
(3) (4) (1) (2)
I
(3)

I
300 I 300
I
(I)
I 700
I
I 700
I
(z)
)
Block Two
t S,P.

(e)
M.P. I S.P.
(I) (7)
M.P.
(z)
IBlockTwo

700
(e)
700 J I
300 II 3 0 0 I BIock Three
I
(7)
i sP.

(5)
MP
(e)
IMP (7)
S.P.

(6)
Block Three

I
300
I
I 500 J 700
I
lI 700 Block Four I S P. MP I SP
I I
M P [Block Four
I I I I
(6) (5) (3i (S) (4) (5)
(a) (b)
I 2 Main Plot Order
I
I' I
13
I 2
I
14
Z
16
I

15
J Block One(4)
Fohgue Test Order
2
I
I J Block Two(2)
I
8 7 5 Fatigue Test Order

2
2
I I
:5
I
I
I
2
4
Block Threed~1~
Fatigue Test Order
2

I
9
2
I0
1'1 2

12
I
l
I
I

II
Block Four~3J~
~

Fatigue Test Order

(c)

(a) Random assignment of main plot treatments within each block, then time order of
tempering (j).
(b) Random assignment of split plot treatments within each main plot, then time order
of each surface preparation (k).
(c) Random assignment of test order of blocks, main plots within blocks, and split plots
within main plots gives fatigue test order shown.
FIG. 5--Randomization for split plot example.

treatments are, of course, quite impractical! Accordingly, test programs

which circumvent potential analytical difficulties associated with processing
specimens in batches should be employed whenever possible. Such programs
generally have nested organizational structures, for example, as in the
elementary split plot test program described next.

Elementary Split Plot Test Program

Consider now the problem of modifying the organizational structures
of the CRD and RCB test programs in Tables 5a and b to circumvent
statistical difficulties caused by processing (applying) treatments in batches.
In this case it is necessary to go back to the original preparation of the
experimental units to restructure the experiment.
Organizational Structure--Consider again the preparation of the 16
experimental units per bar of stock, Fig. 2a. The nested program is restruc-
tured here from that beginning.
First, divide each bar into eight pieces, each 1 ft long, each termed main
plots; then subdivide each main plot into two blanks, each termed split
plots. The intent here is to apply the primary treatments to the main plots
and the secondary treatments to the split plots. In this example tempering
is the primary treatment because heat treatment precedes surface prepara-
tion. To give the example some additional practical structure, also form
four material blocks, Fig. 4a, before forming the main plots and split
plots, Figs. 4b and 4c.
Now assign the two tempering temperatures, 300 and 700~ to the two
main plots within each block, Fig. 5a. In this example these tempering
processes require eight independent runs, four at each temperature. The
time order of these eight tempering runs is indicated by (j) below each
main plot.
Surface preparations are assigned as illustrated in Fig. 5b, and all tests
are then conducted in the randomized time order indicated in Fig. 5c.
The critical distinction between this split plot design and the RCB
design of Table 5b is the hierarchy involved, namely, in the RCB the four
treatment combinations are assigned at random within each block (with no
specific attention given to which processing treatment is applied first);
whereas in the split plot design the primary treatments are assigned to the
main plots before the secondary treatments are assigned to the split plots.
This critical distinction leads to markedly different statistical analyses! TM
Replication--The split plot and RCB designs differ in their replication of
treatments. Recall the RCB example test program required 16 independent
furnace runs for tempering. This split plot example requires only eight,
Fig. 5a. But both test programs require 16 independent surface prepara-
J0 For statistical analysis o f elementary split plot programs, refer to Snedecor and Cochran
16] or Little and Jebe [9].

tions, eight each for both shot peening and mechanical polishing, Fig. 5b.
The decrease in replication is, of course, the direct consequence of the
decrease in experimental units for tempering from 16 in the RCB to the
eight main plots in Fig. 4d. Each of the eight split plot pairs displayed in
this diagram are subunits with regard to tempering, the main plot treat-
ment. These 16 split plots are experimental units, however, with regard to
surface preparation, the split plot treatment.
Now for purposes of comparison, consider the split plot program where
only two tempering runs are conducted, one at 300~ and the other at 700~
Each of these tempering runs forms one main plot with eight split plots
within it. Thus there is only one replicate of each tempering temperature
treatment, whereas there are eight replicates of each surface preparation
treatment.
The split plot program with only one replicate of the tempering tempera-
ture treatments is perhaps as inappropriate from a statistical point of view
as the RCB with eight replicates of the tempering temperature treatments
is from a materials processing point of view. The former program is
statistically "impotent," the latter program is generally unacceptable
with regard to cost and time.
The obvious solution of course is to compromise between these unaccept-
able extremes, namely, to balance both main plot and split plot replication
such that the test program is both reasonably simple and inexpensive, and
statistically sound and effective, q'hese dual requirements may be satisfied
in general by enumerating and comparing several different split-plot test
programs, each time changing the numbers of split plots per main plot, the
numbers of main plots per bar of stock, the amount of bar stock, etc.,
until some suitable compromise is apparent.

Summary and Perspective

The key issue in planning fatigue experiments is whether or not the
proposed experimental design most effectively accomplishes its goal, not
whether the given design is well known and easily identified. Accordingly
the CRD and RCB test programs do not in any universal way connote
well-planned experiments. In fact, these programs may reflect stereotyped
thinking in which the test situation is forced to conform with the experi-
mental design, rather than vice versa.
Well-planned experiments occur when the organizational structure of
the test program is directly compatible with (1) test objectives, (2) number
and nature of the blocking variables, (3) nature and hierarchy of the
processing and preparation of specimens, and (4) various practical and
economical constraints imposed on different aspects of the test program.
The simplest approach to satisfying these compatibility requirements is to
let the steps required in processing and preparing the test specimens estab-
lish the skeleton of the test program. Addition of the required comparisons

and the relevant blocking variables serve to define the organizational

structure of the test program. Lastly, mechanical randomization establishes
the remaining details of the test program conduct.
I f the test p r o g r a m being considered cannot be summarized by means of
a reasonably simple diagram, chances are the test will not effectively
achieve its stated objective.
Well-planned experiments have one additional feature. The method of
analysis is generally quite apparent in examining the organizational struc-
ture. That is, if one truly understands the organizational structure of the
test program, the corresponding statistical analysis should be relatively
straightforward.
Generally speaking, competent statistical analysis involves analysis of
variance ( A N O V A ) when dealing with comparative test programs. The
detailed steps of a typical analysis, for a given example RCB test program,
are presented in Chapter 2.

APPENDIX
Mechanical Randomization and Random Number Tables
Table 6 lists 10 000 uniformly distributed random numbers (namely, each digit
0, 1, 2 . . . . ,9 having an equal probability of occurring at each entry).
For engineering purposes these 10 000 numbers may simply be viewed as a
rearrangement of a stream of 10 000 individual digits, one following the other in a
single column (vector). This stream starts with the entries: 5-1-8-6-0- . . . The
51st and subsequent entries are: 4 - 5 - 5 - 1 - 5 - . . . , and so forth.

Starting Point
The starting point for drawing numbers from the random stream is the next
number following the finish of the previous use of the table. (The first time the
table is used, the experimenter should use the last digit of his telephone number
(or similar device) to find a "random" starting point. For example, if this digit is 9,
the experimenter starts his stream in column 00 at row 10.)

Random Number Sequence

Suppose there are 16 treatments to be assigned to experimental units. The num-
bers of interest are clearly those two digit numbers starting with zero and those
two digit numbers starting with one whose second entry is six or less. Thus, the
procedure is to read down the random number stream in pairs of digits, for example,
starting at column 00 and row 10 as in the example above, read: 93, 06, 49, 29, 95,
71, 44, 86, 25, 79, 41, 05, 99, 48, 53, 74, 64, 89, 95, 80, 45, 51, 51, 87, 02, 64, 62, 78,
46,05(repeated) . . . . 0 8 . . . 0 4 . . . 0 1 . . . 1 5 . . . 1 4 . . . 0 7 . . . 1 I . . . 1 2 . . . 13...
1 6 . . . 0 9 . . . 10. The next starting point is column 09, row 28.
Suppose next that four treatments within a block are to be assigned random
time orders. Single digit numbers less than five (ignoring 0) are of interest. Thus,
proceeding from the last point used in the random number stream, the sequence is:
6, 8, 9, 3, 7, 2 , . . . 1 , . . . 4. The next starting point is column 10, row 13.
For a stream of 2500 uniformly distributed random numbers, 1 through 16 only,
see Ref 9. For random permutations of the integers 1 through 16, see Ref 29 and 30.

TABLE 6--Random number tables (Snedecor and Cochran [6])."

TEN THOUSAND RANDOMLY ASSORTED DIGITS

00-04 05--09 10--14 15-19 20-24 25-29 30-34 35-39 40-44 45-49

00 54463 22662 65905 70639 79365 67382 29085 69831 47058 08186
01 15389 85205 18850 39226 42249 90669 96325 23248 60933 26927
02 85941 40756 82414 02015 13858 78030 16269 65978 01385 15345
03 61149 69440 11286 88218 58925 03638 52862 62733 33451 77455
04 05219 81619 10651 67079 92511 59888 84502 72095 83463 75577
05 41417 98326 87719 92294 46614 50948 64886 20002 97365 30976
06 28357 94070 20652 35774 16249 75019 21145 05217 47286 76305
07 17783 00015 10806 83091 91530 36466 39981 62481 49177 75779
08 40950 84820 29881 85966 62800 70326 84740 62660 77379 90279
09 82995 64157 66164 41180 10089 41757 78258 96488 88629 37231
10 96754 17676 55659 44105 47361 34833 86679 23930 53249 27083
11 34357 88040 53364 71726 45690 66334 60332 22554 90600 71113
12 06318 37403 49927 57715 50423 67372 63116 48888 21505 80182
13 62111 52820 07243 79931 89292 84767 85693 73947 22278 11551
14 47534 09243 67879 00544 23410 12740 02540 54440 32949 13491
15 98614 75993 84460 62846 59844 14922 48730 73443 48167 34770
16 24856 03648 44898 09351 98795 18644 39765 71058 90368 44104
17 96887 12479 80621 66223 86085 78285 02432 53342 42846 94771
18 90801 21472 42815 77408 37390 76766 52615 32141 30268 18106
19 55165 77312 83666 36028 28420 70219 81369 41943 47366 41067
20 75884 12952 84318 95108 72305 64620 91318 89872 45375 85436
21 16777 37116 58550 42958 21460 43910 01175 87894 81378 10620
22 46230 43877 80207 88877 89380 32992 91380 03164 98656 59337
23 42902 66892 46134 01432 94710 23474 20423 60137 60609 13119
24 81007 00333 39693 28039 10154 95425 39220 19774 31782 49037
25 68089 01122 51111 72373 06902 74373 96199 97017 41273 21546
26 20411 67081 89950 16944 93054 87687 96693 87236 77054 33848
27 58212 13160 06468 15718 82627 76999 05999 58680 96739 63700
28 70577 42866 24969 61210 76046 67699 42054 12696 93758 03283
29 94522 74358 71659 62038 79643 79169 44741 05437 39038 13163
30 42626 86819 85651 88678 17401 03252 99547 32404 17918 62880
31 16051 33763 57194 16752 54450 19031 58580 47629 54132 60631
32 08244 27647 33851 44705 94211 46716 11738 55784 95374 72655
33 59497 04392 09419 89964 51211 04894 72882 17805 21895 83864
34 97155 13428 40293 09985 58434 01412 69124 82171 59058 82859
35 98409 66162 95763 47420 20792 61527 20441 39435 11859 41567
36 45476 84882 65109 96597 25930 66790 65706 61203 53634 22557
37 89300 69700 50741 30329 11658 23166 05400 66669 48708 03887
38 50051 95137 91631 66315 91428 12275 24816 68091 71710 33258
39 31753 85178 31310 89642 98364 02306 24617 09609 83942 22716
40 79152 53829 77250 20190 56535 18760 69942 77448 33278 48805
41 44560 38750 83635 56540 64900 42912 13953 79149 18710 68618
42 68328 83378 63369 71381 39564 05615 42451 64559 97501 65747
43 46939 38689 58625 08342 30459 85863 20781 09284 26333 91777
44 83544 86141 15707 96256 23068 13782 08467 89469 93842 55349
45 91621 00881 04900 54224 46177 55309 17852 27491 89415 23466
46 91896 67126 04151 03795 59077 11848 12630 98375 52068 60142
47 55751 62515 21108 80830 02263 29303 37204 96926 30506 09808
48 85156 87689 95493 88842 00664 55017 55539 17771 69448 87530
49 07521 56898 12236 60277 39102 62315 12239 07105 11844 01117

Used by permission from Statistical Methods by George W. Snedecor and William G.

Cochran. 9 1967 by the Iowa State University Press.
Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authoriz
CHAPTER I ON FUNDAMENTAL CONCEPTS 31

TABLE 6--Random number tables (Snedecor and Cochran [6]) (continued).

TEN THOUSAND RANDOMLY ASSORTED DIGITS

50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 95-99

00 59391 58030 52098 82718 87024 82848 04190 96574 90464 29065
Ol 99567 76364 77204 04615 27062 96621 43918 01896 83991 51141
O2 10363 97518 51400 25670 98342 61891 27101 37855 06235 33316
03 86859 19558 64432 16706 99612 59798 32803 67708 15297 28612
04 11258 24591 36863 55368 31721 94355 34936 02566 80927 08188
05 95068 88628 35911 14530 33020 80428 39936 31855 34334 64865
06 54463 47237 73800 91017 36239 71824 83671 39892 60518 37092
07 16874 62677 57412 13215 31389 62233 80827 73917 82802 84420
08 92494 63157 76593 91316 03505 72389 96363 52887 01087 66091
09 15669 56689 35682 40844 53256 81872 35213 09840 34471 74441
10 99116 75486 84989 23476 52967 67104 39495 39100 17217 74073
11 15696 10703 65178 90637 63110 17622 53988 71087 84148 11670
12 97720 15369 51269 69620 03388 13699 33423 67453 43269 56720
13 11666 13841 71681 98000 35979 39719 81899 07449 47985 46967
14 71628 73130 78783 75691 41632 09847 61547 18707 85489 69944
15 40501 51089 99943 91843 41995 88931 73631 69361 05375 15417
16 22518 55576 98215 82068 10798 86211 36584 67466 69373 40054
17 75112 30485 62173 02132 14878 92879 22281 16783 86352 00077
18 80327 02671 98191 84342 90813 49268 95441 15496 20168 09271
19 60251 45548 02146 05597 48228 81366 34598 72856 66762 17002
20 57430 82270 10421 05540 43648 75888 66049 21511 47676 33444
21 73528 39559 34434 88596 54086 71693 43132 14414 79949 85193
22 25991 65959 70769 64721 86413 33475 42740 06175 82758 66248
23 78388 16638 09134 59880 63806 48472 39318 35434 24057 74739
24 12477 09965 96657 57994 59439 76330 24596 77515 09577 91871
25 83266 32883 42451 15579 38155 29793 40914 65990 16255 17777
26 76970 80876 10237 39515 79152 74798 39357 09054 73579 92359
27 37074 65198 44785 68624 98336 84481 97610 78735 46703 98265
28 83712 06514 30101 78295 54656 85417 43189 60048 72781 72606
29 20287 56862 69727 94443 64936 08366 27227 05158 50326 59566
30 74261 32592 86538 27041 65172 85532 07571 80609 39285 65340
31 64081 49863 08478 96001 18888 14810 70545 89755 59064 07210
32 05617 75818 47750 67814 29575 10526 66192 44464 27058 40467
33 26793 74951 95466 74307 13330 42664 85515 20632 05497 33625
34 65988 72850 48737 54719 52056 01596 03845 35067 03134 70322
35 27366 42271 44300 73399 21105 03280 73457 43093 05192 48657
36 56760 10909 98147 34736 33863 95256 12731 66598 50771 83665
37 72880 43338 93643 58904 59543 23943 11231 83268 65938 81581
38 77888 38100 03062 58103 47961 83841 25878 23746 55903 44115
39 28440 07819 21580 51459 47971 29882 13990 29226 23608 15873
40 63525 94441 77033 12147 51054 49955 58312 76923 96071 05813
41 47606 93410 16359 89033 89696 47231 64498 31776 05383 39902
42 52699 45030 96279 14709 52372 87832 02735 50803 72744 88208
43 16738 60159 07425 62369 07515 82721 37875 71153 21315 00132
44 59348 11695 45751 15865 74739 05572 32688 20271 65128 14551
45 12900 71775 29845 60774 94924 21810 38636 33717 67598 82521
46 75086 23537 49939 33595 13484 97588 28617 17979 70749 35234
47 99495 51434 29181 09993 38190 42553 68922 52125 91077 40197
48 26075 31671 45386 36583 93459 48599 52022 41330 60651 91321
49 13636 93596 23377 51133 95126 61496 42474 45141 46660 42338

TABLE 6--Random number tables (Snedecor and Coehran [6]) (continued).

TEN THOUSAND RANDOMLY ASSORTED DIGITS

00-04 05-09 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49

5O 64249 63664 39652 40646 97306 31741 07294 84149 46797 82487
51 26538 44249 04050 48174 65570 44072 40192 51153 11397 58212
52 05845 00512 78630 55328 18116 69296 91705 86224 29503 57071
53 74897 68373 67359 51014 33510 83048 17056 72506 82949 54600
54 20872 54570 35017 88132 25730 22626 86723 91691 13191 77212
55 31432 96156 89177 75541 81355 24480 77243 76690 42507 84362
56 66890 61505 01240 00660 05873 13568 76082 79172 57913 93448
57 48194 57790 79970 33106 86904 48119 52503 24130 72824 21627
58 11303 87118 81471 52936 08555 28420 49416 44448 04269 27029
59 54374 57325 16947 45356 78371 10563 97191 53798 12693 27928
60 64852 34421 61046 90849 13966 39810 42699 21753 76192 10508
61 16309 20384 09491 91588 97720 89846 30376 76970 23063 35894
62 42587 37065 24526 72602 57589 98131 37292 05967 26002 51945
63 40177 98590 97161 41682 84533 67588 62036 49967 01990 72308
64 82309 76128 93965 26743 24141 04838 40254 26065 07938 76236
65 79788 68243 59732 04257 27084 14743 17520 95401 55811 76099
66 40538 79000 89559 25026 42274 23489 34502 75508 06059 86682
67 64016 73598 18609 73150 62463 33102 45205 87440 96767 67042
68 49767 12691 17903 93871 99721 79109 09425 26904 07419 76013
69 76974 55108 29795 08404 82684 00497 51126 79935 57450 55671
70 23854 08480 85983 96025 50117 64610 99425 62291 86943 21541
71 68973 70551 25098 78033 98573 79848 31778 29555 61446 23037
72 36444 93600 65350 14971 25325 00427 52073 64280 18847 24768
73 03003 87800 07391 11594 21196 00781 32550 57158 58887 73041
74 17540 26188 36647 78386 04558 61463 57842 90382 77019 24210
75 38916 55809 47982 41968 69760 79422 80154 91486 19180 15100
76 64288 19843 69122 42502 48508 28820 59933 72998 99942 10515
77 86809 51564 38040 39418 49915 19000 58050 16899 79952 57849
78 99800 99566 14742 05028 30033 94889 53381 23656 75787 59223
79 92345 31890 95712 08279 91794 94068 49337 88674 35355 12267
80 90363 65162 32245 82279 79256 80834 06088 99462 56705 06118
81 64437 32242 48431 04835 39070 59702 31508 60935 22390 52246
82 91714 53662 28373 34333 55791 74758 51144 18827 10704 76803
83 20902 17646 31391 31459 33315 03444 55743 74701 58851 27427
84 12217 86007 70371 52281 14510 76094 96579 54853 78339 20839
85 45177 02863 42307 53571 22532 74921 17735 42201 80540 54721
86 28325 90814 08804 52746 47913 54577 47525 77705 95330 21866
87 29019 28776 56116 54791 64604 08815 46049 71186 34650 14994
88 84979 81353 56219 67062 26146 82567 33122 14124 46240 92973
89 50371 26347 48513 63915 11158 25563 91915 18431 92978 11591
90 53422 06825 69711 67950 64716 18003 49581 45378 99878 61130
91 67453 35651 89316 41620 32048 70225 47597 33137 31443 51445
92 07294 85353 74819 23445 68237 07202 99515 62282 53809 26685
93 79544 00302 45338 16015 66613 88968 14595 63836 77716 79596
94 64144 85442 82060 46471 24162 39500 87351 36637 42833 71875
95 90919 11883 58318 00042 52402 28210 34075 33272 00840 73268
96 06670 57353 86275 92276 77591 46924 60839 55437 03182 13191
97 36634 93976 52062 83678 41256 60948 18685 48992 19462 96062
98 75101 72891 85745 67106 26010 62107 60885 37503 55461 71213
99 05112 71222 72654 51583 05228 62056 57390 42746 39272 96659

TABLE 6--Random mtmber tables (Snedecor and Cochran [6]) (cont#tued).

TEN THOUSAND RANDOMLY ASSORTED DIGITS

50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 95-99

50 32847 31282 03345 89593 69214 70381 78285 20054 91018 16742
51 16916 00041 30236 55023 14253 76582 12092 86533 92426 37655
52 66176 34047 21005 27137 03191 48970 64625 22394 39622 79085
53 46299 13335 12180 16861 38043 59292 62675 63631 37020 78195
54 22847 47839 45385 23289 47526 54098 45683 55849 51575 64689
55 41851 54160 92320 69936 34803 92479 33399 71160 64777 83378
56 28444 59497 91586 95917 68553 28639 06455 34174 11130 91994
57 47520 62378 98855 83174 13088 16561 68559 26679 06238 51254
58 34978 63271 13142 82681 05271 08822 06490 44984 49307 62717
59 37404 80416 69035 92980 49486 74378 75610 74976 70056 15478
60 32400 65482 52099 53676 74648 94148 65095 69597 52771 71551
61 89262 86332 51718 70663 11623 29834 79820 73002 84886 03591
62 86866 09127 98021 03871 27789 58444 44832 36505 40672 30180
63 90814 14833 08759 74645 05046 94056 99094 65091 32663 73040
64 19192 82756 20553 58446 55376 88914 75096 26119 83898 43816
65 77585 52593 56612 95766 10019 29531 73064 20953 53523 58136
66 23747 16364 05096 03192 62386 45389 85332 18877 55710 96459
67 45989 96257 23850 26216 23309 21526 07425 50254 19455 29315
68 92970 94243 07316 41467 64837 52406 25225 51553 31220 14032
69 74346 59596 40088 98176 17896 86900 20249 77753 19099 48885
70 87646 41309 27636 45153 29988 94770 07255 70908 05340 99751
71 50099 71038 45146 06146 55211 99429 43169 66259 97786 59180
72 10127 46900 64984 75348 04115 33624 68774 60013 35515 62556
73 67995 81977 18984 64091 02785 27762 42529 97144 80407 64524
74 26304 80217 84934 82657 69291 35397 98714 35104 08187 48109
75 81994 41070 56642 64091 31229 02595 13513 45148 78722 30144
76 59537 34662 79631 89403 65212 09975 06118 86197 58208 16162
77 51228 10937 62396 81460 47331 91403 95007 06047 16846 64809
78 31089 37995 29577 07828 42272 54016 21950 86192 99046 84864
79 38207 97938 93459 75174 79460 55436 57206 87644 21296 43395
80 88666 31142 09474 89712 63153 62333 42212 06140 42594 43671
81 53365 56134 67582 92557 89520 33452 05134 70628 27612 33738
82 89807 74530 38004 90102 11693 90257 05500 79920 62700 43325
83 18682 81038 85662 90915 91631 22223 91588 80774 07716 12548
84 63571 32579 63942 25371 09234 94592 98475 76884 37635 33608
85 68927 56492 67799 95398 77642 43913 91853 08424 81450 76229
86 56401 63186 39389 88798 31356 89235 97036 32341 33292 73757
87 24333 95603 02359 72942 46287 95382 08452 62862 97869 71775
88 17025 84202 95199 62272 06366 16175 97577 99304 41587 03686
89 02804 08253 52133 20224 68034 50865 57868 22343 55111 03607
90 08298 03879 20995 19850 73090 13191 18963 82244 78479 99121
91 59883 01785 82403 96062 03785 03488 12970 64896 38336 30030
92 46982 06682 62864 91837 74021 89094 39952 64158 79614 78235
93 31121 47266 07661 02051 67599 24471 69843 83696 71402 76287
94 97867 56641 63416 17577 30161 87320 37752 73276 48969 41915
95 57364 86746 08415 14621 49430 22311 15836 72492 49372 44103
96 09559 26263 69511 28064 75999 44540 13337 10918 79846 54809
97 53873 55571 00608 42661 91332 63956 74087 59008 47493 99581
98 35531 19162 86406 05299 77511 24311 57257 22826 77555 05941
99 28229 88629 25695 94932 30721 16197 78742 34974 97528 45447

Chapter 2 Necessary Steps in Statistical

Analysis for Planned Experiments

Introduction
Proper statistical analysis of planned fatigue experiments involves
writing the mathematical model relevant to the given organizational
structure (for example, a split plot program), transforming the observed
data as required (for example, taking log life), performing the appropriate
statistical analysis pertinent to the stated test objective, and examining the
residual errors to ascertain as well as possible whether the transformed data
meet the assumptions underlying the mathematical model and analysis.
This overall process is mandatory for all fatigue test programs, even though
the individual steps may differ considerably in their detail and rigor depend-
ing on the situation.
For test programs whose primary objective is comparative in nature
(that is, to compare the fatigue behavior of two or more materials or com-
ponents), it is recommended that the statistical analysis methodology known
as analysis of variance (ANOVA) be used. The necessary steps in analysis
of variance are illustrated in this Chapter for the specific case of the paired
comparison (RCB) test program (see Fig. 3, Chapter 1). ANOVA for
other test programs may be found in Refs 1, 6-9.

Necessary Steps in Analysis of Variance (ANOVA) for a Paired Comparison

Test Program

Step 1: Write the Mathematical Model

The correct mathematical model depends on the organizational structure
of the planned experiment, which in turn has been dictated by the test
objective and the various time, cost, and physical constraints placed on the
test conduct, if the test program organizational structure consists of N
sets (blocks) of paired comparisons, Table 7, the appropriate mathematical
model is
Y,, = /a + r, + ~j + e,i
where
Y. = random observation associated with the i*h treatment in the j,h
block,
34
Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Copyright*1975
Downloaded/printed by by ASTMInternational www.astm.org
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 2 ON NECESSARYSTEPS 35

u = mean (average) o f all conceptually possible m e a s u r e m e n t s

(observations),
r, = exact incremental change in response associated with i (i = 1, 2),
that is, the true t r e a t m e n t effect,
~j = exact incremental change in response associated with block j
( j = 1, N), t h a t is, the true block effect, and
e,, = r a n d o m error associated with the i th treatment m e a s u r e d in the
j ~ block, t h a t is, the r a n d o m deviation of Y~ f r o m (~ + r~ + ~ ) .

The objective o f A N O V A is to estimate the m a g n i t u d e o f each t r e a t m e n t

effect r, relative to the (estimated) m a g n i t u d e o f the s t a n d a r d deviation o f
the r a n d o m error. Valid statistical analyses are relatively straightforward
when independent estimates of these quantities are available. If the esti-
m a t e d treatment effects are all small relative to the scatter of the data, little
statistical or engineering significance should be given to the " o b s e r v e d
t r e a t m e n t differences." O n the other hand, if any treatment effect estimate
is quite large c o m p a r e d to the scatter o f the data, it is unlikely the observed
t r e a t m e n t difference can be r e a s o n a b l y attributed to mere r a n d o m variabil-
ity o f data. In this case, the t r e a t m e n t effect is statistically significant, b u t
engineering significance depends on its " c o s t effectiveness."
The p u r p o s e of structure in a p l a n n e d experiment is to permit unbiased
estimation of each r,, as well as an independent estimate o f the s t a n d a r d
deviation o f the r a n d o m error. The a p p r o p r i a t e w o r k i n g model for estima-
tion in the paired c o m p a r i s o n test p r o g r a m is

[^ caret denotes estimate]

TABLE 7--Example fatigue data for a paired comparison test program. (The number o f
blocks (b = 4) is made small deliberately to permit the reader to follow the numerical
details o f each step o f the analysis.)

Treatments
A B
Block Cycles to Failure Remarks

b[,~]
1 t~.] 6 620 000(2) 7 880 000o.) Vendor A
2E~] 4 310 000c2) 5 210 000o.) Vendor B
3cx~ 15 020 000o.) 9 070 000(2) Vendor C
4t4] 7 380 000(1) 3 740 000(2) Vendor D

NOTE I : Test conduct requires that both specimens in each block be tested back to back
in random time order on the same test machine at the same stress level. But different test
machines may be used to test different blocks.
NOTE 2: The time order of testing blocks is denoted by btkb assuming a single machine
is used in testing. If two or more test machines are used, the blocks assigned to each given
machine are tested in random time order.

where
y~j = numerical value observed (measured) for the i th treatment in
the jth block, namely the observed datum,
average of all observed datum values, y , , for example, the average
of the values in Table 8,
estimated treatment effect. (~ = t, - ~, where L is the average of
all measurements involving the i th treatment),
b~ = estimated block effect. (1)~ = D~ - }, where D~ is the average of
both measurements involving the jth block, and
~ = estimated value of the r a n d o m error, c o m m o n l y termed the
residual or the residual error, associated with the i th treatment in
the jth block.

M o d e l A s s u m p t i o n s - - F i r s t , the model is predicated on the assumption

that the planned experiment has the paired comparison structure of Fig. 3.
The mathematical model is not valid if the test p r o g r a m does not incor-
porate a p r i o r i the paired comparisons (the blocks) as shown.
Additional assumptions underlying the given mathematical model will
vary depending on the situation, but certain assumptions are c o m m o n to
most A N O V A : (a) The model assumes that treatment and block effects are
simply additive, either in the original or transformed scale. Thus blocks
must be chosen such that there is no interaction with treatments, namely,
such that r~ does not depend on the block in which it is measured. If experi-
ence or intuition infers the possibility of an interaction, t~ is not an accept-
able blocking variable. (If r is of interest in the given experiment, it should
be regarded as a second treatment (factor) and an interaction term (r~),~ is
added to the mathematical model. Introduction of two factors into the

TABLE 8--Example data o f Table 6, transformed using logs base 10. a (The treatment averages
and the block averages are subsequently used to compute the treatment effect estimates
and the residual errors e~i).

Treatment

Block A B Block Average

1 6.82086 6.89653 6.85869

2 6.63448 6.71684 6.67566
3 7.17667 6.95761 7.06714
4 6.86806 6.57287 6.72046

Treatment Average Grand Average

6.87502 6.78596 6.83049

a Although generally speaking logs base e are preferable in modern computation, logs
base 10 are used herein to conform to common engineering practice.

7.2
7.1
Average(A) = ~. = 6.875
7.0
6.9 l - m
~. 6.8 Grand Average = 6.83
m
6.7
0
6.6 Average (8) = B = 6.786

6.5
" 6.4
o
-- 6.3 -
Identification
.I-
-J 6,2
6.1 9 Treatment A

6.0 Im Treatment B
0
Z
I I I I I I I I
Z
I 2 3 4 5 6 7 8 Time Order of Testing Specimens

L--.5.--.J L-- I --J L--..2 - . . . J L--.4.J Black Order of Testing

FIG. 6--Plot of example data in terms of time order of testing.

t,n

overall experiment requires further consideration of the blocking procedure

and the amount of replication. Moreover, the two treatments should be
applied in a factorial arrangement when practical.). (b) It is assumed the
random variables Y,j are statistically independent, that is, the values of the
respective ~,j do not depend on one another. Investigations have shown
that if the residual errors are correlated serially, the effect on statistical
inference regarding median life may indeed be serious. But most correla-
tions among errors can be handled from a practical point of view by proper
randomization, that is, mechanical randomization may be regarded as
producing a close approximation to independence of errors. (c) It is also
assumed the variance of E~j is constant (homoscedastic) regardless of the
block or the treatment for which it is measured. Thus, data transformations
serve two essential purposes: (1) variance stabilization (block to block
uniformity of variation), and (2) normalization, that is, making the trans-
formed residual errors follow a normal distribution; (d) finally, it is assumed
that the residual errors follow a normal distribution with zero mean and
(homoscedastic) variance 0-2. Since the variance 0-2 does not depend (by
assumption) on the block in which it is measured, the individual e,~ terms
may be pooled to estimate 0-2. For the paired comparison example,

#~-= [Z#,,2]/(N,,,ocks- 1)

in which the summation acts over all residuals.

There are, in addition to the foregoing common ANOVA assumptions,
certain assumptions which serve primarily to categorize or classify various
analysis of variance models. Discussion of such assumptions and models is
beyond the scope of this manual [10].

Step 2: Plot the Transformed Data

Table 8 lists the example data of Table 7, transformed to logs base 10.
In this tabular format the block averages and the treatment averages are
computed easily. Figure 6 displays the complete data set in terms of time
order of testing, as well as indicating the respective treatment averages.
In this example the second test in each block gave a shorter life than
the first test. If this behavior were to continue for even one more block,
it would warrant further study to find its cause (if any). Clearly one should
always be on the lookout for an environmental effect in the laboratory
(for example, humidity changes) or time trends associated with changes
in test machine performance. Whether or not there may be an overall
time trend downward in this example, as testing continues, depends
on the number of machines used simultaneously in testing. If a single
machine had been used in all four blocks, it would be prudent to be on the
lookout for a time trend, that is, the probability the second life is smaller
in all four blocks is (1/2) 4 = 1/16 under the (null)assumption that the

second life is equally likely to be smaller or larger than the first life in each
block (and that block outcomes are independent of one another.)
Generally speaking, if the data show any significant statistical trends,
such trends (differences) will be quite evident in the appropriate plot.
Thus, data plotting should always accompany (and preferably precede) the
corresponding statistical analysis. In certain cases data plotting may even
indicate a priori the nature of the outcome of the statistical analysis. For
example, the individual treatment differences in Fig. 7 display considerable
variability about their average value. Hence, it is unlikely that the observed
treatment differences will lead to a rejection of the null hypothesis of no
actual treatment effects.

Step 3." Explain the Data (by appropriate parameter estimation)

The preliminary data analysis in Table 8 may now be extended to estimate
the individual values of the terms in the mathematical model. For the
paired comparison working model and its associated ANOVA
treatment effect estimate~,) =
= treatment average(o - grand average = t-~ - .~
block effect estimate o) = / ~ = block averageo) - grand average = / b -
residual erroq,~) = ~,~ = observed value - estimated value
= - + L +

(A-e)

0.5

04.

4
0.:5
|

Z 0.2
0

a
o~ o.= Average Difference
(0 08906)
=
o
oo Ira-

o:, Block Number

(in time order
-0.1
E3
-

of testtng)

-02 5 I 2

FIG. 7--Plot of example data displaying(average) treatment difference, A-B, within blocks.

These expressions were used in conjunction with the transformed data in

Table 8 to compute the entries in Table 9. For example, f~ for i = 1 equals
6.87502 - 6.83049 = +0.04453; bj f o r j = 2 equals 6.67566 - 6.83049 =
- 0 . 1 5 4 8 3 ; and ( , for i = 1 a n d j = 2 equals 6.63448 - (6.83049 +
0.04453 - 0.15483) = 0.08571.
An elementary check of the entries of Table 9 is possible by making use
of the fact that for the RCB A N O V A , El, = 0.0 and ~bj = 0.0. In addition,
2:~,~ = 0.0 for individual treatments as well as both treatments simultaneous-
ly. For example, ~/~ = (0.02820 - 0.15483 + 0.23665 - 0.11003) = 0.0
and given i = 1, Egl~ = ( - 0 . 0 8 2 3 6 - 0.08571 + 0.06500 + 0.10307) -- 0.0.
It is important to understand that the foregoing elementary computations
are valid only because the organizational structure of the paired comparison
p r o g r a m and the corresponding A N O V A lead to drastic simplifications
analytically. If the test program had deviated inadvertently from the paired
comparison format, the desired estimates could be inextricably confounded
with block effects. Accordingly, simple straightforward statistical analysis

T A B L E 9--Data "explanation" in terms o f incremental effects associated with structured

program.

y,, = ~ + ~ + ~ + ~,

i= 1 i=1

6.82086 6.83049 + 0.04453 + 0.02820 + 0.08236 j = l

6.63448 6.83049 + 0.04453 - 0.15483 - 0.08571 j = 2
3.17667 6.83049 + 0.04453 + 0.23665 + 0.06500 j = 3
6.86806 6.83049 + 0.04453 - 0.11003 + 0.10307 j = 4
i = 2 i = 2

6.89653 6.83049 - 0.04453 -+- 0.02820 + 0.08236 j = 1

6.71684 6.83049 - 0.04453 - 0.15483 + 0.08571 j = 2
6.95761 6.83049 - 0.04453 + 0.23665 - 0.06500 j = 3
6. 57287 6.83049 - 0.04453 - 0. 11003 - 0. 10307 j = 4
Degrees o f F r e e d o m a
8 = 1 + 1 + 3 + remainder

~ T h e degrees o f f r e e d o m are established as follows. First, the data y~j are viewed as a
c o l u m n vector w h o s e elements are the eight i n d e p e n d e n t observations ( m e a s u r e m e n t s )
listed. This c o l u m n vector thus h a s eight degrees o f freedom (when plotted in N = 8 space).
T h e d a t a vector is the s u m o f the respective c o l u m n vectors }, F, ~, a n d ~. T h e c o l u m n
vector ~ is comprised o f a single (repeated) element. It t h u s h a s a single degree o f freedom.
T h e c o l u m n vector ?`has two distinct (repeated) elements, b u t one set o f repeated elements
is related to the,other set, namely, Y.i'~ = 0. T h u s , ?'has one (2 -- 1) degree o f freedom. T h e
c o l u m n vector b is c o m p r i s e d o f four distinct (repeated) elements, which are related by the
(single) constraint Zb~ = 0. T h u s it h a s three (4 -- 1) degrees o f freedom. Finally, the error
( c o l u m n ) vector h a s three degrees o f freedom, namely, 8 -- (1 + 1 + 3) = 3. Its degrees o f
freedom are obtained by subtraction, because the constraints involved are usually m u c h
m o r e tedious to e n u m e r a t e directly.

and parameter estimation are associated only with test programs based on
properly planned organizational structures.
The paired comparison data are "explained" (decomposed) in Table 9.
The magnitude of the random error has been estimated for each test
observation. These residual errors are assumed to be normally distributed.
Hence, if the normality assumption is correct, the ordered residuals should
plot (lie) "near" a straight line on normal probability paper. But this
example was made small deliberately to simplify the calculations for the
reader, and therefore the data are so meager that a plot of the residual
errors on normal probability paper is hardly worthwhile. Nevertheless,
Figure 8 presents the required plot for the data of Table 9. The plotting
position used to locate (plot) the datum point along the nonlinear P-scale
is (for the normal distribution)
j - (3/8)
eplottmg position --
N + (1/4)
in which N is the number of residuals and j = l, N. The relevant plotting
positions and the ordered residuals are summarized in Table 10.
As evident in Fig. 8 the ordered residuals lie near the straight line repre-
senting the estimated normal distribution (which passes throughout the
point (0.0, 0.0) and has a slope equal to (1.0/estimated standard deviation)
= 1.0/0.139 = 7.2.) The standard deviation e for this plot is computed
using the expression e2 = [Z~2]/(Nblocks __ 1), in which the summation
acts over all residual errors.

Step 4: Statistical Analysis (generally analysis of variance (ANOVA))

The statistical analysis appropriate for the given planned experiment
depends on the test program and its stated objectives. For test programs
whose primary objective is comparative in nature, it appears the minimum
statistical analysis should proceed through the methodology of analysis of
variance. Then, assuming certain treatment effects are found to be statis-

T A B L E lO--Summary of data Jbr plotting ordered residuals.

Ordinate Abscissa

P Nonlinear Ordered Residual Errors ~

j Scale (smallest to largest)
i 0.147 --0.08571
2 0.382 --0.08236
3 0.618 +0.06500
4 0.853 +0.10307

F o r i = 1, namely, T r e a t m e n t A. N o t e that for i = 2

( T r e a t m e n t B), ~2j = --~,~.

(Y=) Z
c:
3.0
o
Z

2.0
N
Treatment A O ~ " Slope = 0.1--~ = +

1,0 Treatment B m \ - ,'.,.I ~

Z
Z
0

0.0 Z
~-(o.o,o.o)
Z
f l
J
-I.0 li~ I _.

-2.0

-3.0 I I I I I I I I I I I I I X(LINEAR)
-0.10 0.0 0.10
Ordered Residuals

F I G . 8--Plot of ordered residual errors[(i = 1) O ; (i = 2) BB].

tically significant, corresponding confidence or probability (or tolerance)

intervals may also be particularly appropriate in engineering interpretation
of the test program results.
The logarithmic transformation is commonly used in the statistical analy-
sis of fatigue life data. This transformation, when coupled with the usual
assumptions of ANOVA, implies the life data follow a two-parameter
log-normal distribution.
First, the raw fatigue data of Table 7 are transformed as illustrated in
Table 8, and then the treatment and block effects, and the residual errors
are estimated as illustrated in Table 9. Finally, the corresponding observed
value of the F statistic is computed from the data of Table 9 using the
expression
UStreatments
Fobserved
M S r e s l d u a l error

NOTE: MS signifies mean square

Definitions and illustrative calculations of MStreat. . . . t s and MSres,du~ .....
and the resulting Fob..... d, based on the information in Table 9, are as follows
2; [estimated treatment effects ]2
MStreatments
degrees of freedom for treatments
z?'2
= (Ntreatments = 2 for paired
Ntreatments-I comparison test programs)

= [(+0.04453) 2 + (+0.04453) 2 + (+0.04553) 2 +

(+0.04453) 2 + ( - 0 . 0 4 4 5 3 ) 2 + ( - 0 . 0 4 4 5 3 ) 2 +
( - 0 . 0 4 4 5 3 ) 2 + (-0.04453)2]/(2 - 1) = 0.01586
2; [residual errors ]2
M S r e s , d u ~ l error =
degrees of freedom for residual errors
Z~,7
- (Nbtock~ = 4 for this example)
Nbloek -1

= [(-0.08236) 2 + ( - 0 . 0 8 5 7 1 ) 2 + (+0.06500) 2 +
(+0.10307) 2 + (+0.08236) 2 + (+0.08571) 2 +
( - 0 . 0 6 5 0 0 ) 2 + (-0.10307)2]/(4 - 1) = 0.01932
0.O1586
Fob ..... d -- -- 0.82
0.01932
Table 11 presents the classical format for summarizing the foregoing
calculations, namely, the ANOVA table. The tabulated value of the F
statistic (with 1 and 3 degrees of freedom) at the 5 percent significance level
is 10.1, refer Natrella [1], Table A.5. The observed value is clearly non-

TABLE ll--Analysis o f variance (ANOVA) for the example paired

comparison data. DF = degrees o f freedom, M S = mean square (com-
puted in text), and F = M S t ~ t , , e , t / M S .... d ~ ......

Source DF MS F

Treatments 1 0.01586 0.82 (observed)

Blocks (3) ...
Error 3 0.0i932

Tabulated: Fl,n,0 95 = 10.1, therefore the observed F = 0.82 is not

significant.
NOTE: d- = 0.139 = ~ / ~ = v/O..01932 refer Fig. 8.

significant, namely, 0.82 is much smaller than 10.1. This outcome indicates
that an observed F-value on the order of 0.82 could easily occur by chance
when sampling randomly from the populations described by the foregoing
mathematical model with ~-a = TB = 0, that is, the " o b s e r v e d " treatment
effect estimates m a y be attributed to e (random variability) rather than r
(fixed treatment differences). However, it must be noted that the sample
size is so small that the statistical power of the test is very poor. u
The sample size given in Table 7 is so small, in fact, that if it had been
suspected a p r i o r i that the treatment effect would indeed be small, then the
test p r o g r a m should not have been conducted. Such small test programs
are effective only when the treatment effects measured are large c o m p a r e d
to ~. The outcome of the foregoing analysis does not indicate there is no
treatment effect. Rather, it merely indicates that the actual treatment effect,
if any, is too small to be detected reliably with the given sample size.
This latter interpretation may or may not be of value in the given situation.
Note that the computational results for blocks (vendors) have been
omitted from the analysis of variance table. This experiment provides no
statistical basis for comparing the vendors. There are two reasons for this:
first, the vendors were not selected at r a n d o m from the universe of all
possible sources (and seldom are), and second, the vendors are not truly
replicated. ( F r o m a broader organizational structure point of view, Treat-
ments A and B m a y be regarded as split plot treatments within each
vendor.)
I f vendors were of interest in this program, and if a single machine had
been used in testing, then a re-examination of Fig. 6 leaves the implication
that Vendor D might have basis for a complaint regarding a long-range
downward trend of the data, possibly due to deterioration of machine
performance. But, of course, if vendors were of interest, then a different
test p r o g r a m would have been used, and the vendor effects would have
been isolated from possible long-range time trends. F'roper test planning

xt The power of the test should be considered in sizing the program. Consult Refs 11-13.

eliminates much of the potential confusion in the subsequent interpretation

of the test outcome.

Summary
Most statistical analyses of fatigue life data should involve an analysis of
variance (ANOVA) or covariance (ANOCOVAR), or the equivalent in
special situations. For complex fatigue programs, a professional statistician
should be consulted to plan the program prior to testing and to conduct
the subsequent analyses. But for simple test programs, say of the type in
the foregoing paired comparison example, both the test planning and the
ANOVA are sufficiently straightforward to be carried out competently by
nonstatisticians, especially by fatigue researchers with some statistical back-
ground. In this regard, Ref i should be of considerable value to beginners
because it not only emphasizes applications, it also presents numerous nu-
merical examples. For a broader view of statistical analysis, Refs 6-8 may be
of interest, because each, to a different degree, emphasizes both theory and
application. Reference 9 pertains specifically to the fatigue test situation,
emphasizing planned experiments and the associated analysis of variance.

Chapter 3--Planning S-N and

Response Tests

Introduction
The key step in planning any fatigue test, and especially those involving
only a few specimens, is to define clearly before testing the primary test
objective. If the test data will be regarded as preliminary or exploratory,
as is common in the generation of much research data, then test planning
is markedly different than when the tests are expected to generate reliable
life data.

Conventional S-N Tests

Preliminary and Exploratory Tests

Preliminary and exploratory tests generally have stringent cost and time
constraints. Accordingly, the primary test objective for these tests is to
learn as much as possible about large fatigue effects in the shortest period
of time, using the fewest possible specimens. These restrictions usually
result in S-N tests with a single specimen being tested at most if not all of
the stress (strain) levels selected for the test program.
Although such single specimen S-N test programs generate considerable
information per specimen, experience indicates clearly that the results of
these tests are not sufficiently precise to specify accurately the magnitude
of a relatively small fatigue effect. (In fact, even a compilation of numerous
fatigue effect data based on single specimen S-N tests may not provide
a reliable estimate of the magnitude of a small fatigue effect.) In perspective,
single specimen S-N tests provide reasonably accurate test results only when
the fatigue effect being studied is sufficiently large to make negligible the
intrinsic variability of the fatigue life data inflated by the nuisance variabil-
ity associated with various other test conditions and variables that have
not been controlled or balanced.

Sample Sizes: Case 1, Shape of S-N Curve "Known"--In most situations,

especially when testing laboratory specimens (as opposed to components),
the basic shape of the S-N curve is well known. Thus, it is unnecessary to
laboriously trace out the S-N curve point by point, specimen by specimen.
It is seldom worthwhile to use more than four to six stress levels in testing,
and perhaps only two will suffice, Fig. 9. The remaining specimens should be
46
Copyright by ASTM
Copyright Int'l
9 1975 by (all rights
ASTM reserved); Sun
International Jan 3 20:27:44 EST 2016
www.astm.org
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions autho
CHAPTER 3 ON PLANNING S-N AND RESPONSETESTS 47

S
log S

Reverse Curvoture
(a) S4 Straight
S3 _
Forward Curvature

Log N

S
log S
--"-~4 " ~ I Stroight
(b)

Curvafure

Log N

log S

Fatigue Limit

Log N

FIG. 9--Three elementary types o f S-N curves (E 468-72 T) with straight and curved
domains. Generally two stress levels per domain are sufficient to describe the curve shape
adequately. Note that fatigue limits should be estimated using either the up-and-down strategy
or the two-point method.

used to provide replication of stress levels previously tested. Generally,

long-life specimens warrant greater replication than shorter-life specimens
(because generally the variability of long-life data exceeds that of shorter-
life data). On the other hand, if the number of specimens available equals
approximately twice the number of stress levels to be tested, uniform
replication (two specimens at each stress level) provides a satisfactory
allocation for research data.

Sample Sizes: Case 2, Shape of S-N Curve "Unknown"--When no

prior information regarding the shape of the S-N curve is available, the
number of single specimen stress levels may be increased to six to eight,
assuming the resulting S-N curve thereby takes on a recognized shape,
for example, one of the forms displayed in Fig. 9, or perhaps increased to
ten to twelve, assuming the resulting S-N curve has certain peculiarities.
There is seldom if ever any reason for testing at more than twelve different
stress levels, regardless of the complexity of the S-N curve, and the number
of peculiarities that may occur.
Once the S-N curve takes on a recognizable form, the testing emphasis
should immediately shift from curve form definition to replication, namely,
precise specification of the median S-N curve.

Test Planning---The first goal of planning preliminary and exploratory

tests is to obtain a random sample. This random sample helps assure that
the specimens tested are representative of the conceptual universe of
specimens about which both statistical and engineering inference will be
made. For example, if there are two primary suppliers for a proposed
material, each vendor's stock should be randomly sampled to obtain
(n/2) test blanks. If the "additional variability" involved by sampling from
both suppliers is even suspected to be large enough to obscure the fatigue
effect being studied, then the simple single specimen S-N test program is
clearly inappropriate for the study. Analogously, it is generally preferable
to select one specimen from each of ten bars of stock rather than selecting
ten specimens from one bar of stock.
The second goal of planning preliminary and exploratory tests is to
balance the potentially spurious effects of nuisance variables by appropriate
planned grouping (blocking). For example, if there are two vendors and four
test machines, each vendor's product should be tested on each test machine
the same number of times. Then, even if no attempt is made to estimate
the machine effect, if any, the fairness of the given test program cannot be
questioned. In contrast, it is difficult to defend the fairness of a test program
in which all of one vendor's stock are tested on one specific machine, while
all of another vendor's stock are tested on a different machine. Similarly,
if the test program involves several treatments, say corrosion specimens
versus specimens tested "in air," it is difficult to defend a test program
which tests all "in air" specimens, then modifies the test setup, and sub-
sequently tests the remaining specimens in various corrosive environments.
It is equally poor practice to conduct the "in air" tests on a separate test
machine. (The planning of comparative tests is discussed in Chapter 1.)
The third goal of planning preliminary and exploratory tests is to allow
for possible test equipment difficulties and modifications. This can often
be accomplished by conducting the S-N tests in small time blocks to form a

sequence (series) of "mini tests." This technique avoids problems associated

with mandatory test setup modifications in the middle of a long stream of
related tests.

Replicated S-N Tests

After the gross fatigue effects have been identified and their approximate
magnitudes estimated by exploratory (research) tests, the usefulness of
further single specimen S-N tests diminishes markedly. Replication then
becomes the central issue. Figure 10 displays an acceptable compromise
between the number of stress levels used in testing and the amount of
replication at each level.
Generally, as the amount of replication increases the number of stress
levels decreases. Often the reason is simply that the need for reliable data
is accompanied by a tighter understanding of the fatigue lives of interest
in the given situation. In other cases, the reason is merely that the total
test time or test specimens available is fixed. In either case, the number of
stress levels used in a replicated S-N test program should be no more than
that used in the prior preliminary and exploratory test program, and
perhaps as few as two stress levels.

Two Stress Levels Only--Whenever the range of stress levels is suffi-

ciently small that the S-N curve may be described reasonably by a straight
line in the life interval of interest, then only two stress levels should be
used in testing, $2 = Sm.~xand $1 = Snlin. If the variance of log life is uni-

9 Fretting

9 No Fretting
130 A IA 9b d I IItlL J

OOe
125

v" 120

I10 9 ~
I05
I00100 2o0 500 10o0 2000

N, C y c F e s x 10 - 3

F I G . lO--Aeceptable compromise between number o f stress levels and replication at each

level. Note that the middle stress level for the straight line S-N curve couM be deleted without
affecting the fitted curve, and also that either otte o f the middle two stress levels for the other
S-N curve could be deleted without markedly affecting the faired (sketched) curve.

form along the life interval of interest (that is, homoscedastic variance)
and the total number of specimens available for testing is fixed, the allo-
cation of specimens should be as follows: test one half the specimens at
S~, the other half at $2. But if test time, cost, or the precision of some
point on the S-N curve is critical, Ref 9 or 14 should be consulted for
optimal specimen allocation.

Three Stress Levels Only--Whenever experience indicates the S-N curve

is smooth with a gradual forward curvature, three more or less equally
spaced stress levels may be used in testing 12, assuming the life range o f
interest can be limited, for example, to the range associated with $1, $2, 5'3
in Fig. 9b. Otherwise, perhaps even four stress levels may be considered.
Each datum point along the S-N curve is established with equal precision
when
(n,/n,) = (~ 2/~ 2)

where n~ is the number of specimens tested at stress level S,, whose standard
deviation of log life is a,. For uniform variance conditions (that is, homo-
scedasticity) this elementary allocation scheme dictates equal sample sizes
at each stress level. Such allocation, however, may be relatively inefficient
when estimating parameters in polynomial regression, or when predicting
the location of fitted points on the polynomial regression curve. (Consult
the references in Ref 14.)

Three Stress Levels Perhaps--On occasion it may appear desirable to

check the adequacy of the linear model [8, 9, or 13] for the S-N curve in
the life range of interest. In this situation three stress levels may be used.
However, it is not sound policy to test at three stress levels just as a.pre-
caution that the S-N curve may display a slight forward curvature. (Either
a slight forward, or a slight reverse curvature will be observed in 50 percent
of repeated experiments when the true model is indeed linear!).

Four Stress Levels at Most--If the range of fatigue life of interest is

sufficiently wide that it is decided to fit a second order expression to the
S-N data, perhaps four stress levels should be considered to permit estab-
lishing the statistical significance of the quadratic term. (When only three
levels are used in testing the fitted quadratic passes through all three Y~
averages, allowing r~o assessment of the deviations of the averages from the
fitted curve.) On the other hand, if five or more levels are used in testing,
either the replication at each stress level suffers markedly or the number of
specimens required for the test program exceeds practical limitations. If
replication must be sacrificed to extend the range of fatigue life of interest,

'~ If the S-N curve is to be described in terms of Log S, then equal spacing refers to
Log S, not S.

then the test program appropriate for this situation is that corresponding
to preliminary or exploratory testing.
Test Planning--The time order of testing is important in testing groups of
specimens at several stress levels. If tests are conducted at say three stress
levels, one specimen should be tested at each stress level in random time
order before starting the second set of replicates, etc. In particular it is
mandatory not to test all of the specimens allocated to the highest stress
level before testing all of the specimens allocated to the next lowest level,
etc. The latter technique guarantees analytical difficulties if an equipment
malfunction occurs, or if a fixutre modification is required. Moreover, the
random assignment of stress levels in time order usually assures an inde-
pendent stress setting for each new test, that is, assures true replication of
the stress level treatment. If two or more test machines are used, the time-
machine blocks may be confounded with as many other nuisance variables
as desired.
When blocks appear in the organizational structure of the test program,
analysis of variance is recommended to separate the nuisance variability
from the variability associated with true replication. Generally the overall
analysis is simplified when it is assumed the underlying distribution of log
life is normal, that is, fatigue life is log normally distributed.

Replication Guidelinesfor S-N Testing

The replication guidelines below are based on our definition for percent
replication:
total number of

I
different stress levels|
used in t e s t i n ~ J
~o replication = 100 I-
total number of |
specimens tested _J

Type of Test Percent Replication

Preliminary and exploratory (exploratory 17 to 33 min

research and development tests)
Research and development testing of 33 to 50 min
components and specimens
Design allowables data 50 to 75 min
Reliability data 75 to 88 min

It is also evident that the minimum number of specimens required in

testing depends on the type of test program conducted. The following
recommendations appear reasonable.

Minimum Number
Type of Test of Specimens

Preliminary and exploratory (exploratory research and 6 t o 12

development tests)
Research and development testing of components and 6 t o 12
specimens
Design allowables data 12 to 24
Reliability data 12 to 24

Discussion of Guidelines
Generally speaking when the shape of the S-N curve is known on the
basis of experience or a literature survey, then at least 50 percent replication
is desirable. The minimums given in the guidelines are justified only when
necessitated by compelling considerations. When the shape of the S-N
curve is unknown, the percent replication should increase markedly as the
test program progresses. In particular, at least three out of four specimens
should be replicates by the time the number of stress levels used in testing
reaches the order of six to eight.
It is advisable to allocate at least two specimens for purpose of replication
in all S-N test programs, including those involving the minimum recom-
mended number of specimens. These two specimens may either be used to
replicate the highest and lowest stress levels used in the test program, or
both may be used to replicate only the lowest stress level. The latter alloca-
tion is usually slightly conservative in estimating the variability of the
fatigue life data (namely, tends to give a larger estimate of the variability
than occurs for replicating both the highest and lowest stress levels, because
the variability of fatigue life data tends to increase as the median life
increases).

Fatigue Strength and Fatigue Limit Tests

Probit Method
In the well-known Probit method of conducting fatigue strength and
fatigue limit test programs, a group of specimens is tested at each of several
uniformly spaced stress levels. Such data may be used subsequently to fit a
fatigue strength response curve, commonly termed the P-S curve [9,15,16].
However, the Probit method is relatively inefficient for estimating the
median fatigue limit from a statistical point of view, and it usually requires
more specimens than are available for fatigue testing. Consequently the
Probit method is not recommended in this manual. Rather, the two-point
method presented later should be used whenever the number of test speci-
mens is limited.

If the number of specimens available for testing is adequate for the

Probit method, say 30 or more specimens, then Fig. 11 may be used to
allocate specimens to stress levels to obtain the most precise estimate of the
median fatigue limit. Note that testing at any arbitrary stress level (what-
ever the spacing and whatever the number of specimens tested) is approxi-
mately equally effective in estimating the median fatigue limit whenever
the true probability of failure is between roughly 0.3 and 0.7 for symmetrical
distributions such as the normal and logistic [17] response curves. For
other point estimates however, optimal allocations are not nearly so
flexible [9].

O. 70

Normal
0.60 Distribution \

/ /
0.50

Weight
Extreme Value
w
/ -Smallest
0.40 Distribution

[w = (dP/dY)2]
PQ J
I

0.30

020 / / Logistic \
Distribution
\
\
O.iO
\
o.oo
o.o o.I o.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 I.O
P

FIG. 11--Plot o f statistical weight w (per specimen) versus P, the true probability o f
failure at the given stress level used in testing. For symmetrical distributions the weight w
is directly proportional to the effectiveness o f that specimen in locating the median fatigue
limit.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
54 M A N U A L O N STATISTICAL PLANNING AND ANALYSIS

77.5

750 X X X X , -
I0 14 22 28
725 X O--'X----O'mX X---<)--" X - - X "--<)-- X - -
5 9 II 15 15 19 21 23 25 27 29
Cn 70.0 o x-o--o o-o o-
"4 16 18 20 24 26 30
6Z 5 o o o
CO 7 17
65.0 - o
2
6 2 . 5 OI , 0 No FaiLure
X Foilure
60.0

Test Number

FIG. 12--Illustration of the up-and-down test strategy.

Up-and-Down Method
Up-and-Down Strategy--Specimens are tested sequentially, one at time
in the most elementary version of this strategy, as indicated in Fig. 12.
The first specimen is tested at a stress level approximately equal to the
median fatigue limit estimated by experience or using preliminary S-N
data. If this specimen survives a specified number of cycles N* (or a given
duration), the next specimen is tested at a higher stress level. Otherwise, it
is tested at a lower stress level. Generally, for convenience of analysis the
spacing of the increments between adjacent stress levels is uniform (con-
stant). However, arbitrary spacing is quite permissible when it is to be
accompanied by a Probit-type analysis [1] or [9].
The primary advantage of the up-and-down strategy over the Probit
method strategy is that the former estimates the median fatigue limit much
more rapidly in terms of number of specimens expended. However, once
the approximate median fatigue limit has been estimated by up-and-down
testing, this strategy no longer retains its advantage, and may indeed
become relatively inefficient.

Two-Point Method
Two-Point Strategy--The two-point strategy [9,18] is the same as the up-
and-down strategy until the first test result such that two nonzero and
nonunity proportions failed are observed, 13 after which specimens are
tested only at the two corresponding stress levels. In Fig. 12 the second
nonzero, nonunity proportion failed is observed at specimen 9. Subse-
quently all tests should be conducted at either $1 (70.0) or $2 (72.5). Note
that the two-point strategy avoids testing four specimens at $3 (75.0)
where the statistical weight associated with the specimens is diminished
markedly. In addition, testing of specimen 17 at an equally ineffective
stress level is avoided.

Is The up-and-down strategy is used until there are two P-values to plot on the proba-
bility paper o f Fig. 13.

Two-Point Analysis--Analysis for the two-point method is usually quite

straightforward. The median fatigue limit often may be estimated merely
by plotting the two proportions failed (at $1 and &) on probability paper,
and interpolating as required. (Note that the "fitted" response curve
passes through the two points and is a straight line on the appropriate
probability paper, Fig. 13.) The asymptotic variance for the median point
(~'~0) on the fitted response curve is (ignoring all specimens tested at stress
levels other than $1 and $2) [18] 14
- 7
(~(Y~0) -- ($2 $1) 2 1- -N1~1 + U2~z _J

For the normal and logistic distributions, 950 equals zero when estimating
the median fatigue limit.
Example--Estimate the median fatigue limit for the data in Fig. 12
assuming the specimens 10, 14, 22, 28, and 17 have not been tested (omitted
according to the two-point strategy). Also ignoring specimens l, 2, 3, and 7,
the resulting data are

70.0 72.5 0.200 0.636

(2//10) (7/11)
Assuming an underlying normal distribution (950 = 0.0), in Fig. 13:
Y1 = -0.842 (from Normal tables)
~'2 = +0.349 (from Normal tables)
Hence, $50 = 70.0 + 2.50 (0.842/(0.842 + 0.349)) = 71.77. Next, reading
~ and ~2 from Fig. 11,
f)l = 0.49 ~.., = 0.61
1 r (0"73)2 (1.77)'
d'~Y"~~ - (2.5) 2 I_(10)(0.49) + (ll)(0.6D_l = 0.o92
Thus, the estimated 2-sigma limits are: i 0 . 6 1 - . Accordingly, the 95 percent
confidence limits stated in terms of percent survival at N* cycles with
~50 = 71.77 are: Plo = 27 percent, Ph, = 73 percent. The corresponding
approximate 95 percent confidence limits for the median fatigue limit are
(Fig. 13): Sso.lo = 70.5, ~o,h, = 73.1.
Test Planning--Preliminary testing is absolutely essential to the conduct
of efficient up-and-down and two-point tests. Preliminary tests can be
avoided only when ancillary information or experience can produce
sufficiently accurate initial estimates to minimize test time (wasted) prior

i4 T h i s v a r i a n c e is m i m m u m w h e n (for fixed Ntotat) Ntt~'t2(S2 - S~0) 2 = N'2~,fl(St -- $5o) ~.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
y(LINEAR) P(NONLLNEAR) Normal probab| I I ty paper U~
O~

3:
(Y=) >
z
c:
3.0

SI S2 0
z

2.0 f 1.960
L645
.975
.95 I
1.282 .90 "
1.0 + 0.61 -1
>.
0"842 "80 P = 0"636 i z
^
Y2 =
0.524 .'tO 2 "~ 9
z_
z
+0.349 0.Z53 .60 0
o.o o.o .5o r~, ,''r F I
.o.~53 Ao I ,-- z
o
- - -0.5Z4 .30 ~ --
J
z
"~1 : 1 9 .o.,,,~ .~O-~l = O. 2 0 0 ' 9~ - ~ I,,';
-o.842 -,.o- ,......~. ~ I_o.6, , _. ,. .,<

-- -I.282 .10 / () I l ,I

-,.6,5 .o5 " I o ' ~-

.2.o- .L~o .o25 " I ..... ~ I d

.2.326.0I ~ J ,! , ~
- o= I I
..~ I I 1 I I I "~1 I Ill II l I t
68 69 70 71 72 7:5

S, KSI

F I G . 13--Analysis for the two-poiltt strategy. The fitted response curve (a straight line) passes tkroug]t the two points used in testing (St, PI;
S~, PO.

to obtaining the first response reversal (namely, the X at specimen 5 in

Fig. 12).
If the first few tests are conducted above the fatigue limit, Fig. 14, these
data can also serve to establish an approximate S-N curve. Thus specimens
1, 2, and 3 in Fig. 12 are "wasted" in the context of providing no informa-
tion for sketching the S-N curve. On the other hand, if the number of
specimens is restricted severely and if the S-N curve is of no particular
interest (and if coaxing can be ignored, say either from a knowledge of
the material behavior, or by using widely spaced stress levels), then a
single specimen could be used to home in on the median fatigue limit.
But, as illustrated in Fig. 15, a considerable amount of test time can be
wasted by starting the test program at a stress level that is too low or by
using stress increments that are too small in approaching the median
fatigue limit. Thus, considering its limitations, this single specimen tech-
nique cannot be generally recommended.
One of the basic goals of preliminary testing is to obtain an idea of the
stress levels which are likely to cause failure and those which are not. This
information is then used to establish the incremental spacing between
successive stress levels. Ideally this spacing should be in the range of (2/3)
to (3//2) the standard deviation of the underlying distribution. For steel
laboratory specimens the spacing should be on the order of 1 to 3 percent
of the ultimate strength of the material [16]. (But, for manufactured parts,
the standard deviation may be as large as 20 percent of the median fatigue
limit, or even larger.) If the spacing is too wide the test outcomes will
eventually "ping-pong" back and forth between X's and O's. If the spacing
is too narrow, the test outcomes will exhibit numerous reversals on each
side of the median fatigue limit.
Sample Sizes--The two-point strategy generally requires eight to ten
specimens to obtain two nonzero, nonunity points for plotting on proba-
bility paper. If too few specimens are tested, or if the spacing of the tests
that have been conducted is so wide that two nonzero, nonunity points
are not available for plotting, the median fatigue limit can be estimated
only if the standard deviation of the underlying distribution is assumed
known. If too few specimens have been tested, Refs 17, 19, or 20 should be
consulted for tabulated quantities appropriate for estimating the median
fatigue limit and its 2-sigma confidence interval. In the wide spacing situa-
tion the median fatigue limit may be estimated as lying halfway between
the two stress levels used in testing, but there is no valid estimate of the
corresponding 2-sigma confidence interval.
Two or More Test Machines--When two or more test machines are used
to speed data output, it is advantageous for purposes of test simplicity to
view each machine as generating an independent estimate of the median
fatigue limit. Then the test strategy for each individual test machine is

X~
"X Z
Z
\_ Z
G3
x,, Z

Z
0
w v

.~- II.-
3 4 5 Number
Log Cycles
Specimen Number Specimen Life

F I G . 1 4 - - S - N data generated by the first f e w specimens in a two-point test program.

t~
-p
\
c~
o
Z

x
Z
z
Z
0 0
0
w
Z
R.efest Runouts "7

I la fb Number Log Cycles t.n

Specimen Number Specimen Life

FIG. 15--Preliminary testing ill which a single test specimen is used. (Note) This procedure is not r e c o m m e n d e d except in extreme situations involving
expensive components.

t~
*,0

independent of the outcomes for the other test machines. The overall
results may subsequently be combined using a Probit-type analysis [1] or
[9] (confounding machine differences with the median fatigue limit).
When using several machines to speed response data output care should
be taken to assure that individual machine calibration differences do not
inadvertently introduce a bias into the slope estimate of the response curve.
Note that if a given test machine has a calibration error of magnitude 6, the
slope of the overall response curve will be biased if all its tests are con-
ducted on one side of the median fatigue limit. But this bias will be dimin-
ished markedly by an attempt to balance the stress levels above and below
the median fatigue limit.

Chapter 4 Distribution-Free Statistical

Tests for Comparing Two Censored Life
Samples (in a CRD Test Program)

Nomenclature

a Size of the Type I error, that is, the probability of incorrectly

rejecting the null hypothesis Y = X
DNF Did not fail, that is, a runout in fatigue testing
J~ Number of failed X specimens
Jy Number of failed u specimens
My Total number of Y specimens
N= Total number of X specimens
RH Number of run-outs (suspended tests) for Halperin's statistic
(a random variable)
% Number of Y failures used to terminate the test program for
Young's statistic (Type ll censoring)
U Mann and Whitney's statistic, namely, the count of all X, Y pairs
in which a Y precedes an X
UH Halperin's statistic, namely, the count of all X, Y pairs in which a Y
precedes an X, corrected for Type I censoring
Ur Young's statistic, namely, the count of all X, Ypairs in which an X
precedes a Y, corrected for Type II censoring
u Observed value of U
uh Observed value of Utl
uy Observed value of Ur
X Reference population (material)
X Random life associated with X
x, i *~' smallest observed X life
Y Test population (material)
Y Random life associated with Y
y, i th smallest observed Ylife

Introduction

All measurements are comparative in nature. Sometimes, however, the

comparisons pertain to data separated so distinctly by time, geography,
and personnel that the comparison must be qualified, if not discounted
61
Copyright by ASTM
Copyright Int'l by
9 1975 (allASTM
rights reserved); Sun Jan 3 20:27:44
International EST 2016
www.astm.org
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
62 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

entirely. On the other hand, when the comparison is made by the same
personnel in the same laboratory at the same time, both the physical
interpretation and the statistical inference clearly have greater validity.
Moreover, if one of the samples pertains to a reference population (say
the material, processing, configuration, etc., presently being used), then
the interpretation and inference are enhanced even further.
We shall assume that all data are generated using a completely ran-
domized design (CR D) test program; namely, we assume that all data are
generated at the same time ~ by the same personnel using the same equip-
ment, and that all sample allocations are established using random number
tables. These assumptions, in effect, validate the use of the statistical tables
presented herein.
Let X be the reference population. Assume Nx specimens are tested, and
the data, ordered from smallest to largest, are: xl, x2 . . . . . XN~. Further,
let Y be the other (test) population. The respective ordered Y data are:
yl, Y2. . . . , Yiy. Usually, My < Nx for practical reasons.

Complete (uncensored) Data

When all N~ and My life values are known, the data are complete (un-
censored). Such data are amenable to a number of elementary statistical
analyses [13], some of which assume that the form of the distributions of
the two life populations are known. The statistical tables provided herein,
however, are for use with statistical analyses that are distribution-free,
namely, the statistical tests do not require the explicit assumption of a
specific life distribution form.

Censored Data
Usually, when an individual fatigue test specimen does not fail within a
reasonable period of time, the test is suspended, and the only life informa-
tion available is that the given specimen endured a specific test duration.
Or, when the resources allocated to the given test program are limited, it
may be necessary at some point in time to suspend all fatigue tests then
still in progress. In either ease, the test data are censored.
There are a number of different categorizations of censored data. In fact,
each distinct set of rules for suspending individual tests leads to a different
category (and statistical analysis) of censored data. For example, the overaU
test program could be suspended as soon as the first X failure is observed,
or it could be suspended as soon as the third failure is observed (counting
both X and Y), or it could be suspended as soon as all specimens have
either failed or endured a specific test duration. Each of these three cases

~a That is, during a specified time period in which it may reasonably be assumed the test
conditions and environment are invariant for practical purposes.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions autho
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 63

requires a different set o f statistical tables to perform the c o r r e s p o n d i n g

statistical analysis. I~
In addition, there m a y be alternative statistical analyses for a specific
set o f rules used to suspend the test p r o g r a m . Consequently, some dis-
cretion must be used to limit the n u m b e r o f statistical tables presented
herein, b o t h for the sake o f e c o n o m y , and by keeping the statistical analyses
as simple as possible. Accordingly, we shall present only two sets o f sta-
tistical tables for censored data, and refer the reader to the literature for
other tables and analyses. These two tables should suffice to cover m o s t
applications in fatigue and materials testing. Other distribution-free
statistical analyses are presented in Refs 21 and 22.

Test of Hypolhests
The classical statistical test o f hypothesis technique is receiving m o r e
and more criticism as this "decision m a k i n g " technique finds b r o a d e r
application. Nevertheless, it provides a starting point in analysis and
interpretation of fatigue life data.
The null hypothesis for c o m p a r i n g two life populations is simply that
b o t h distributio.ns are identical, whatever the distributions and the p a r a m e -
ter values. Then, if the null hypothesis is true, there would be (Nx + M~)
factorial equally likely permutations for the c o m b i n e d (X plus Y) set o f
ordered data. O f these permutations, only a small fraction will c o r r e s p o n d
to the situation where m o s t X's precede, say, the first observed Y (termed
yl). Hence, if that were the given test p r o g r a m outcome, we have either
observed an unlikely event if the null hypothesis is true, or a m o r e likely
event if indeed Y > X . : Usually, if the probability of the given event is
less than 0.10 assuming the null hypothesis is true, we opt for the decision
that the null hypothesis is not true, that is, we reject the null hypothesis.
N o t e that if indeed the null hypothesis is true, we would incorrectly reject
it less than 0.10 X 100 percent of the time in repeated applications (repeated
trials). This mistake is termed a Type i error.
On the other hand, we might also m a k e the mistake of not rejecting the
null hypothesis when it is false. This mistake is called a Type II error, The

~6The two most common types of censoring are referred to as Types I and II in the
statistical literature. Type I pertains to censoring based on test t~me endured, namely, each
individual test is suspended as soon as the specimen has survived a specific prestated test
duration. For example, run-outs in fatigue testing (at say N* cycles) correspond to Type I
censoring. Note that the number of suspended tests is a random variable for Type I censor-
lng. In contrast, Type II censoring pertains to a censoring plan based on a prespecified
number of censored observations, for example, suspending the overall test program as
soon as the second Y failure is observed (to accelerate the comparison).
: Y > X Is read : Y is stochastically greater than X; namely, F(z) > G(z)for all z, where
F and G are the cumulative distribution functtons of X and Y, respectively. A loose verbal
translation is that Y is greater than X in distribution. X and Y are random variables.

larger the Type I I error the smaller the power of the statistical test of
hypothesis, namely, the smaller the probability of correctly rejecting the
null hypothesis when it is false.
For a given significance (Type l error) level, the power of a statistical
test of hypothesis is controlled primarily by the sample sizes used in the
test program. But test planning is also important. Thus consultation with a
statistician to assure reasonable statistical power before becoming com-
mitted to an expensive test program is quite prudent.

Alternative Hypotheses
Types I and II errors are not independent of one another. In order to
tabulate the significance levels pertinent to testing the null hypothesis,
we must properly state the alternative hypothesis, namely, what we con-
elude if we reject the null hypothesis.
Elementary alternative hypotheses are of two types: one-sided or two-
sided. A one-sided alternative hypothesis is used when the question is
whether Y is an improvement over X, that is, the alternative hypothesis is
Y > X. On the other hand, the two-sided alternative is used when the
question is whether Y is different than X, either Y > X or Y < X. In
summary

Test of Hypothesis: One-Sided Alternative

null hypothesis: Y = X
alternative hypothesis: Y > X

Test of Hypothesis: Two-Sided Alternative

null hypothesis: Y = X
alternative hypothesis: Y ~ X
(Y > X or Y < X)

All tables given herein pertain to the one-sided alternative hypothesis

Y > X, where X is the reference population and Y is the test population.

Complete (uncensored) Data

1. Suppose the following data are observed

X: 212, 329, 192, 401,231,345 (Nx = 6)

Y: 112, 143, 177, 151 (My = 4)

Consider the null hypothesis Y = X relative to the alternative hypothesis

Y > X. Answer: No analysis is necessar2~. Visual inspection of the data
suffices to indicate that we would not reject the hypothesis Y = X in favor
of the alternative hypothesis Y > X. (But note that the alternative hypothe-
sis X > Y looks attractive.)

2. Suppose the following data are observed

X: 112, 143, 177, 151,231,345 (Nx = 6)

Y: 212, 329, 192, 401 (My = 4)

Consider the null hypothesis Y = X relative to the alternative hypothesis

Y > X. Answer: Visual inspection does not suffice in this case. Thus, we
first order each set of data from smallest to largest

X: 112, 143, t5t, 177, 231,345

Y: 192, 212, 329, 401

Next, we order the entire collection of data, but retaining the X, Y identity
of the individual datum

112, 143, 151, 177, 192, 212, 231,329, 345,401

X X X X Y Y X Y X Y

Now, the test statistic used to refer to statistical tables must be defined for
these data. Note that the alternative hypothesis Y > X is neither over-
whelming nor unreasonable.
In overall perspective, Mann and Whitney's U statistic provides the best
distribution-free statistical test available for the two sample life com-
parison situation with no censoring, Table 12.

Mann and Whitney's U Statistic

Mann and Whitney [23] introduced an elementary statistic U which is
equal to the number of times a Y precedes an X in the combined ordered
sample of Nx X's and M~ Y's. They show that the limiting form of the
distribution of U is normal if M,j and N~ go to infinity in any arbitrary
manner. The power of the U statistic is considered in Ref 24.
Example--Given Nx = 6 and M u = 4. Suppose the test outcome stated
in terms of the combined ordered sample is

XXXX Y Y X Y X Y
The observed value of U, termed u, is 5. (Note that each of the first two
Y's precede two X's, and thereby contribute four counts to u. The third
Y precedes one X and thus contributes the final count to u.)
Example--Given N~ = 10 and M~ = 2. Suppose the test outcome is
XXXX Y XXXXXX Y
In this case u = 6.

Mann and Whitney's Tables

Table 12 consists of ten columns: two columns to identify M~ and Nx,
and four pairs of columns, each pair denoted u and Prob. The four u

columns state the critical significance values of u~ for which (if the null
hypothesis is true)
Prob [U < u~] < , ~
where a, the size of the Type I error, is equal to 0.10, 0.05, 0.025, and 0.01,
respectively. The four adjacent columns denoted Prob state the exact size
of the significance test for the corresponding values of u~.
Given the tables and the observed value u of U, the appropriate sig-
nificance test of size a is
reject the null hypothesis if u < uo, or
"accept" the null hypotheses if u > u~
Example--Given the outcome X X X X Y Y X Y X Y, consider the
statistical significance at the 0.10 level.
Answer--For My = 4 and N~ = 6, Table 12 indicates that a value of U
equal to 5 or smaller would occur 0.0857 X 100 percent of the time by
chance if the populations were indeed identical. This probability is suffi-
ciently small that we would usually opt to reject the null hypothesis and
believe instead that Y is better than X. (Note that a value of U equal to one
or smaller would occur by chance less than 1 percent of the time. This
probability is clearly sufficiently small that we would surely opt to reject
the null hypothesis and believe instead that Y is better than X.)
Example--Given the outcome X X X X Y X X X X X X Y, consider the
statistical significance at the 0.10 level.
Answer--For Mu = 2 and N~ = 10, examination of Table 12 indicates
that a value of U equal to 3 or smaller is regarded as significant statistically
at the 0.10 level (exact size 0.0909). Thus, the observed value u = 6 does
not support a decision to reject the null hypothesis, that is, we have no
overwhelming reason to suspect the null hypothesis even though the odds
favor the proposition that Y is better than X.

Censored Data
Run-Outs
Case / - - S u p p o s e each individual test is terminated as soon as the
specimen fails or is a run-out at say N* cycles. In this case, the overall data
consist of the ordered life data and the numbers of X and Y run-outs.
1. Suppose the following data are observed
X: 212, 329, 500(DNF) is, 192, 500(DNF),
401,231,345 (Nx = 8)
Y." 112, 143, 177, 151 (My = 4)

18DNF = did not fail, that is, is a run-out.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 67

T A B L E 12,--Mann and Whitney's U statistic, a

U,~ ua Ua u,~
M~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 0 0.1000 0 0.0 0 0.0 0 0.0

10 0 0.0909 0 0.0 0 0.0 0 0.0
11 0 0.0833 0 0.0 0 0.0 0 0.0
12 0 0.0769 0 0.0 0 0.0 0 0.0
13 0 0.0714 0 0.0 0 0.0 0 0.0
14 0 0.0667 0 0.0 0 0.0 0 0.0
15 0 0.0625 0 0.0 0 0.0 0 0.0
3 0 0.I000 0 0.0 0 0.0 0 0.0
4 0 0.0667 0 0.0 0 0.0 0 0.0
5 1 0.0952 0 0.0476 0 0.0 0 0.0
6 I 0.0714 0 0.0357 0 0.0 0 0.0
7 I 0.0556 0 0.0278 0 0.0 0 0.0
8 2 0.0889 i 0.0444 0 0.0222 0 0.0
9 2 0.0727 I 0.0364 0 0.0182 0 0.0
I0 3 0.0909 I 0.0303 0 0.0152 0 0.0
II 3 0.0769 I 0.0256 0 0.0128 0 0.0
12 4 0,0989 2 0.0440 I 0.0220 0 0.0
13 4 0.0857 2 0.0381 1 0.0190 0 0.0095
14 5 0.I000 3 0.0500 i 0.0167 0 0.0083
15 5 0.0882 3 0.0441 i 0.0147 0 0.0074
2 0 0.I000 0 0.0 0 0.0 0 0.0
3 1 0.1000 0 0.0500 0 0.0 0 0.0
4 i 0.0571 0 0.0286 0 0.0 0 0.0
5 2 0.0714 I 0.0357 0 0.0179 0 0.0
6 3 0.0833 2 0.0476 I 0.0238 0 0.0
7 4 0,0917 2 0.0333 I 0.0167 0 0.0083
8 5 0.0970 3 0.0424 2 0.0242 0 0,0061
9 5 0.0727 4 0.0500 2 0.0182 I 0.0091
I0 6 0.0804 4 0.0385 3 0.0245 I 0.0070
II 7 0.0852 5 0.0440 3 0.0192 1 0.0055
12 8 0,0901 5 0.0352 4 0.0242 2 0.0088
13 9 0.0946 6 0.0411 4 0.0196 2 0.0071
14 I0 0.0985 7 0.0456 5 0.0235 2 0.0059
15 i0 0.0821 7 0.0380 5 0.0196 3 0.0086
2 0 0.0667 0 0.0 0 0.0 0 0.0
3 i 0.0571 0 0.0286 0 0.0 0 0.0
4 3 0.I000 i 0.0286 0 0.0143 0 0.0
5 4 0.0952 2 0.0317 I 0.0159 0 0.0079
6 5 0.0857 3 0.0333 2 0.0190 I 0.0095
7 6 0.0818 4 0.0364 3 0.0212 I 0.0061
8 7 0.0768 5 0.0364 4 0.0242 2 0.0081
9 9 0.0993 6 0.0378 4 0.0168 3 0.0098
10 10 0.0939 7 0.0380 5 0,0180 3 0.0070
II II 0.0886 8 0.0388 6 0.0198 4 0.0088
12 12 0.0852 9 0.0390 7 0.0209 5 0.0099
13 13 0,0815 i0 0.0395 8 0.0223 5 0.0076
14 15 0.0961 ii 0.0395 9 0.0232 6 0.0088
15 16 0.0924 12 0.0400 I0 0.0243 7 0.0098
2 1 0.0952 0 0.0476 0 0.0 0 0.0

aSee Note at end of table, p. 70.

TABLE 1 2 - - M a n n and Whitney's U statistic (continued).

u~ u~ u~ ua
Mr N~ 0.10 Prob 0.05 Prob 0.025Prob 0.01 Prob

5 3 2 0.0714 1 0.0357 0 0.0179 0 0.0

5 4 4 0.0952 2 0.0317 1 0.0t59 0 0~
5 5 5 0.0754 4 0.0476 2 0.0159 1 0.0079
5 6 7 0.0887 5 0.0411 3 0.0152 2 0.0087
5 7 8 0,0745 6 0.0366 5 0.0240 3 0.0088
5 8 I0 0.0855 8 0.0466 6 0.0225 4 0.0093
5 9 12 0,0949 9 0,0415 7 0.0210 5 0.0095
5 I0 13 0.0823 II 0.0496 8 0,0200 6 0.0097
5 II 15 0.0902 12 0.0449 9 0.0190 7 0.0096
5 12 17 0.0973 13 0.0409 II 0.0242 8 0.0097
5 13 18 0.0867 15 0.0473 12 0.0230 9 0.0097
5 14 20 0.0930 16 0.0435 13 0.0218 I0 0.0097
5 15 22 0.0987 18 0.0491 14 0.0209 II 0.0097
6 3 3 0,0833 2 0.0476 1 0.0238 0 0.0
6 4 5 0,0857 3 0.0333 2 0.0190 I 0,0095
6 5 7 0.0887 5 0.0411 3 0.0152 2 0.0087
6 6 9 0.0898 7 0.0465 5 0.0206 3 0.0076
6 7 II 0.0903 8 0.0367 6 0.0175 4 0.0070
6 8 13 0.0906 I0 0,0406 8 0.0213 6 0.0100
6 9 15 0.0905 12 0.0440 i0 0.0248 7 0.0088
6 I0 17 0.0903 14 0.0467 II 0.0210 8 0.0080
6 II 19 0.0901 16 0.0491 13 0.0238 9 0.0073
6 12 21 0,0899 17 0.0415 14 0.0207 II 0,0091
6 13 23 0,0896 19 0.0437 16 0.0231 12 0.0084
6 14 25 0.0893 21 0.0457 17 0.0204 13 0.0077
6 15 27 0.0890 23 0.0474 19 0.0224 15 0.0092
7 4 6 0.0818 4 0.0364 3 0.0212 1 0.0061
7 5 8 0.0745 6 0.0366 5 0,0240 3 0.0088
7 6 Ii 0.0903 8 0.0367 6 0.0175 4 0.0070
7 7 13 0.0825 II 0.0487 8 0.0189 6 0.0087
7 8 16 0.0946 13 0.0469 I0 0.0200 7 0.0070
7 9 18 0.0869 15 0.0454 12 0.0209 9 0.0082
7 I0 21 0.0966 17 0.0439 14 0o0215 II 0.0093
7 II 23 0.0896 19 0.0427 16 0.0221 12 0.0077
7 12 26 0.0978 21 0.0416 18 0.0225 14 0.0085
7 13 28 0.0913 24 0.0484 20 0.0228 16 0.0093
7 14 31 0,0984 26 0.0469 22 0.0230 17 0.0079
7 15 33 0.0926 28 0.0455 24 0.0233 19 0.0086
8 5 I0 0.0855 8 0.0466 6 0.0225 4 0.0093
8 6 13 0.0906 I0 0.0406 8 0.0213 6 0o0100
8 7 16 0.0946 13 0.0469 i0 0.0200 7 0.0070
8 8 19 0.0974 15 0.0415 13 0.0249 9 0.0074
8 9 22 0.0998 18 0.0464 15 0.0232 II 0.0076
8 I0 24 0.0864 20 0.0416 17 0.0217 13 0.0078
8 II 27 0.0887 23 0.0454 19 0.0204 15 0.0079
8 12 30 0.0907 26 0.0489 22 0.0237 17 0.0079
8 13 33 0.0923 28 0.0445 24 0.0223 20 0.0099
8 14 36 0.0938 31 0.0475 26 0.0211 22 0.0098
8 15 39 0.0951 33 0.0437 29 0.0237 24 0.0097

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions a
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 69

TABLE 1 2 - - M a n n and Whitney's U statistic (continued).

tt. u~ tl~ ua
M~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 6 15 0.0905 12 0.0440 I0 0.0248 7 0.0088

9 7 18 0,0869 15 0.0454 12 0,0209 9 0.0082
9 8 22 0,0998 18 0.0464 15 0.0232 ii 0.0076
9 9 25 0,0951 21 0.0470 17 0~ 14 0.0094
9 I0 28 0.0912 24 0.0474 20 0.0217 16 0.0086
9 II 31 0.0877 27 0,0476 23 0.0232 18 0.0079
9 12 35 0.0965 30 0.0477 26 0.0245 21 0.0092
9 13 38 0.0929 33 0.0478 28 0.0217 25 0.0085
9 14 41 0.0897 36 0.0478 31 0.0228 26 0.0096
9 15 45 0.0968 39 0.0478 34 0.0238 28 0.0089
i0 7 21 0.0966 17 0.0439 14 0,0215 II 0.0093
I0 8 24 0.0864 20 0.0416 17 0.0217 13 0.0078
i0 9 28 0.0912 24 0.0474 20 0.0217 16 0.0086
i0 I0 32 0.0952 27 0.0446 23 0o0216 19 0.0093
i0 ii 36 0.0986 31 0.0493 26 0.0215 22 0.0098
I0 12 39 0.0901 34 0.0465 29 0.0213 24 0.0084
I0 13 45 0.0931 37 0.0441 33 0.0247 27 0.0089
I0 14 47 0.0958 41 0.0478 36 0.0242 30 0.0093
I0 15 51 0.0981 44 0.0455 39 0.0238 33 0~
Ii 8 27 0.0887 23 0.0454 19 0.0204 15 0.0079
II 9 31 0.0877 27 0.0476 23 0.0232 18 0.0079
II I0 36 0.0986 31 0.0493 26 0.0215 22 0.0098
Ii II 40 0.0966 34 0.0440 30 0.0237 25 0.0096
ii 12 44 0.0948 38 0.0454 33 0.0219 28 0.0094
Ii 13 48 0.0931 42 0.0467 37 0.0237 31 0.0092
Ii 14 52 0.0916 46 0.0477 40 0.0221 34 0.0090
II 15 57 0.0991 50 0.0486 44 0.0236 37 0.0088
12 9 35 0.0965 30 0.0477 26 0.0245 21 0.0092
12 I0 39 0.0901 34 0.0465 29 0.0213 24 0.0084
12 II 44 0.0948 38 0.0454 35 0.0219 28 0.0094
12 12 49 0.0989 42 0.0444 37 0.0225 31 0.0086
12 iB 53 0.0930 47 0.0488 41 0.0229 35 0,0094
12 14 58 0.0966 51 0.0475 45 0.0232 38 0.0087
12 15 63 0.0998 55 0.0468 49 0.0254 42 0.0093
13 i0 43 0.0931 37 0.0441 55 0.0247 27 0.0089
13 II 48 0,0931 42 0,0467 37 0.0237 31 0.0092
13 12 53 0.0930 47 0.0488 41 0.0229 35 0.0094
IB 13 58 0.0928 51 0.0454 45 0.0221 39 0.0095
13 14 63 0.0926 56 0.0472 50 0.0241 43 0.0097
13 15 68 0.0923 61 0.0489 54 0.0232 47 0.0097
14 ii 52 0.0916 46 0.0477 40 0,0221 34 0.0090
14 12 58 0.0966 51 0.0475 45 0.0232 38 0.0087
14 13 63 0.0926 56 0.0472 50 0.0241 43 0.0097
14 14 69 0.0968 61 0.0469 55 0.0249 47 0,0093
14 15 74 0.0930 66 0.0466 59 0.0229 51 0.0089
15 12 63 0.0998 55 0,0463 49 0.0234 42 0.0093
15 13 68 0.0923 61 0.0489 54 0.0232 47 0.0097
15 14 74 0.0930 66 0,0466 59 0.0229 51 0.0089
15 15 80 0.0936 72 0.0488 64 0,0227 56 0.0093

N o t e for Table 12.

For
l<Mu<15
[2,Mu - 3]m~x < Nx < 15
where
U = count of all X, Y pairs in which a Y precedes an X. The smaller the value
of U, the less likely it would occur if Y = X,
Mu = size of Y (test) sample, and
N~ = size of X reference sample.
NOTE: When the statistic tabulated has no value such that a significance test is
possible at the desired level, this condition is stipulated by the probability entry 0.0.

Consider the null hypothesis Y = X relative to the alternative hypothesis

Y>X.

Answer--No analysis is necessary. Visual inspection of the data suffices

to indicate that we would not reject the hypothesis Y = X in favor of the
alternative hypothesis Y > X. (But note that the alternative hypothesis
X > Y looks attractive.)
2. Suppose the following data are observed

X: 112, 143, 177, 151,231,345 (N~ -- 6)

Y: 212, 329, 500(DNF), 192, 401 (M~ = 5)

Consider the null hypothesis Y = X relative to the alternative hypothesis

Y>X.

Answer- Visual inspection does not suffice in this case. Thus, we first
order each set of data from smallest to largest

X: 112, 143, 151, 177, 231,345

Y: 192, 212, 329, 4 0 1 , 5 0 0 ( D N F )

Next, we order the entire collection of data, separating the run-outs from
the life data but retaining the X, Y identity of the individual datum.
Life Data
112, 143, 151, 177, 192, 212, 231,329, 345, 401
X X X X Y Y X Y X Y
Suspended Tests
X's: 0
Y's: 1

Now, the test statistic used to refer to statistical tables must be defined for
these data.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 71

Halperin's UH statistic is appropriate for a test program whose individual

tests are terminated as soon as the specimen either fails or is a run-out
(namely, for Type I censoring), Table 13.

Halperin' s UH Statistic
Halperin [25] devised a modification of Mann and Whitney's U statistic
which accounts for certain individual tests being suspended as soon as the
specimen has survived a pre-established duration, say N* cycles. He shows
that the limiting form of the distribution of UII is normal if My and N~ go
to infinity in any arbitrary manner. The power of the UH statistic is con-
sidered in R e f 26.
Halperin's Un statistic is defined as

UH = UL(life data) + Uc(eensored data)

in which UL is M a n n and Whitney's U statistic for the observed life data,

and Uc is the term used to account for the censored data. Uc is computed
as follows
Uc = J~(Nx -- J~)

where Jr and Jy are the numbers of failed X and Y specimens, respectively,

and (Nx - Jx) and (M u -- J~) are the corresponding numbers of run-outs.
The total number of run-outs is denoted RH.

Example--Given Nx = 6 and M~ = 5. Suppose the test outcome stated

in terms of the combined ordered sample is

XXXX YY X Y X Y (Y)DNF

The observed value of U~t, termed u~,, is 5 + 0 = 5. (The life data con-
tributes a count of 5 to uh; and the suspended data contributes a count of
zero, namely, 4(6 - 6) = 0.)

Example--Given Nx = 15 and M,, = 4. Suppose the test outcome is

XXXX Y XXXXXX Y (5 X's, 2 Y's)DNF

In this case, uh = 6 + 2(5) = 16 (RH = 7).

Halperin's Tables
Table 13 consists of eleven columns. The first three identify My, N~, and
RH, respectively. The last eight columns consist of four pairs of columns,
each pair denoted uh and Prob. The four uh columns state the critical
significance values of (ut,), for which (if the null hypothesis is true and
given RH)
Prob[UH < (uh).] _-< c~

CopyrightbyASTMInt'l(allrightsreserved);SunJan 320:27:44EST2016
Downloaded/printedby
UniversityofWashington(UniversityofWashington)pursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.
72 MANUAk ON STATISTICAL PLANNING AND ANALYSIS

where a, the size of the Type I error, is equal to 0.10, 0.05, 0.025, and 0.01,
respectively. The four adjacent columns denoted Prob state the exact size
of the significance test for the corresponding values of (Uh),.
Given the tables and the values of RH and Uh, the appropriate significance
test of size c~ is
reject the null hypothesis if uh < (Uh),, or
"accept" the null hypothesis if Uh > (Uh),

E x a m p l e - - G i v e n the outcome X X X X Y Y X Y X Y ( Y ) D N F , consider

the statistical significance at the 0.10 level.

A n s w e r - - F o r My = 5 and N~ = 6, with RH = 1, Table 13 indicates a

value of UH equal to 5 or smaller would occur 0.0411 • 100 percent of the
time by chance if the populations were indeed identical. This probability
is so small that we usually opt to reject the null hypothesis and believe
instead that Y is better than X. (In this example, UH could be as large as 7,
and we would still reject the null hypothesis at the 0.10 level with exact
probability equal 0.0887.)

E x a m p l e - - G i v e n the outcome X X X X Y X X X X X X Y(5 X's, 2 Y's)DNF,

consider the statistical significance at the 0.10 level.

A n s w e r - - F o r M u = 4, Nx = 15, and RH = 7, examination of Table 13

indicates that a value of UH equal to 1l or smaller is regarded as sig-
nificant statistically at the 0.10 level (exact size 0.0921). Thus, the observed
value uh = 16 does not support a decision to reject the null hypothesis.

Acceh, rated Compartsons

Case 2--Suppose to speed the decision making process it is decided to

terminate the overall test program as soon as the ruth Y failure has occurred,
regardless of the time at which it occurs. In this case, the overall data
consist of the ordered life data and the numbers of suspended X and Y
specimens.
1. Given Nx = 8, My = 4, and r~ = 3, suppose the following data are
observed
Y: 112, 143, 151(= y3), (151 +) suspended
X: (8 X's 151 +) suspended

Consider the null hypothesiz Y = X relative to the alternative hypothesis

Y>X.

A n s w e r - - N o analysis is necessary. Visual inspection of the data suffices

to indicate that we would not reject the hypothesis Y = X in favor of the
alternative hypothesis Y > X. (But note that the alternative hypothesis
X > Y looks attractive.) (Text continues on p. 112.)

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAl TESTS 73

T A B L E 13--Halperin's U~ statistic."

Uha ah= Uha U~

Mu N~ Ru 010 Prob 0.05 Prob 0.025 Prob 0.01 Prob

1 9 0 0 0.1000 0 0.0 0 0.0 0 0.0

1 9 1 0 0.1000 0 0.0 0 0.0 0 0.0
I I0 0 0 0.0909 0 0.0 0 0.0 0 0.0
I i0 i 0 0.0909 0 0.0 0 0.0 0 0.0
I II 0 0 0.0833 0 0.0 0 0.0 0 0.0
I ii I 0 0.0833 0 0.0 0 0.0 0 0.0
I 12 0 0 0.0769 0 0.0 0 0.0 0 0.0
1 12 1 0 0.0769 0 0.0 0 0.0 0 0.0
1 13 0 0 0.0714 0 0.0 0 0.0 0 0.0
I 13 i 0 0.0714 0 0.0 0 0.0 0 0.0
i 14 0 0 0.0667 0 0.0 0 0,0 0 0.0
i 14 I 0 0.0667 0 0.0 0 0.0 0 0.0
I 15 0 0 0.0625 0 0.0 0 0.0 0 0.0
i 15 1 0 0.0625 0 0.0 0 0.0 0 0.0
2 3 0 0 0.1000 0 0.0 0 0.0 0 0.0
2 3 I 0 0. I000 0 0.0 0 0.0 0 0.0
2 3 2 0 0.I000 0 0.0 0 0.0 0 0.0
2 4 0 0 0.0667 0 0.0 0 0.0 0 0.0
2 4 1 0 0.0667 0 0.0 0 0.0 0 0.0
2 4 2 0 0.0667 0 0.0 0 0.0 0 0.0
2 5 0 1 0.0952 0 0.0476 0 0.0 0 0.0
2 5 I 1 0.0952 0 0.0476 0 0.0 0 0.0
2 5 2 0 0.0476 0 0.0476 0 0.0 0 0,0
2 6 0 1 0.0714 0 0.0357 0 0.0 0 0.0
2 6 I I 0.0714 0 0.0357 0 0.0 0 0.0
2 6 2 0 0.0357 0 0.0357 0 0.0 0 0.0
2 7 0 1 0.0556 0 0.0278 0 0.0 0 0.0
2 7 1 I 0.0556 0 0.0278 0 0.0 0 0.0
2 7 2 1 0.0833 0 0.0278 0 0.0 0 0.0
2 7 3 I 0.0833 0 0.0 0 0.0 0 0.0
2 8 0 2 0.0889 i 0.0444 0 0.0222 0 0.0
2 8 I 2 0.0889 i 0.0444 0 0.0222 0 0.0
2 8 2 1 0.0667 0 0.0222 0 0.0222 0 0.0
2 8 3 1 0.0667 0 0.0 0 0.0 0 0.0
2 9 0 2 0.0727 I 0.0364 0 0.0182 0 0.0
2 9 I 2 0.0727 i 0.0364 0 0.0182 0 0.0
2 9 2 2 0.0909 0 0.0182 0 0.0182 0 0.0
2 9 3 I 0.0545 0 0.0 0 0.0 0 0.0
2 i0 0 3 0.0909 I 0.0303 0 0.0152 0 0.0
2 I0 1 3 0.0909 I 0.0303 0 0.0152 0 0.0
2 I0 2 2 0.0758 1 0.0455 0 0.0152 0 0.0
2 I0 3 2 0.0909 i 0.0455 0 0,0 0 0.0
2 I0 4 2 0.0909 0 0.0 0 0.0 0 0.0
2 II 0 3 0.0769 I 0.0256 0 0.0128 0 0o0
2 II 1 3 0.0769 1 0.0256 0 0.0128 0 0.0
2 ii 2 3 0.0897 I 0.0385 0 0.0128 0 0.0
2 ii 3 2 0.0769 I 0.0385 0 0.0 0 0.0
2 II 4 2 0.0769 0 0.0 0 0.0 0 0.0
2 12 0 4 0.0989 2 0.0440 I 0.0220 0 0.0

=See note at end o f table, p 111

TABLE 13--Halperin's Ua statistic (continued).

Uha Uha Uha Hha

M~ Nx RH 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

2 12 1 4 0.0989 2 0.0440 1 0.0220 0 0.0

2 12 2 3 0.0789 1 0.0330 0 0.0110 0 0.0
2 12 3 3 0.09A9 I 0.0330 0 0.0 0 0.0
2 12 4 2 0.0659 0 0.0 0 0.0 0 0.0
2 13 0 4 0.0857 2 0.0381 I 0.0190 0 0.0095
2 13 I 4 0.0857 2 0.0381 I 0.0190 0 0.0095
2 13 2 4 0.0952 2 0.047(~ 0 0.0095 0 0.0095
2 13 3 3 0.0857 1 0.0286 0 0.0 0 0.0
2 13 4 3 0.0952 0 0.0 0 0.0 0 0.0
2 13 5 3 0.0952 0 0.0 0 0.0 0 0.0
2 14 0 5 0.1000 3 0.0500 1 0.0167 0 0.0083
2 14 1 5 0.1000 3 0.0500 I 0.0167 0 0.0083
2 14 2 4 0.0833 2 0.0417 1 0.0250 0 0.0083
2 14 3 4 0.1000 2 0.0500 I 0.0250 0 0.0
2 14 4 3 0.0833 2 0.0500 0 0.0 0 0.0
2 14 5 3 0.0833 0 0.0 0 0.0 0 0.0
2 15 0 5 0.0882 3 0.0441 1 0.0147 0 0.0074
2 15 I 5 0,0882 3 0.0441 1 0.0147 0 0.0074
2 15 2 5 0.0956 2 0.0368 I 0.0221 0 0.0074
2 15 3 4 0.0882 2 0.0441 I 0.0221 0 0.0
2 15 4 3 0.0735 2 0.0441 0 0.0 0 0.0
2 15 5 3 0.0735 0 0.0 0 0.0 0 0.0
3 2 0 0 0.1000 0 0.0 0 0.0 0 0.0
3 2 i 0 0. I 0 0 0 0 0.0 0 0.0 0 0.0
3 2 2 0 0.1000 0 0.0 0 0.0 0 0.0
3 2 3 0 0.I000 0 0.0 0 0.0 0 0.0
3 3 0 I 0.I000 0 0.0500 0 0.0 0 0.0
3 3 1 I 0.1000 0 0.0500 0 0.0 0 0.0
3 3 2 1 0.1000 0 0.0500 0 0.0 0 0.0
3 3 3 0 0.0500 0 0.0500 0 0.0 0 0.0
3 4 0 1 0.0571 0 0.0286 0 0.0 0 0.0
3 4 I I 0.0571 0 0.0286 0 0.0 0 0.0
3 4 2 1 0.0571 0 0.0286 0 0.0 0 0.0
3 4 3 0 0.0286 0 0.0286 0 0.0 0 0.0
3 5 0 2 0.0714 1 0.0357 0 0.0179 0 0.0
3 5 i 2 0.0714 1 0.0357 0 0.0179 0 0.0
3 5 2 2 0.0893 I 0.0357 0 0.0179 0 0.0
3 5 3 I 0.0714 0 0.0179 0 0.0179 0 0.0
3 5 4 I 0.0714 0 0.0 0 0.0 0 0.0
3 6 0 3 0.0833 2 0.0476 I 0.0238 0 0.0
3 6 I 3 0.0833 2 0.0476 1 0.0238 0 0.0
3 6 2 3 0.0952 1 0.0238 1 0.0238 0 0.0
3 6 3 2 0.0833 I 0.0476 0 0.0119 0 0.0
3 6 4 I 0.0476 1 0.0476 0 0.0 0 0.0
3 7 0 4 0.0917 2 0.0333 1 0.0167 0 0.0083
3 7 1 4 0.0917 2 0.0333 1 0.01(',7 0 0.0083
3 7 2 3 0.0667 2 0.0417 I 0.0167 0 0.00~5
3 7 3 3 0.0833 1 0.0333 0 0.0083 0 0.0085
3 7 4 2 0.0833 1 0.0333 0 0.0 0 0.0

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 75

TABLE 13--Halperin's U a statistic (continued).

U hot ~lh a ll hct lib a

Mu N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

3 7 5 2 0.0R33 0 0.0 0 0.0 0 0.0

3 8 0 5 0.0970 3 0.0424 2 0.0242 0 0.0041
3 ~ i 5 0.0970 3 0.0424 2 0.0242 00.OO&1
3 8 2 40.07RP 3 0.04~5 1 0.0121 00.OOAL
3 R 3 4 0.0970 2 0.0424 I 0.0242 0 0.0061
3 R 4 3 0.0970 i 0.0242 I 0.0242 0 0.0
3 8 5 2 0.0606 0 0.0 0 0.0 0 0.0
3 9 0 5 0.0727 4 0.0500 2 0.0182 1 0.0091
3 9 i 5 0.0727 4 0.0500 2 0.0182 i 0.0091
3 9 2 5 0.0818 3 0.0364 2 0.0227 1 0.0091
3 9 3 5 0.i000 3 0.0455 i O.OIR2 0 0.0045
3 9 4 4 0. i 0 0 0 2 0.0455 I 0.0182 0 0.0
3 9 5 3 0.0909 2 0.0455 0 0.0 0 0.0
3 9 6 3 0.0909 0 0.0 0 0.0 0 0.0
3 i0 0 6 0.0804 4 0.0385 3 0.0245 1 0.0070
3 10 i 6 0.0R04 4 0.0385 3 0.0245 1 0.0070
3 i0 2 6 0.0909 4 0.0455 2 0.0175 1 0.0070
3 10 3 5 0.0769 3 0.0350 2 0.0245 0 0.0035
3 i0 4 5 0.0979 2 0.0550 1 0.0140 0 0.0
3 i0 5 3 0.0699 2 0.0350 0 0.0 0 0.0
3 lO 6 3 0.0699 0 0.0 0 0.0 0 0.0
3 ii 0 7 0.0A52 5 0.0440 3 0.0192 1 0.0055
3 ii i 7 0.0R52 5 0.0440 3 0.0192 1 0.0055
3 Ii 2 7 0.0934 5 0.0495 3 0.0220 i 0.0055
3 ii 3 6 0.0852 4 0.0440 2 0.0192 0 0.0027
3 ii 4 5 0.0769 3 0.0440 1 0.0110 0 0.0
3 II 5 4 0.0824 2 0.0275 0 0.0 0 0.0
3 ii 6 4 0.0962 0 0.0 0 0.0 0 0.0
3 ii 7 4 0.0962 0 0.0 0 0.0 0 0.0
3 12 0 8 0.0901 5 0.0352 4 0.0242 20.OOSS
3 12 1 8 0.0901 5 0.0352 4 0.0242 2 0.00R8
3 12 2 8 0.09R9 5 0.0396 3 0.0176 1 0.0044
3 12 3 7 0.0879 5 0.04S4 3 0.0220 1 0.00R8
3 12 4 6 0.0835 4 0.0484 2 0.0220 1 0.0088
3 12 5 5 0.0879 3 0.0640 2 0.0220 0 0.0
3 12 6 4 0.0769 3 0.0440 0 0.0 0 0.0
3 12 7 4 0.07A9 0 0.0 0 0.0 0 0.0
3 13 0 9 0.0946 6 0.0411 4 0.0196 2 0.0071
3 13 i 9 0.0946 6 0.0411 4 0.0196 2 0.0071
3 13 2 R 0.0804 6 0.0464 4 0.0232 20.O()S9
3 13 3 8 0.0929 5 0.0393 3 0.0179 1 0.0071
3 13 4 7 0.0H57 5 0.0500 2 0.0179 1 0.0071
3 13 5 6 0.0893 3 0.0357 2 0.0179 0 0.0
3 13 6 5 0.0893 3 0.0357 0 0.0 0 0.0
3 13 7 5 0.I000 0 0.0 0 0.0 0 0.0
3 13 8 5 0.1000 0 0.0 0 0.0 0 0.0
3 14 0 I0 0 . 0 9 8 5 7 0.0456 5 0.0235 2 0.0059
3 14 1 I0 0 . 0 9 8 5 7 0.0456 5 0.0235 2 0.0059
3 14 2 9 0.0~38 7 0.0500 4 0.0191 2 0.0074

TABLE 13--Halperin' s Un statistic (continued).

Uh= Ith~ Uh= Uh~

Mu N~ Rn 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

3 14 3 9 0.0956 6 0.0456 4 0.0235 1 0.0059

3 14 4 8 0.0912 5 0.0412 3 0.0235 1 0.0059
3 14 5 7 0.0882 4 0.0441 2 0.0147 0 0.0
3 14 6 6 0.0956 3 0.0294 0 0.0 0 0.0
3 I4 7 5 0.0824 0 0.0 0 0.0 0 0.0
3 14 8 5 0.0824 0 0.0 0 0.0 0 0.0
3 15 0 I0 0.0821 7 0.0380 5 0.0196 3 0.0086
3 15 i i0 0.0821 7 0.0380 5 0.0196 3 0.0086
3 15 2 i0 0.0882 7 0.0417 5 0.0221 3 0.0098
3 15 3 i0 0.0993 7 0.0490 4 0.0196 2 0.0086
3 15 4 9 0.0931 6 0.0466 3 0.0196 1 0.0049
3 15 5 8 0.0919 5 0.0490 3 0.0245 0 0.0
3 15 6 7 0.0980 4 0.0429 3 0.0245 0 0.0
3 15 7 6 0.0944 4 0.0429 0 0.0 0 0.0
3 15 8 5 0,0686 0 0.0 0 0~ 0 0.0
4 2 0 0 0.0667 0 0.0 0 0.0 0 0.0
4 2 i 0 0.0667 0 0.0 0 0.0 0 0.0
4 2 2 0 0.0667 0 0.0 0 0.0 0 0.0
4 2 3 0 0.0667 0 0.0 0 0.0 0 0.0
4 2 4 0 0,0667 0 0.0 0 0.0 0 0.0
4 3 0 1 0.0571 0 0.0286 0 0.0 0 0.0
4 3 i 1 0.0571 0 0.0286 0 0.0 0 0.0
4 3 2 1 0.0571 0 0.0286 0 0.0 0 0.0
4 3 3 1 0.0571 0 0.0286 0 0.0 0 0.0
4 3 4 0 0.0286 0 0.0286 0 0.0 0 0.0
4 4 0 3 0.1000 i 0.0286 0 0.0143 0 0.0
4 4 i 3 0.1000 1 0.0286 0 0.0143 0 0.0
4 4 2 2 0.0571 1 0.0286 0 0.0143 0 0.0
4 4 3 2 0.0857 1 0.0286 0 0.0143 0 0.0
4 4 4 1 0.0714 0 0.0143 0 0.0143 0 0.0
4 4 5 1 0.0714 0 0.0 0 0.0 0 0.0
4 5 0 4 0.0952 2 0.0317 1 0.0159 0 0.0079
4 5 1 4 0.0952 2 0.0317 1 0.0159 0 0.0079
4 5 2 3 0.0635 2 0.0317 1 0.0159 0 0.0079
4 5 3 3 0.0794 2 0.0476 1 0.0159 0 0.0079
4 5 4 2 0.0714 1 0.0397 0 0.0079 0 0.0079
4 5 5 1 0.0397 1 0.0397 0 0.0 0 0.0
4 6 0 5 0.0857 3 0.0333 2 0.0190 1 0.0095
4 6 I 5 0.0857 3 0.0333 2 0.0190 1 0.0095
4 6 2 5 0.0952 3 0.0381 2 0.0190 1 0.0095
4 6 3 4 0.0R10 3 0.0476 1 0.0095 1 0.0095
4 6 4 3 0.0619 2 0.0429 1 0.0238 0 0.0048
4 6 5 2 0.0714 1 0.0238 1 0.0238 0 0.0
4 6 6 2 0.0714 0 0.0 0 0.0 0 0.0
4 7 0 6 0.0818 4 0.0364 3 0.0212 1 0.0061
4 7 I 6 0.0818 4 0.0364 3 0.0212 1 0.0061
4 7 2 6 0.0909 4 0.0394 3 0.0242 1 0.0061
4 7 3 5 0.0727 3 0.0303 2 0.0182 I 0.0061
4 7 4 5 0.1000 3 0.0594 1 0.0152 0 0.0030

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 77

TABLE 13 H a l p e r i n ' s Ur~ statistic ( c o n t i n u e d ) .

Uha Uh~ Igha Uh=

M, N= RH 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

4 7 5 3 0.0758 2 0.0455 1 0.0152 0 0.0

4 7 6 2 0.0455 2 0.0455 0 0.0 0 0.0
4 8 0 7 0.0768 5 0.0364 4 0.0242 2 0.0081
4 8 1 7 0.0768 5 0.0364 4 0.0242 2 0.0081
4 8 2 7 0.0848 5 0.0404 3 0.0162 2 0.0081
4 8 3 6 0.0747 5 0.0485 30.020Z 1 0.0040
4 8 4 6 0,0990 4 0.0465 2 0,0182 0 0.0020
4 8 5 5 0.0909 2 0.0303 1 0.0101 0 0.0
4 8 A 3 0.0707 2 0.0303 0 0.0 0 0.0
4 8 7 3 0.0707 0 0.0 0 0.0 0 0.0
4 9 0 9 0.0993 6 0.0378 4 0.0168 3 0.0098
4 9 1 9 0.0993 6 0.0378 4 0.0168 3 0.0098
4 9 2 8 0.0811 6 0.0420 4 0.0182 2 0.0056
4 9 3 8 0.0965 5 0.0336 4 0.0238 2 0.0084
4 9 4 7 0.0909 5 0.0462 3 0.0182 1 0.0070
4 9 5 6 0.0909 4 0.0490 2 0.0210 1 0.0070
4 9 6 4 0.0769 3 0.0490 2 0.0210 0 0.0
4 9 7 4 0.0979 3 0.0490 0 0.0 0 0.0
4 9 S 4 0,0979 0 0.0 0 0.0 0 0.0
4 I0 0 i0 0.0939 7 0.0380 5 0.0180 3 0.0070
4 i0 1 I0 0.0939 7 0.0380 50.OiRO 3 0.0070
4 i0 2 9 0.0779 7 0.0420 5 0.0200 3 0.0080
4 10 3 9 0.0909 7 0.0500 5 0.0240 B 0.0100
4 i0 4 8 0.0869 6 0.0490 4 0.0230 2 0.0090
4 I0 5 7 0.0849 5 0.0450 3 0.0250 1 0.0050
4 I0 6 6 0.0949 3 0.0350 2 0.0150 0 0.0
4 I0 7 4 0.0699 3 0.0350 0 0,0 0 0.0
4 i0 8 4 0.0699 0 0.0 0 0.0 0 0.0
4 ii 0 II 0.0886 8 0.0388 6 0.0198 4 0.0088
4 Ii i ii 0.0886 8 0.0388 6 0.0198 4 0.0088
4 Ii 2 ii 0.0960 8 0.0425 6 0.0220 4 0.0095
4 Ii 3 I0 0.0872 7 0.0366 5 0.0176 3 0.0073
4 II 4 9 0.0828 7 0.0476 5 0.0242 3 0.0095
4 ii 5 8 0.0842 6 0.0476 3 0.0183 1 0.0037
4 II 6 7 0.0842 4 0.0405 Z 0.0110 0 0.0
4 ii 7 5 0.0769 3 0.0256 0 0.0 0 0.0
4 II 8 5 0.0923 0 0.0 0 0.0 0 0.0
4 11 9 5 0.0923 0 0.0 0 0.0 0 0.0
4 12 0 12 0.0852 9 0.0590 7 0.0209 5 0.0099
4 12 i 12 0.0852 9 0.0390 7 0.0209 5 0.0099
4 12 2 12 0.0918 9 0.0429 7 0.0231 4 0.0071
4 12 3 Ii 0.0824 9 0.0500 6 0.0203 4 0.0093
4 12 4 ii 0.0995 8 0.0478 5 0.0181 3 0.0071
4 12 5 9 0.0797 7 0.0467 5 0.0247 2 0.0082
4 12 6 8 0.0824 5 0.0412 3 0.0192 2 0.0082
4 12 7 7 0.0962 4 0.0385 3 0.0192 0 0.0
4 12 8 6 0.I000 4 0.0385 0 0.0 0 0.0
4 12 9 5 0.0692 0 0.0 0 0.0 0 0.0
4 13 0 13 0.0815 i0 0.0395 8 0.0223 5 0.0076

T A B L E 1 3 - - H a l p e r i n ' s U n statistic ( c o n t i n u e d ) .

Uh~ Uh~ biba biba

Mu N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

4 13 1 13 0.0815 I0 0.0395 8 0.0223 5 0.0076

4 13 2 13 0.0874 I0 0.0429 8 0.0244 5 0.0084
4 13 3 13 0.0987 I0 0.0500 7 0.0210 4 0.0071
4 13 4 12 0.0954 9 0.0475 6 0.0206 4 0.0097
4 13 5 ii 0.0945 8 0,0483 5 0.0189 2 0.0063
4 13 6 I0 0.0987 7 0.0483 4 0.0231 2 0.0063
4 13 7 8 0.0882 5 0.0441 3 0.0147 0 0.0
4 13 8 7 0.I000 4 0.0294 0 0.0 0 0.0
4 13 9 6 0.0~82 0 0.0 0 0.0 0 0.0
4 13 I0 6 0.0882 0 0.0 0 0.0 0 0.0
4 14 0 15 0.0961 II 0.0395 9 0.0232 6 0.0088
4 14 1 15 0.0961 Ii 0.0395 9 0.0232 6 0.0088
4 14 2 14 0,0840 II 0,0428 8 0,0190 6 0.0098
4 14 3 14 0.0944 II 0.0490 8 0.0225 5 0.0078
4 14 4 13 0.0905 I0 0.0474 7 0.0212 4 0.0075
4 14 5 12 0.0915 9 0,0474 6 0.0212 3 0.0082
4 14 6 Ii 0.0931 8 0~ 5 0.0245 2 0.0049
4 14 7 I0 0.0984 6 0.0458 4 0.0229 0 0.0
4 14 8 8 0.0961 5 0.0412 4 0.0229 0 0.0
4 14 9 7 0,0961 5 0.0412 0 0.0 0 0.0
4 14 i0 6 0,0686 0 0,0 0 0.0 0 0.0
4 15 0 16 0.0924 12 0.0400 i0 0.0243 7 0.0098
4 15 I 16 0.0924 12 0.0400 I0 0.0243 7 0.0098
4 15 2 16 0.0978 12 0.0431 9 0.0201 6 0.0077
4 15 3 15 0.0903 12 0.0493 9 0.0235 6 0.0095
4 15 4 14 0.0869 ii 0.0467 8 0.0224 5 0.0085
4 15 5 13 0.0864 i0 0.0477 7 0.0219 4 0.0090
4 15 6 12 0.0903 9 0.0477 6 0.0245 3 0.0090
4 15 7 ii 0.0921 7 0.0451 4 0.0181 3 0.0090
4 15 8 9 0,0903 6 0.0470 4 0.0181 0 0.0
4 15 9 8 0.0975 5 0.0325 0 0.0 0 0.0
4 15 10 7 0.0~51 0 0.0 0 0.0 0 0.0
4 15 II 7 0.0851 0 0.0 0 0.0 0 0.0
5 2 0 1 0.0952 0 0.0476 0 0.0 0 0.0
5 2 1 1 0.0952 0 0.0476 0 0.0 0 0.0
5 2 2 1 0.0952 0 0.0476 0 0.0 0 0.0
5 2 3 1 0.0952 0 0.0476 0 0,0 0 0.0
5 2 4 1 0.0952 0 0.0476 0 0.0 0 0.0
5 2 5 0 0.0476 0 0.0476 0 0.0 0 0.0
5 3 0 2 0.0714 i 0.0357 0 0.0179 0 0.0
5 3 1 2 0.0714 1 0.0357 0 0.0179 0 0.0
5 3 2 2 0.0714 1 0.0357 0 0.0179 0 0.0
5 3 3 2 0.0714 i 0.0357 0 0.0179 0 0.0
5 3 4 1 0.0357 i 0.0357 0 0.0179 0 0.0
5 3 5 0 0.0179 0 0,0179 0 0.0179 0 0.0
5 4 0 4 0.0952 2 0.0317 1 0.0159 0 0.0079
5 4 I 4 0.0952 2 0o0317 i 0o0159 0 0.0079
5 4 2 3 0.0556 2 0.0317 i 0.0159 0 0.0079
5 4 3 3 0.0714 2 0.0317 1 0.0159 0 0.0079

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 79

TABLE 13--Halperin's Ua statistic (continued).

Id ha l~ hrt ld hot tt hct

M~ N~ Ru 0.10 Prob 0.05 Prob 0.025 Ptob 0.01 Prob

5 4 4 3 0.0952 I 0.0159 I 0.0159 0 0.0079

5 4 5 2 0.0873 I 0.0476 0 0.0079 0 0.0079
5 4 6 1 0.0476 1 0.0476 0 0.0 0 0.0
5 5 0 5 0.0754 4 0.0476 2 0.0159 1 0.0079
5 5 1 5 0.0754 4 0.0476 2 0o0159 1 0.0079
5 5 2 5 0.0794 3 0.0278 2 0.0159 1 0.0079
5 5 3 5 0.0952 3 0.0357 2 0.0159 1 0.0079
5 5 4 4 0.0853 3 0.0476 I 0.0079 i 0.0079
5 5 5 3 0.0635 2 0.0437 1 0.0238 0 0.0040
5 5 6 2 0.0~33 1 0.0238 1 0.0238 0 0.0
5 5 7 2 0.0833 0 0.0 0 0.0 0 0.0
5 6 0 7 0.0887 5 0.0411 3 0.0152 2 0.0087
5 6 I 7 0.0887 5 0.0411 3 0.0152 2 0.0087
5 6 2 7 0.0952 5 0.0433 3 0.0152 2 0.00R7
5 6 3 6 0.0801 4 0.0325 3 0.0195 2 0.00R7
5 6 4 5 0.0649 4 0.0455 2 0.0152 1 0.0043
5 6 5 5 0.0996 3 0.0346 2 0.0238 0 0.0022
5 6 6 3 0.0779 2 0.0455 1 0.0130 0 0.0
5 6 7 2 0.0455 2 0.0455 0 0.0 0 0.0
5 7 0 8 0.0745 6 0.0366 5 0.0240 3 0.0088
5 7 1 8 0.0745 6 0,0366 5 0.0240 3 0,0088
5 7 2 8 0.0808 6 0.0391 4 0.0164 3 0.0088
5 7 3 8 0.0947 6 0.0467 4 0.0189 2 0.0051
5 7 4 7 0.0859 5 0.0379 3 0.0152 2 0.0088
5 7 5 6 0.0896 4 0.0391 3 0.0202 1 0.0076
5 7 6 5 0.0833 3 0.0455 1 0.0076 1 0.0076
5 7 7 3 0.0707 2 0.0265 0 0.0 0 0.0
5 7 8 3 0.0707 0 0.0 0 0.0 0 0.0
5 8 0 i0 0.0855 R 0.0466 6 0.0225 4 0.0093
5 8 1 i0 0.0855 8 0.0466 6 0.0225 4 0.0093
5 8 2 10 0.0925 7 0.0350 6 0.0241 3 0.0054
5 8 3 9 0.0808 7 0.0412 5 0.0186 3 0.0070
5 8 4 8 0.0754 6 0.0381 5 0.0233 3 0.0093
5 8 5 7 0.0746 5 0.0357 4 0.0241 2 0.0085
5 8 6 6 0.0785 4 0.0396 2 0.0163 1 0.0047
5 8 7 5 0.0979 3 0.0435 2 0.0163 0 0.0
5 8 8 4 0.0979 3 0.0435 0 0.0 0 0.0
5 8 9 4 0.0979 0 0.0 0 0.0 0 0.0
5 9 0 12 0.0949 9 0.0415 7 0.0210 5 0.0095
5 9 1 12 0.0949 9 0.0415 7 0.0210 5 0.0095
5 9 2 ii 0.0789 9 0.0445 7 0.0225 50.OlO0
5 9 3 ii 0.0914 8 0.0380 6 0.0185 4 0.0075
5 9 4 i0 0.0879 8 0.0485 6 0.0245 3 0.0060
5 9 5 9 0.0879 7 0.0480 5 0.0230 3 0.0080
5 9 6 8 0.0954 5 0.0330 3 0.0180 1 0.0030
5 9 7 7 0.0979 4 0.0455 2 0.0105 0 0.0
5 9 8 5 0.0979 3 0.0280 0 0.0 0 0.0
5 9 9 4 0.0629 0 0.0 0 0.0 0 0.0
5 i0 0 13 0.0823 II 0.0496 8 0.0200 6 0.0097

TABLE 13--Halperin's UR statistic (continued).

Hha Hha gha Hha

M~ N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

5 i0 I 13 0.0823 II 0.0496 8 0.0200 6 0.0097

5 I0 2 13 0.0882 I0 0.0406 8 0.0216 5 0.0067
5 i0 3 12 0.0803 I0 0.0470 7 0.0176 5 0.0080
5 i0 4 12 0.0976 9 0.0440 7 0.0226 5 0.0100
5 I0 5 II 0.0969 8 0.0436 6 0.0236 3 0.0053
5 I0 6 9 0.0819 7 0.0453 5 0.0220 2 0.0070
5 I0 7 8 0.0886 5 0.0420 3 0.0186 2 0.0070
5 I0 8 6 0.0886 4 0.0420 3 0.0186 0 0.0
5 i0 9 5 0.0839 4 0.0420 0 0.0 0 0.0
5 i0 I0 5 0.0839 0 0.0 0 0.0 0 0.0
5 ii 0 15 0.0902 12 0.0449 9 0.0190 70. 0096
5 II i 15 0.0902 12 0.0449 9 0.0190 7 0.0096
5 ii 2 15 0.0962 i2 0.0433 9 0.0204 6 0.0071
5 II 3 14 0.0881 II 0.0426 9 0.0240 6 0.0085
5 II 4 13 0.0847 I0 0.0405 8 0.0222 5 0.0069
5 ii 5 12 0.0861 9 0.0403 7 0.0220 4 0.0071
5 II 6 ii 0.0907 R 0.0437 6 0.0231 3 0.0082
5 ii 7 9 0.0769 7 0.0449 4 0.0208 2 0 . 004R
5 ii R 8 0.0929 5 0.0449 3 0.0128 0 0.0
5 11 9 6 0.0865 4 0.0288 0 0.0 0 0.0
5 Ii I0 5 0.0577 0 0.0 0 0.0 0 0.0
5 12 0 17 0.0973 13 0.0409 Ii 0.0242 8 0.0097
5 12 I 17 0.0973 13 0.0409 II 0.0242 8 0.0097
5 12 2 16 0.0853 13 0.0438 10 0.0197 7 0 . 0073
5 12 3 16 0.0957 12 0.0394 i0 0.0228 7 0.0086
5 12 4 15 0.0923 i2 0.0483 9 0.0213 60. 0079
5 12 5 14 0.0939 Ii 0.0478 8 0.0212 5 0.0074
5 12 6 13 0.0971 9 0.0398 7 0.0220 4 0.0082
5 12 7 ii 0.0882 8 0.0430 5 0.0204 3 0.0090
5 12 8 I0 0.0973 6 0.0430 4 0.0204 3 O. 0 0 9 0
5 12 9 7 0.0814 5 0.0407 4 0.0204 0 0.0
5 12 10 6 0.0747 5 0.0407 0 0.0 0 0.0
5 I2 iI 6 0.0747 0 0.0 0 0.0 0 0.0
5 13 0 18 0.0867 15 0.0473 12 0.0230 9 0.0097
5 13 i I~ 0.0867 15 0.0473 12 0.0230 9 0.0097
5 13 2 18 0.0920 14 0.0404 12 0.0247 8 0.0076
5 13 3 17 0.0851 14 0.0459 II 0.0217 8 0.0089
5 13 4 17 0.0990 13 0.0439 i0 0.0207 70. 0079
5 13 5 15 0.0836 12 0.0445 9 0.0205 6 0.0083
5 13 6 14 O.OB71 Ii 0.0462 8 0.0223 5 0.0077
5 13 7 13 0.0964 9 0.0392 7 0.0229 3 0.0065
5 13 8 ii 0.0850 8 0.0474 5 0.0229 3 0.0065
5 13 9 9 0.0882 6 0.0441 4 0.0147 0 0.0
5 13 i0 7 0.0784 5 0.0294 0 0.0 0 0.0
5 13 ii 7 0.0924 0 0.0 0 0.0 0 0.0
5 13 12 7 0.0924 0 0.0 0 0.0 0 0.0
5 14 0 20 0.0930 16 0.0435 13 0.0218 i0 0 . 0 0 9 7
5 I4 i 20 0.0930 16 0.0435 13 O.02iR i0 O. 0 0 9 7
5 14 2 20 0.0981 16 0.0464 i3 0.0234 9 0.0077

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 81

TABLE 13---Halperin's U n statistic (continued),

$t h a U ha bl h a U ha

My N= RH 0.]0 Prob 0.05 Prob 0.025 Prob O . O l Prob

5 14 3 19 0.0912 15 0.0423 12 0 . 0 2 1 0 9 0.0090

5 14 4 18 0.0890 14 0.0407 II 0 . 0 1 9 6 8 0.0083
5 14 5 17 0.0895 13 0.0409 i0 0 . 0 1 9 9 7 0.0083
5 14 6 16 0.0934 12 0.0435 9 0.0212 6 0.0087
5 14 7 15 0.0999 ii 0.0470 8 0.0229 4 0.0078
5 14 8 13 0.0939 9 0.0409 6 0.0229 3 0.0048
5 14 9 ii 0.0867 7 0.0433 5 0.0217 0 0.0
5 14 I0 9 0.0939 6 0.0397 5 0.0217 0 0.0
5 14 Ii 8 0.0965 6 0.0397 0 0.0 0 0.0
5 14 12 7 0.0681 0 0.0 0 0.0 0 0.0
5 15 0 22 0.0987 18 0.0491 14 0 . 0 2 0 9 Ii 0 . 0 0 9 7
5 15 1 22 0.0987 18 0.0491 14 0 . 0 2 0 9 11 0 . 0 0 9 7
5 15 2 21 0.0884 17 0.0428 14 0 . 0 2 2 4 I0 0 . 0 0 7 9
5 15 3 21 0.0971 17 0.0481 13 0 . 0 2 0 1 i0 0 . 0 0 9 1
5 15 4 20 0.0945 16 0.0466 13 0 . 0 2 4 3 9 0.0085
5 15 5 19 0.0949 15 0.0468 i2 0 . 0 2 4 6 8 0.0084
5 15 6 18 0.0994 14 0,0491 IO 0 . 0 2 0 7 7 0.0088
5 15 7 16 0.0898 12 0.0442 9 0.0217 5 0.0081
5 15 8 14 0.0858 ii 0.0470 7 0.0217 4 0.0081
5 15 9 13 0.0921 9 0.0488 6 0.0244 4 0.0081
5 15 I0 ii 0.0975 7 0.0433 5 0.0163 0 0.0
5 15 ii 9 0.0937 6 0.0298 00.0 00.0
5 15 12 8 0.0830 0 0.0 0 0.0 0 0.0
5 15 13 8 0.0830 0 0.0 0 0.0 0 0.0
6 3 0 3 0.0833 2 0.0476 1 0.0238 0 0.0
6 3 1 3 0.0833 2 0.0476 1 0.0238 0 0.0
6 3 2 3 0.0833 2 0.0476 i 0.0238 0 0.0
6 3 3 3 0.0833 2 0.0476 i 0.0238 0 0.0
6 3 4 2 0.0476 2 0.0476 1 0.0238 0 0.0
6 3 5 2 0.0952 I 0.0238 i 0.0238 0 0.0
6 3 6 I 0.0833 0 0.0119 0 0.0119 0 0.0
6 3 7 1 0.0833 0 0.0 00.0 00.0
6 4 0 5 0.0857 3 0.0333 2 0.0190 I 0.0095
6 4 I 5 0.0857 3 0.0333 2 0.0190 1 0.0095
6 4 2 5 0.0905 3 0.0333 2 0.0190 i 0.0095
6 4 3 5 0.i000 3 0.0333 2 0.0190 I 0.0095
6 4 4 4 0.0810 3 0.0476 2 0.0190 1 0.0095
6 4 5 3 0.0667 2 0.0381 I 0.0095 I 0.0095
6 4 6 3 0.0905 1 0.0333 0 0.0048 0 0.0048
6 4 7 1 0.0333 1 0.0333 00.0 00.0
6 5 0 7 0.0887 5 0.0411 3 0.0152 2 0.0087
6 5 I 7 0.0887 5 0.0411 3 0.0152 2 0.0087
6 5 2 7 0.0931 5 0,0433 3 0.0152 2 0.0087
6 5 3 6 0.0736 5 0.0476 3 0.0152 2 0.0087
6 5 4 6 0,0931 4 0.0368 3 0.0216 2 0.0087
6 5 5 5 0.0779 3 0.0303 2 0.0173 I 0.0043
6 5 6 4 0.0866 3 0.0411 i 0.0152 0 0.0022
6 5 7 2 0.0606 1 0.0152 I 0.0152 0 0.0
6 5 8 2 0.0606 0 0.0 00.0 00.0

TABLE 13--Halperin's Un statistic (continued).

U ha Idha Uha U hct

Mu N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

6 6 0 9 0.0898 7 0.0665 5 0.0206 3 0.0076

6 6 1 9 0,0898 7 0.0465 5 0,0206 3 0.0076
6 6 2 9 0.0952 7 0.0487 5 0.0216 3 0,0076
6 6 3 8 0.0801 6 0.0379 5 0,0238 3 0.0076
6 6 4 8 0.0996 6 0,0476 4 0.0184 2 0.0043
6 6 5 7 0.0942 5 0,0390 3 0.0152 2 0.0087
6 6 6 5 0.0660 4 0,0433 3 0.0206 1 0.0076
6 6 7 5 0,09R5 2 0.0303 1 0.0076 1 0.0076
6 6 8 3 0.0909 2 0.0303 0 0.0 0 0.0
6 6 9 3 0.0909 0 0,0 0 0.0 0 0.0
6 7 0 II 0.0903 8 0.0367 6 0,0175 4 0.0070
6 7 1 Ii 0.0903 R 0.0367 6 0.0175 4 0.0070
6 7 2 II 0,0962 8 0.0385 6 0.0181 4 0.0070
6 7 3 I0 0.0839 8 0,0443 6 0.0204 4 0.0082
6 7 4 9 0.0763 7 0.0373 5 0.0163 4 0.0099
6 7 5 8 0.0746 6 0.0361 5 0.0210 3 0.0082
6 7 6 7 0.0775 5 0.0355 4 0.0233 2 0.0076
6 7 7 6 0.0857 4 0.0408 2 0.0163 1 0.0041
6 7 8 4 O.OR16 3 0.0490 2 0.0163 0 0.0
6 7 9 3 0.0490 3 0.0490 0 0.0 0 0.0
6 8 0 13 0.0906 i0 0.0406 8 0.0213 6 0.0100
6 8 i 13 0.0906 10 0.0406 8 0.0213 6 0.0100
6 8 2 13 0.0962 I0 0.0430 8 0.0223 5 0.0067
6 8 3 12 0.0856 i0 0.0493 7 0.0173 5 0.0073
6 ~ 4 II 0.0793 9 0.0443 7 0.0213 5 0.0093
6 S 5 I0 0.0796 8 0.0430 6 0.0206 4 0.0083
6 8 6 9 0.0829 7 0,0443 5 0.0203 3 0.0063
6 8 7 8 0,0979 6 0.0490 4 0.0233 2 0.0093
6 8 8 6 0.0839 4 0.0466 2 0.0093 2 0.0093
6 8 9 4 0.0699 3 0.0280 0 0o0 0 0.0
6 8 I0 4 0.0699 0 0o0 0 0o0 0 0.0
6 9 0 15 0~ 12 0.0440 I0 0.0248 7 0.0088
6 9 1 15 0.0905 12 0.0440 I0 0.0248 7 0.0088
6 9 2 15 0.0961 12 0.0466 9 0.0190 7 0.0092
6 9 3 14 0.0867 II 0.0400 9 0.0214 6 0.0070
6 9 4 13 0,0819 Ii 0~ 8 0.0190 6 O.OOR8
6 9 5 12 0.0829 I0 0,0490 7 0.0176 5 0.0072
6 9 6 II 0.0869 8 0.0368 6 0.0194 4 O.OORO
6 9 7 I0 0.0979 7 0.0406 5 0.0182 3 0.0098
6 9 8 8 0.0867 5 0.0392 3 0.0168 2 0.0056
6 9 9 6 0,0923 4 0.0420 3 0,0168 0 0.0
6 9 i0 5 0.0923 4 0.0420 0 0.0 0 0.0
6 9 II 5 0.0923 0 0.0 0 0.0 0 0.0
6 I0 0 17 0.0903 14 0,0467 II 0.0210 8 0.0080
6 i0 1 17 0,0903 14 0.0467 ii 0.0210 8 0.0080
6 I0 2 17 0.0957 14 0.0496 II 0.0222 8 0.0084
6 i0 3 16 0.0875 13 0.0437 i0 0.0189 8 0.0096
6 10 4 15 0.0835 12 0.0412 I0 0.0230 7 0.0080
6 i0 5 14 0.0844 II 0.0401 9 0.0225 6 0.0077

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 83

T A B L E 13--Halperin's UH statistic (continued).

Id ha Id hct bl h a lib=
My Nx RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

6 10 6 13 0.0894 10 0.0426 8 0.0230 5 0.0076

6 10 7 12 0.0988 9 0.0481 6 0.0186 4 0.0087
6 I0 8 i0 0.0944 7 0.0385 5 0.0245 2 0.0035
6 I0 9 R 0.0892 5 0.0420 3 0.0105 0 0.0
6 i0 i0 6 0.0892 4 0.0262 0 0.0 0 0.0
6 i0 ii 5 0.0577 0 0.0 0 0.0 0 0.0
6 ii 0 19 0.0901 16 0.0491 13 0.0238 9 0.0073
6 ii I 19 0.0901 16 0.0491 13 0.0238 9 0.0073
6 Ii 2 19 0.0952 15 0.0415 12 0.0193 9 0.0077
6 ii 3 18 0.0878 15 0.0467 12 0.0220 9 0.0086
6 Ii 4 17 0.0846 14 0.0443 ii 0.0200 8 0.0077
6 ii 5 16 0.0858 13 0.0442 10 0.0199 7 0.0071
6 11 6 15 0.0899 12 0.0470 9 0.0206 6 0.0078
6 ii 7 14 0.0995 I0 0.0413 R 0.0238 5 0.0074
6 ii R 12 0.0939 9 0.0452 7 0.0249 3 0.0068
6 ii 9 I0 0.0882 7 0.0475 4 0.0170 3 0.0068
6 ii i0 R 0.0984 5 0.0373 4 0.0170 0 0.0
6 ii ii 6 0.0747 5 0.0373 0 0.0 0 0.0
6 Ii 12 6 0.0747 0 0.0 0 0.0 0 0.0
6 12 0 21 0.0899 17 0.0415 14 0.0207 11 0.0091
6 12 i 21 0.0899 17 0.0415 14 0.0207 Ii 0.0091
6 12 2 21 0.0946 17 0.0440 14 0.0220 ii 0.0096
6 12 3 20 0.0880 17 0.0492 14 0.0249 i0 0.0081
6 12 4 19 0.0850 16 0.0472 13 0.0232 i0 0.0099
6 12 5 18 0.0863 15 0.0473 12 0.0233 9 0.0097
6 12 6 17 0.0901 13 0.0400 ii 0.0241 8 0.0099
6 12 7 16 0.0995 12 0.0449 9 0.0207 6 0.0079
6 12 8 14 0.0920 I0 0.0407 8 0.0234 4 0.0075
6 12 9 12 0.0928 9 0.0452 6 0.0249 3 0.0045
6 12 IO i0 0.0928 6 0.03R5 5 0.0249 0 0.0
6 12 II 8 0.0995 6 0.0498 5 0.0249 0 0.0
6 12 12 7 0.0924 6 0.0498 0 0.0 0 0.0
6 12 13 7 0.0924 0 0.0 0 0.0 0 0.0
6 13 0 23 0.0896 19 0.0437 16 0.0231 12 0.0084
6 13 i 23 0.0896 19 0.0437 16 0.0231 12 O.OOR4
6 13 2 23 0.0941 19 0.0462 16 0.0245 12 0.0088
6 13 3 22 0.0879 18 0.0423 15 0.0218 Ii 0.0075
6 13 4 21 0.0854 18 0.0496 14 0.0205 ii 0.0091
6 13 5 20 0.0864 17 0.0499 13 0.0204 i0 0.0091
6 13 6 19 0.0904 15 0.0431 12 0.0217 9 0.0094
6 13 7 18 0.0986 14 0.0470 ii 0.0235 7 0.0075
6 13 8 16 0.0924 12 0.0449 9 0.0206 6 0.0093
6 13 9 14 0.0913 ii 0.0495 7 0.0217 4 0.0077
6 13 i0 12 0.0898 8 0.0449 5 0.0170 4 0.0077
6 13 II 9 0.0851 6 0.0341 5 0.0170 0 0.0
6 13 12 8 0.0924 6 0.0341 0 0.0 0 0.0
6 13 13 7 0.0632 0 0.0 0 0.0 0 0.0
6 14 0 25 0.0893 21 0.0457 17 0.0204 13 0.0077
6 14 1 25 0.0893 21 0,0457 17 0.0204 13 0.0077

TABLE 13--Halperin's Un statistic (continued).

Uha Uhat lthct Uhct

My N~ Rn 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

6 14 2 25 0.0936 21 0 . 0 4 8 2 17 0.0216 13 0 . 0 0 8 2
6 14 3 24 0.0878 20 0 . 0 4 4 5 17 0.0242 13 0 . 0 0 9 2
6 14 4 24 0.0996 19 0 . 0 4 2 8 16 0.0231 12 0 . 0 0 R 6
6 14 5 22 0.0864 18 0 . 0 4 3 3 15 0.0231 11 0 . 0 0 8 3
6 14 6 21 0.0903 17 0 . 0 4 5 4 14 0.0244 i0 0 . 0 0 8 8
6 14 7 20 0.0972 16 0 . 0 4 9 7 12 0.0215 9 0.0099
6 14 R iR 0.0932 14 0 . 0 4 7 0 Ii 0.0246 7 0.0079
6 i4 9 16 0.0899 I2 0 . 0 4 4 4 9 0.0217 5 0.0087
6 14 iO I4 0.0921 IO 0 . 0 4 4 4 7 0.0249 4 0.0054
6 i4 Ii i2 0.0954 8 0.0477 6 0.0238 0 0.0
6 14 12 9 0.0851 7 0.0443 6 0.0238 0 0.0
6 14 13 8 0.0775 7 0.0443 o 0.0 0 0.0
6 14 14 R 0.0775 0 0.0 0 0.0 0 0.0
6 15 0 27 0.0890 23 0 . 0 4 7 4 19 0.0224 15 0 . 0 0 9 2
6 15 i 27 0.0890 23 0 . 0 4 7 4 19 0.0224 15 0 . 0 0 9 2
6 15 2 27 0.0930 23 0 . 0 4 9 9 19 0.0237 15 0 . 0 0 9 7
6 15 3 26 0.0876 22 0 . 0 4 6 3 iR 0.0216 14 0 . 0 0 8 6
6 15 4 26 0.0986 21 0 . 0 4 4 9 17 0.0206 13 0 . 0 0 8 0
6 i5 5 25 0.0994 20 0.0454 16 0.0208 i20.OORO
6 15 6 23 0.0899 19 0 . 0 4 7 4 15 0.0218 11 0 . 0 0 ~ 2
6 15 7 22 0.0961 17 0 . 0 4 3 0 14 0.0239 iO 0 . 0 0 9 4
6 15 8 20 0.0924 16 0 . 0 4 8 2 12 0.0224 8 0.0080
6 15 9 18 0.0890 14 0 . 0 4 8 0 Ii 0.0248 6 0.0085
6 15 iO 16 0.0898 12 0 . 0 4 4 9 8 0.0224 5 0.0085
6 15 II 14 0.0912 9 0.0426 6 0.0170 5 0.0085
6 i5 I2 ii 0.0900 8 0.0462 6 0.0170 0 0.0
6 i5 i3 9 0.0791 7 0.0316 0 0.0 0 0.0
6 I5 14 9 0.0922 0 0.0 0 0.0 0 0.0
6 15 15 9 0.0922 0 0.0 0 0.0 0 0.0
7 4 0 6 0.0818 4 0.0364 3 0.0212 1 0.0061
7 4 i 6 0.0818 4 0.0364 3 0.0212 I 0.0061
7 4 2 6 0.0848 4 0.0364 3 0.0212 1 0.0061
7 4 3 6 0.0909 4 0.0364 3 0.0212 1 0.0061
7 4 4 5 0.0697 4 0.0455 3 0.0212 1 0.0061
7 4 5 5 0.0939 3 0.0333 2 0.0121 1 0.0061
7 4 6 4 0.0879 3 0.0485 1 0.0061 1 0.0061
7 4 7 3 0.0667 2 0.0455 1 0.0242 0 0.0030
7 4 8 1 0.0242 I 0.0242 1 0.0242 0 0.0
7 5 0 8 0.0745 6 0.0366 5 0.0240 3 0.0088
7 5 i 8 0,0745 6 0,0366 5 0.0240 3 0.0088
7 5 2 8 0.0770 6 0.0379 5 0.0240 3 0.0088
7 5 3 8 0.0846 6 0.0404 4 0.0152 3 0.0088
7 5 4 7 0.0707 6 0.0480 4 0.0189 3 0.0088
7 5 5 7 0.0909 5 0.0404 4 0.0240 2 0.0051
7 5 6 6 0.0947 4 0.0366 3 0.0202 1 0.0025
7 5 7 5 0.0985 3 0.0278 2 0.0189 0 0.0013
7 5 8 3 0.0808 2 0.0455 1 0.0101 0 0.0
7 5 9 2 0.0455 2 0.0455 0 0.0 0 0.0
7 6 0 II 0.0903 8 0.0367 6 0.0175 4 0.0070

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 85

TABLE 13 Halperin's Un statistic (continued).

Uha ~ha ~ha ~ha

My N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

7 6 1 ii 0.0903 8 0.0367 6 0.(}175 4 0.0070

7 6 2 ii 0.0944 8 0.0379 6 0.0181 4 0.0070
7 6 3 I0 0.0798 8 0.0414 6 0.0192 4 0,0070
7 6 4 I0 0.0956 7 0.0338 6 0.0227 4 0.0087
7 6 5 9 0.0897 7 0.0431 5 0.0186 3 0.0064
7 6 6 8 0.0915 6 0.0443 5 0.0245 3 0.0093
7 6 7 7 0.0985 5 0.0455 3 0.0128 2 0.0087
7 6 8 5 0.0699 3 0.0373 2 0.0210 1 0.0047
7 6 9 3 0.0699 2 0.0210 2 0.0210 0 0.0
7 6 I0 3 0,0699 0 0.0 0 0.0 0 0o0
7 7 0 13 0.0825 Ii 0.0487 8 0.0189 6 0.0087
7 7 1 13 0.0825 ii 0.0487 8 0.0189 6 0.0087
7 7 2 13 0.0862 10 0.0579 8 0.0195 6 0.0090
7 7 3 13 0.0962 i0 0.0420 8 0.0213 6 0.0096
7 7 4 12 0.0886 9 0.0355 7 0.0172 5 0.0070
7 7 5 II 0.0851 9 0.0472 7 0.0219 5 0.0093
7 7 6 I0 0.0889 8 0.0478 6 0.0221 4 0.0084
7 7 7 9 0.0982 6 0.0370 5 0.0227 3 0.0064
7 7 8 7 0.0839 5 0.0350 3 0.0186 1 0.0023
7 7 9 5 0.0839 3 0.0350 2 0.0105 0 0.0
7 7 i0 4 0.0962 3 0.0350 0 0.0 0 0.0
7 7 ii 4 0.0962 0 0.0 0 0.0 0 0.0
7 8 0 16 0.0946 13 0.0469 10 0.0200 7 0.0070
7 8 1 16 0.0946 13 0.0469 i0 0.0200 7 0.0070
7 8 2 16 0.0995 13 0.0491 I0 0.0208 7 0.0071
7 8 3 15 0.0886 12 0.0420 I0 0.0230 7 0.0078
7 8 4 14 0.0828 12 0.0499 9 0.0193 7 0.0092
7 8 5 13 0.0810 Ii 0.0476 8 0.0179 6 0.0079
7 8 6 12 0.0861 i0 0.0494 7 0.0171 5 0.0065
7 8 7 II 0.0926 8 0.0382 6 0.0197 4 0.0078
7 8 8 9 0,0839 7 0.0448 5 0.0186 3 0.0099
7 8 9 7 0.0709 5 0,0448 3 0,0186 2 0.0056
7 8 I0 5 0.0839 3 0.0186 3 0.0186 0 0.0
7 8 ii 4 0.0513 0 0.0 0 0.0 0 0.0
7 9 0 18 0.0869 15 0.0454 12 0.0209 9 0.0082
7 9 i 18 0,0869 15 0.0454 12 0,0209 9 0.0082
7 9 2 18 0.0913 15 0.0476 12 0.0219 9 0.0085
7 9 3 17 0.0823 14 0.0414 12 0.0242 9 0.0094
7 9 4 17 0.0957 14 0.0490 II 0.0215 8 0.0078
7 9 5 16 0.0960 13 0.0477 i0 0.0201 7 0.0066
7 9 6 14 0.0808 ii 0.0371 9 0.0205 7 0.0096
7 9 7 13 0.0888 i0 0.0411 8 0.0215 5 0.0068
7 9 8 II 0.0790 9 0.0497 6 0.0178 4 0.0080
7 9 9 I0 0.0986 7 0.0599 4 0.0178 2 0.0031
7 9 i0 7 0.0839 5 0.0472 3 0.0105 0 0.0
7 9 ii 5 0.0692 4 0.0288 0 0.0 0 0.0
7 9 12 5 0.0692 0 0.0 0 0.0 0 0.0
7 I0 0 21 0.0966 17 0.0439 14 0.0215 Ii 0.0095
7 I0 1 21 0.0966 17 0.0439 14 0.0215 ii 0.0093

TABLE 13--Halperin's Ua statistic (continued).

U ha I,l h a I~ h a l t hct
My N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

7 I0 2 20 0.0846 17 0.0461 14 0.0226 Ii 0 . 0 0 9 6

7 I0 3 20 0.0932 16 0.0409 13 0.0192 I0 0 . 0 0 7 8
7 I0 4 19 0.0892 i6 0.0481 13 0.0227 IO 0 . 0 0 9 3
7 I0 5 18 0.0897 i5 0.0474 12 0.0222 9 O.OOA5
7 I0 6 17 0.0934 14 0.0496 Ii 0.0222 8 0.0085
7 10 7 15 0.0839 12 0.0432 i0 0.0245 7 0.0090
7 i0 R 14 0.0969 II 0.0479 8 0.0220 5 0.0062
7 I0 9 12 0.0947 9 0.0450 7 0.0234 3 0,0062
7 I0 10 i0 0.0947 7 0,0494 4 0.0170 3 0.0062
7 i0 ii 7 0.0882 5 0.0407 4 0.0170 0 0.0
7 I0 12 6 0.0882 5 0,0407 0 0,0 0 0.0
7 I0 13 6 0.0882 0 0.0 0 0.0 0 0.0
7 Ii 0 23 0.0896 19 0.0427 16 0.0221 12 0 . 0 0 7 7
7 II 1 23 0.0896 19 0.0427 16 0.0221 12 0 . 0 0 7 7
7 II 2 23 0.0938 19 0,0448 16 0.0232 12 0 . 0 0 8 0
7 ii 3 22 0.0868 19 0.0495 15 0.0201 12 0 . 0 0 8 9
7 ii 4 22 0.0995 IR 0.0470 15 0.0237 ii 0 . 0 0 7 8
7 ii 5 20 0.0842 17 0.0468 14 0.0232 ii 0 . 0 1 0 0
7 ii 6 19 0.0878 16 0.0492 13 0.0240 9 0.0074
7 Ii 7 18 0.0964 14 0,0431 ii 0.0200 8 0.0077
7 Ii 8 16 0.0913 13 0.0486 i0 0.0240 7 0.0090
7 II 9 14 0.0882 ii 0.0486 R 0.0209 5 0.0090
7 ii i0 12 0.0882 9 0.0434 6 0.0236 3 0.0038
7 ii ii I0 0.0975 6 0.0394 5 0.0249 0 0.0
7 ii 12 7 0.0830 5 0.0249 5 0.0249 0 0.0
7 II 13 6 0.0539 0 o.o 0 0.0 0 0.0
7 12 0 26 0.0978 21 0.0416 18 0.0225 14 0 . 0 0 8 5
7 12 1 26 0.0978 21 0.0416 18 0.0225 14 0 . 0 0 8 5
7 12 2 25 0.0875 21 0.0436 iR 0.0236 14 0 . 0 0 8 9
7 12 3 25 0.0955 21 0.0480 17 0.0209 14 0 . 0 0 9 9
7 12 4 24 0.0928 20 0.0460 17 0.0244 13 0 . 0 0 8 9
7 12 5 23 0.0934 19 0.0459 16 0.0242 12 0 . 0 0 8 6
7 12 6 22 0.0980 18 0.0483 14 0.0198 ii 0 . 0 0 8 6
7 12 7 20 0.0906 16 0.0431 13 0.0218 I0 0 . 0 0 9 5
7 12 8 18 0.0860 15 0.0493 II 0.0191 8 0.0085
7 12 9 17 0.0999 13 0.0474 I0 0.0241 7 0.0090
7 12 10 15 0.0999 Ii 0.0457 8 0.0232 4 0.0065
7 12 ii 13 0.0982 8 0.0432 6 0.0249 4 0.0065
7 I2 I2 9 0.0891 6 0.0341 5 0.0157 00.0
7 12 13 7 0.0681 6 0.0341 0 0.0 00.0
7 12 14 7 0.0681 0 0.0 0 0.0 00.0
7 i3 0 28 0.0913 24 0.0484 20 0.0228 16 0 . 0 0 9 3
7 I3 i 28 0.0913 24 0.0484 20 0.0228 16 0 . 0 0 9 3
7 13 2 28 0.0953 23 0.0425 20 0,0239 16 0 . 0 0 9 8
7 13 3 27 0.0894 23 0.0467 19 0.0215 15 0 . 0 0 8 4
7 13 4 26 0.0870 22 0.0449 19 0.0250 15 0 . 0 0 9 9
7 i3 5 25 0.0878 21 0.0451 18 0.0249 14 0 . 0 0 9 6
7 i3 6 24 0.0920 20 0.0473 16 0.0210 13 0 . 0 0 9 9
7 13 7 23 0.0989 iS 0.0426 15 0.0229 11 0 . 0 0 8 2

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 87

TABLE 13--Halperin's U, statistic (continued).

U h~ bthc~ tl A,~ lib a

M u N~ Rtt 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

7 13 R 2I 0.0949 17 0.0485 13 0.0207 I0 0.0099

7 13 9 19 0.0937 15 0.0465 ii 0.0200 8 0 . OOR6
7 13 I0 17 0.0937 13 0.0481 i0 0.023R 6 0.0097
7 13 Ii 15 0.0937 ii 0.0460 7 0.0221 4 0.0043
7 13 12 12 0.0937 8 0.0460 6 0.0221 0 0.0
7 13 13 9 ().08R5 7 0.0443 6 0.0221 0 0.0
7 13 14 8 0.0830 7 0.0443 0 0.0 0 0.0
7 13 15 8 0.0830 0 0.0 0 0.0 0 0.0
7 14 0 31 0.0984 26 0.0469 22 0.0230 17 0 . 0 0 7 9
7 14 1 31 0.0984 26 0.0469 22 0.0230 17 0 . 0 0 7 9
7 14 2 30 0.0895 26 0.0491 22 0.0242 17 O . O O R 3
7 14 3 30 0.0967 25 0.0455 2[ 0.0219 17 0 . 0 0 9 2
7 14 4 29 0.0944 24 0.0439 20 0.0209 16 0 . 0 0 8 6
7 14 5 28 0.0954 23 0.0441 19 0.0207 15 O . O O R 4
7 14 6 27 0.0994 22 0.0463 18 0.0218 14 0 . 0 0 8 6
7 14 7 25 0.0929 20 0.0423 17 0.0237 13 0 . 0 0 9 5
7 14 8 23 0.0892 19 0.0474 15 0.0219 II O . O O a 3
7 14 9 21 0.0881 17 0.0458 13 0.0213 9 0.0075
7 14 I0 19 0.0895 15 0.0469 ii 0.0198 7 0.0083
7 14 Ii 17 0.0902 13 0.0465 9 0.0227 5 0.0068
7 14 12 15 0.0920 10 0.0465 7 0.0227 5 0.0068
7 14 13 Ii 0.0885 R 0.0443 6 0.0148 0 0.0
7 14 14 9 0.0812 7 0.0295 0 0.0 0 0.0
7 14 15 9 0.0984 0 0.0 0 0.0 0 0.0
7 14 16 9 0.0~84 0 0.0 0 0.0 0 0.0
7 15 0 33 0.0926 28 0.0455 24 0.0233 19 0 . 0 0 8 6
7 15 1 33 0.0926 28 0.0455 24 0.0233 19 O . O O R 6
7 15 2 33 0.0962 28 0.0476 24 0.0244 19 0 . 0 0 9 0
7 15 3 32 0.0910 27 0.0443 23 0.0223 19 0 . 0 1 0 0
7 15 4 31 0.0890 26 0.0430 22 0.0214 18 0 . 0 0 9 4
7 15 b 30 0.0899 25 0.0433 21 0.0214 17 0 . 0 0 9 2
7 15 6 29 0.0934 24 0.0453 20 0.0224 16 0 . 0 0 9 6
7 15 7 28 0.1000 23 0.0489 19 0.0243 14 O . O O R 3
7 15 8 26 0.0966 21 0.0464 17 0.0227 13 0 . 0 0 9 4
7 15 9 24 0.0950 19 0.0458 15 0.0221 ii 0 . 0 0 9 1
7 15 I0 22 0.0970 17 0.0450 13 0.0219 90.OOSl
7 15 ii 20 0.0995 15 0.0480 II 0.0209 6 0.0074
7 15 12 17 0.0867 13 0.0480 R 0.0209 5 0.0046
7 15 13 14 0.0906 9 0.0402 7 0.0201 0 0.0
7 15 14 ii 0.0906 8 0.0377 7 0.0201 0 0.0
7 15 15 i0 0.0964 R 0.0377 0 0.0 0 0.0
7 [5 16 9 0.067[ 0 0.0 0 0.0 0 0.0
8 5 0 I0 0.0855 R 0.0466 6 0.0225 4 0.0093
8 5 1 i0 0.0855 8 0.0466 6 0.0225 4 0.0093
8 5 2 10 0.0878 8 0.0474 6 0.0225 4 0.0093
8 5 3 I0 0.0956 7 0.0350 6 0.0241 4 0.0093
8 5 4 9 0.0816 7 0.0396 5 0.0171 4 0.0093
8 5 5 9 0.0971 7 0.0490 5 0.0202 3 0.0054
8 5 6 8 0.0963 6 0.0443 4 0.0163 3 0.0093

TABLE 13--Halperin's U a statistic ( c o n t i n u e d ) .

lgha Uha I~lha Igha

Mu N~. Ru 0.I0 Prob 0.05 Prob 0.025 Prob 0.01 Prob

8 5 7 7 0.0971 5 0.0373 3 0.0140 2 0.0078

8 5 8 5 0.0754 4 0.0474 B 0.0194 1 0.0070
8 5 9 4 0.0909 2 0.0350 I 0.0070 I 0.0070
8 5 i0 2 0.0350 2 0.0350 0 0.0 0 0.0
8 6 0 13 0.0906 10 0.0406 8 0.0213 6 0.0100
8 6 i 13 0.0906 i0 0.0406 8 0.0213 6 0.0100
8 6 2 13 0.0939 I0 0.0416 8 0.0216 6 0.0100
8 6 3 12 0.0799 i0 0.0450 R 0.0230 5 0.0063
8 6 4 12 0.0919 9 0.0373 7 0.0176 5 0.0073
8 6 5 ii 0.0852 9 0.0453 7 0.0216 5 0.0087
8 6 6 I0 0.0846 8 0.0443 6 0.0193 4 0.0070
8 6 7 9 0.0909 7 0.0443 5 0.0160 3 0.0060
8 6 8 7 0.0723 5 0.0323 4 0.0203 3 0.0083
8 6 9 6 0.0909 4 0.0390 2 0.0150 1 0.0030
8 6 10 4 0.0949 2 0.0150 2 0.0150 0 0.0
8 6 ii 3 0.0549 0 0.0 0 0.0 0 0.0
8 7 0 16 0.0946 13 0.0469 I0 0.0200 7 0.0070
8 7 I 16 0.0946 13 0.0469 i0 0.0200 7 0.0070
8 7 2 16 0.09~4 13 0.0486 i0 0.0205 7 0.0071
8 7 3 15 0.0861 12 0.0406 i0 0.0221 7 0.0075
8 7 4 15 0,0991 12 0.0466 9 0.0180 7 0.0084
8 7 5 14 0.0959 ii 0.0427 9 0.0218 6 0.0067
8 7 6 13 0.0977 i0 0.0420 R 0.0210 6 0.0090
8 7 7 II 0.0796 9 0.0448 7 0.020a 5 0.0075
8 7 8 i0 0.0915 8 0.0480 5 0.0151 4 0.0095
8 7 9 8 0.0872 6 0.0424 5 0.0238 2 0.0070
8 7 10 6 0.0R16 4 0.0443 2 0.0070 2 0.0070
8 7 ii 4 0.0769 3 0.0256 0 0.0 0 0.0
8 7 i2 4 0.0769 0 0.0 0 0.0 0 0.0
8 8 0 19 0.0974 I5 0.0415 13 0.0249 9 0.0074
8 8 I 19 0.0974 15 0.0415 13 0.0249 9 0.0074
8 ~ 2 18 0.0835 15 0.0430 12 0.0195 9 0.0075
8 8 3 18 0.0913 15 0.0469 12 0.0212 9 0.0080
8 R 4 i7 0.0849 14 0.0423 12 0.0245 9 0.0092
8 8 5 16 0.0834 13 0.0397 Ii 0.0222 8 0.0078
8 8 6 15 0.0854 i2 0.0403 10 0.0217 7 0.0067
8 8 7 I4 0.0921 Ii 0.0415 9 0.0230 6 0.0071
8 8 8 12 0.0834 iO 0.0486 R 0.0246 5 0.0075
8 8 9 ii 0.0986 8 0.0464 6 0.0212 4 0.0091
8 8 lO 9 0.0921 6 0.0408 4 0.0221 2 0.0035
8 8 Ii 6 0.0897 4 0.0385 3 0.0128 0 0.0
8 8 12 5 0.i000 4 0.0385 0 0.0 0 0.0
8 8 13 5 0.1000 0 0.0 0 0.0 0 0.0
8 9 0 22 0,0998 18 0.0464 15 0.0232 II 0.0076
8 9 I 22 0.0998 IR 0.0464 15 0.0232 ii 0.0076
8 9 2 21 0.0871 18 0.0482 15 0.0241 ii 0.0078
8 9 3 21 0.0949 17 0.0422 14 0.0203 Ii 0.0084
8 9 4 20 0.0900 17 0.0486 14 0.0234 Ii 0.0096
8 9 5 19 0.0~90 16 0.0473 13 0.0218 i0 0.0087

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 8(2

TABLE 13--Halperin's UH statistic (continued).

nag* U hot ll h a a ha
M~ N~ Rn 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

8 9 6 18 0.0923 15 0.0483 12 0.0220 9 0.0081

8 9 7 17 0.0995 13 0.0404 ii 0.0226 8 O. O O N 4
8 9 8 15 0.0928 12 0.0468 9 0.0191 7 0.0093
8 9 9 13 0.0882 i0 0.0433 8 0.0246 50. 0063
8 9 I0 II 0.0882 8 0.0401 6 0.0216 3 0.0068
8 9 Ii 9 0.0882 6 0.0475 4 0.0204 30. 0068
8 9 12 6 0.0855 4~ 0.0204 4 0.0204 0 0.0
8 9 13 5 0.0529 0 0.0 0 0.0 0 0.0
8 I0 0 24 0.0864 20 0.0416 17 0.0217 13 0 . 0 0 7 8
8 I0 1 24 0.0864 20 0.0416 17 0.0217 13 0 . 0 0 7 8
8 i0 2 24 0.0900 20 0.0433 17 0.0226 13 0 . 0 0 8 0
8 i0 3 24 0.0978 20 0.0473 17 0.0246 13 0 . 0 0 8 7
8 i0 4 23 0.0937 19 0.0441 16 0.0224 12 0 . 0 0 7 5
8 I0 5 22 0.0937 18 0.0433 15 0.0213 12 0 . 0 0 9 2
8 i0 6 21 0.0972 I7 0.0444 14 0.0215 Ii 0 . 0 0 9 0
8 10 7 19 0.0878 16 0.0484 13 0.0230 i0 0 . 0 0 9 5
8 i0 8 18 0.0998 14 0.0434 II 0.0198 8 0.0072
8 I0 9 16 0.0969 12 0.0412 i0 0.0249 7 0.0090
8 10 10 14 0.0998 I0 0.0422 8 0.0223 5 0.0093
8 10 ii 12 0.0988 9 0.0490 ~ 0.0189 3 0.0038
8 i0 12 8 0.0837 6 0.0475 4 0.0113 0 0.0
8 i0 13 6 0.0686 5 0.0294 0 0.0 0 0,0
8 i0 14 6 0.0686 0 0.0 0 0.0 0 0.0
8 11 0 27 0.0887 23 0.0454 19 0.0204 15 0 . 0 0 7 9
8 ii 1 27 0.0887 23 0.0454 19 0.0204 15 0 . 0 0 7 9
8 ii 2 27 0.0923 23 0.0473 19 0.0213 15 0 . 0 0 8 2
8 II 3 27 0.0999 22 0.0429 19 0.0232 15 0 . 0 0 8 9
8 Ii 4 26 0.0966 22 0.0489 18 0.0214 14 0 . 0 0 7 8
8 II 5 25 0,0968 21 0.0484 17 0.0206 14 0 , 0 0 9 7
8 ii 6 23 0.0858 19 0.0411 16 0.0211 13 0 . 0 0 9 6
8 ii 7 22 0.0931 18 0.0446 15 0.0228 ii 0 . 0 0 7 4
8 ii 8 20 0.0882 16 0.0410 13 0.0204 i0 0 . 0 0 8 5
8 ii 9 18 0.0866 15 0.0488 12 0.0243 8 0.0079
8 II I0 16 0.0882 13 0.0482 i0 0.0244 7 0.0085
8 Ii ii 14 0.0921 ii 0.0493 8 0.0240 4 0.0065
8 ii 12 12 0.0921 8 0.0485 5 0.0170 4 O.OOA5
8 ii 13 8 0.0851 6 0.0397 5 0.0170 0 0.0
8 11 14 7 0.0851 6 0.0397 0 0.0 0 0.0
8 ii 15 7 0.0851 0 0.0 0 0.0 0 0.0
8 12 0 30 0.0907 26 0.0489 22 0.0237 17 0 . 0 0 7 9
8 12 1 30 0.0907 26 0.0489 22 0.0237 17 0 . 0 0 7 9
8 12 2 30 0.0942 25 0.0430 22 0.0247 17 0 . 0 0 8 2
8 12 3 Z9 0.0881 25 0.0467 21 0.0220 17 0 . 0 0 9 0
8 12 4 29 0.0987 24 0.0446 20 0.0205 16 O . O O S l
8 12 5 28 0.0994 23 0.0443 20 0.0248 16 0 . 0 0 9 9
8 12 6 26 0.0894 22 0.0460 18 0.0206 14 0 . 0 0 7 7
8 12 7 25 0.0962 21 0.0498 17 0.0222 13 0 . 0 0 8 2
8 12 8 23 0.0919 19 0.0467 15 0.0201 12 0 . 0 0 9 4
8 12 9 21 0.0912 17 0.0452 14 0.0239 I0 0 . 0 0 8 6

TABLE 13--Halperin's Un statistic (continued).

Uha R ha tl ha Uha
Mu N~ R~ 0.I0 Prob 0.05 Prob 0.025 Prob 0.01 Prob

8 12 I0 19 0.0919 15 0 . 0 4 5 2 12 0.0249 8 0.0077

8 12 II 17 0,0982 13 0 . 0 4 6 9 i0 0.0233 6 0.0092
8 12 12 15 0.0982 ii 0 . 0 4 7 9 7 0.0228 4 0.0039
8 12 13 II 0.0920 7 0.0375 6 0.0238 0 0.0
8 12 14 8 0.0783 6 0.0238 6 0.0238 0 0.0
8 12 15 7 0.0511 0 0.0 0 0.0 0 0.0
8 13 0 33 0.0923 28 0 . 0 4 4 5 24 0.0223 20 0.0099
8 13 i 33 0.0923 28 0 . 0 4 4 5 24 0.0223 20 0.0099
8 13 2 33 0.0958 28 0 . 0 4 6 4 24 0.0232 19 0.0083
8 13 3 32 0.0902 27 0 . 0 4 2 7 23 0.0209 19 0.0090
8 13 4 31 0.0878 27 0 . 0 4 8 4 23 0.0239 18 0.0083
8 13 5 30 0.0885 26 0 . 0 4 8 4 22 0.0237 17 0.0079
8 13 6 29 0.0920 24 0 . 0 4 2 5 21 0.0245 16 0.0081
8 13 7 28 0.0989 23 0 . 0 4 5 9 19 0.0216 15 0.0086
8 13 8 26 0.0953 21 0 . 0 4 3 4 18 0.0245 14 0.0099
8 13 9 24 0.0948 19 0 . 0 4 2 0 16 0.0238 12 0.0092
8 13 10 22 0.0957 17 0 . 0 4 2 6 14 0.0232 I0 0.0091
8 13 ii 19 0.0874 15 0 . 0 4 4 3 12 0.0245 8 0.0089
8 13 12 17 0.0920 13 0 . 0 4 6 5 9 0.0219 5 0.0063
8 13 13 15 0.0907 I0 0 . 0 4 8 5 7 0.0232 5 0.0063
8 IB 14 ii 0.0991 8 0.0485 6 0.0148 0 0.0
8 13 15 9 0.0949 7 0.0316 0 0.0 0 0.0
8 13 16 R 0.0632 0 0.0 0 0.0 0 0.0
8 14 0 36 0.0938 31~0.0475 26 0.0211 22 0.0098
8 14 I 36 0.0938 31~0.0475 26 0.0211 22 0.0098
8 14 2 36 0.0972 31 0 . 0 4 9 4 26 0.0220 21 0.0083
8 14 3 35 0.0919 30 0 . 0 4 6 0 26 0.0239 21 0.009I
8 14 4 34 0.0898 29 0 . 0 4 4 4 25 0.0227 20 0.0084
8 14 5 33 0.0905 28 0 . 0 4 4 5 24 0.0226 19 0.008I
8 14 6 32 0.0941 27 0 . 0 4 6 3 23 0.0234 18 0.0083
8 14 7 30 0.08R6 25 0 . 0 4 2 6 21 0.0210 17 0.0090
8 14 8 29 0.0974 24 0 . 0 4 7 6 20 0.0237 15 0.0081
8 14 9 27 0.0966 22 0 . 0 4 6 8 18 0.0232 14 0.0096
8 14 I0 25 0.0987 20 0 . 0 4 7 5 16 0.0226 12 0.0098
8 14 II 22 0.0899 18 0 . 0 4 9 4 14 0.0249 I0 0.0092
8 14 12 20 0.0986 15 0 . 0 4 3 7 12 0.0243 7 0.0090
8 14 I3 I7 0.0845 13 0 . 0 4 7 0 8 0.0201 6 0.0094
8 14 14 I5 0,0953 9 0.0416 7 0.0201 6 0.0094
8 14 15 I0 0.0805 R 0.0402 7 0.0201 0 0.0
8 14 16 9 0.0760 8 0.0402 0 0.0 0 0,0
8 14 17 9 0.0760 0 0.0 0 0.0 0 0.0
8 15 0 39 0.0951 33 0 . 0 4 3 7 29 0.0237 24 0.0097
8 15 I 39 0.0951 33 0 . 0 4 3 7 29 0.0237 24 0.0097
8 15 2 39 0.0983 33 0 . 0 4 5 4 29 0.0246 23 0.0083
8 15 3 38 0.0934 33 0 . 0 4 8 9 28 0.0227 23 0.0090
8 15 4 37 0.0915 32 0 . 0 4 7 5 27 0.0217 22 0.0085
8 15 5 36 0.0923 31 0 . 0 4 7 7 26 0.0216 21 0.0083
8 15 6 35 0.0957 30 0 . 0 4 9 7 25 0.0224 20 0.0085
8 15 7 33 0.0904 28 0 . 0 4 6 2 24 0.0243 19 0.0092

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 91

T A B L E 1 3 - - H a l p e r i n ' s UH statistic ( c o n t i n u e d ) .

u ha Uh~t Uha Uha

Mu N~ R~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

8 15 R 32 0.0990 26 0.0441 22 0.0229 17 0.0084

8 15 9 30 0.0985 24 0.0436 20 0.0223 15 0.0080
8 15 I0 27 0.0888 22 0.0440 18 0.0223 14 0.0099
8 15 ii 25 0.0913 20 0.0461 16 0.0235 II 0.0076
8 15 12 23 0.0987 18 0.0489 14 0.0247 9 0.0091
8 15 13 20 0.0936 15 0.0446 II 0.0236 7 0.0096
8 15 14 18 0.0936 12 0.0481 8 0.0201 6 0.0061
8 15 15 13 0.0919 9 0.0394 7 0.0131 0 0.0
8 15 16 ii 0.0962 9 0.0496 0 0.0 0 0.0
8 15 17 I0 0.0892 9 0.0496 0 0.0 0 0.0
8 15 18 I0 0.0892 0 0.0 0 0.0 0 0.0
9 6 0 15 0.0905 12 0.0440 I0 0.0248 7 0.0088
9 6 i 15 0.0905 12 0.0440 I0 0.0248 7 0.0088
9 6 2 15 0.0931 12 0.0450 9 0.0182 7 0,0088
9 6 3 15 0.0999 12 0.0478 9 0.0190 7 0.0092
9 6 4 14 0.0897 Ii 0.0400 9 0.0208 7 0.0098
9 6 5 13 0.0829 11 0.0472 9 0.0248 6 0.0074
9 6 6 12 0.0807 i0 0.0450 8 0.0220 6 0.0094
9 6 7 ii 0,0R23 9 0.0440 7 0.0192 5 0.0080
9 6 8 i0 0.0933 8 0.0478 6 0.0214 4 0.0074
9 6 9 8 0.0757 6 0.0398 5 0.0256 3 0.0056
9 6 10 7 0.0949 5 0.0380 3 0.0200 1 0.0020
9 6 ii 4 0.0769 3 0.0440 2 0.0110 0 0.0
9 6 12 3 0.0440 3 0.0440 0 0.0 0 0.0
9 7 0 18 0.0869 15 0.0454 12 0.0209 9 0.0082
9 7 1 18 0.0869 15 0.0454 12 0.0209 9 0.0082
9 7 2 18 0.0898 15 0.0466 12 0.0213 9 0.0083
9 7 3 18 0.0965 15 0,0499 12 0.0226 9 0.0087
9 7 4 17 0.0883 14 0.0440 Ii 0,0185 9 0.0094
9 7 5 16 0.0842 13 0.0400 II 0.0220 8 0.0076
9 7 6 15 0.0844 12 0.0385 i0 0.0207 8 0.0098
9 7 7 14 0.0882 II 0.0389 9 0.0201 7 0.0085
9 7 8 13 0.0979 i0 0.0430 8 0.0217 6 0.0094
9 7 9 II 0.0885 8 0.0339 7 0.0245 4 0.0064
9 7 10 9 0.0874 7 0.0455 5 0.0166 3 0.0087
9 7 II 7 0.0769 5 0.0481 3 0.0192 2 0.0048
9 7 12 4 0.0625 3 0.0192 3 0.0192 0 0.0
9 7 13 4 0.0625 0 0.0 0 0,0 0 0.0
9 8 0 22 0.0998 18 0.0464 15 0,0252 II 0.0076
9 8 i 22 0.0998 18 0.0464 15 0.0232 II 0.0076
9 8 2 21 0,0865 18 0.0478 15 0.0238 II 0.0077
9 8 3 21 0.0931 17 0.0412 14 0,0197 Ii 0,0081
9 8 4 20 0.0870 17 0.0464 14 0.0221 II 0.0090
9 8 5 19 0.0843 16 0.0438 13 0.0199 I0 0.0077
9 8 6 18 0.0858 15 0.0432 12 0.0190 I0 0.0099
9 8 7 17 0.0908 14 0.0452 II 0.0190 9 0.0095
9 8 8 15 0.0810 12 0.0390 i0 0.0209 7 0.0066
9 8 9 14 0.0968 Ii 0.0450 9 0.0242 6 0.0082
9 8 I0 12 0.0961 9 0.0430 7 0.0214 5 0.0078

T A B L E 13--Halperin's Un statistic (continued).

~ha ttha Uh~ tlh~

Mu N~ Rn O.lO Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 8 II I0 0.0973 7 0.0362 5 0.0226 3 0.0090

9 8 12 7 0,0905 5 0.0498 3 0.0090 3 0.0090
9 8 13 5 0.0824 4 0.0294 0 0.0 0 0.0
9 8 14 5 0.0824 0 0.0 0 0.0 0 0.0
9 9 0 25 0.0951 21 0.0470 17 0.0200 14 0.0094
9 9 I 25 0.0951 21 0.0470 17 0.0200 14 0.0094
9 9 2 25 0.0985 21 0.0485 17 0.0206 14 0.0096
9 9 3 24 0.0900 20 0.0428 17 0.0221 13 0.0077
9 9 4 23 0.0850 20 0.0483 17 0.0249 13 0.0086
9 9 5 23 0.0995 19 0.0465 16 0.0232 12 0.0076
9 9 6 2i 0.0854 18 0.0470 15 0.0228 12 0.0097
9 9 7 20 0.0914 17 0.0496 14 0.0236 II 0.0097
9 9 8 18 0.0848 15 0.0438 12 0.0200 9 0.0075
9 9 9 17 0.0984 13 0.0407 II 0.0234 8 0.0082
9 9 I0 15 0.0971 II 0.0372 9 0.0224 6 0.0073
9 9 II 12 0.0860 9 0.0385 7 0.0181 4 0.0079
9 9 12 II 0.0982 7 0.0452 5 0.0249 3 0.0045
9 9 13 7 0.0941 5 0,0412 4 0.0147 0 0o0
9 9 14 5 0,0412 5 0,0412 0 0.0 0 0.0
9 I0 0 28 0.0912 24 0.0474 20 0.0217 16 0.0086
9 I0 I 28 0.0912 24 0.0474 20 0.0217 16 0.0086
9 10 2 28 0.0944 24 0.0490 20 0.0225 16 0.0088
9 I0 3 27 0,0871 23 0.0440 20 0.0242 16 0.0095
9 i0 4 27 0.0972 23 0.0495 19 0.0219 15 0.0082
9 I0 5 26 0.0967 22 0.0484 18 0,0208 15 0.0097
9 I0 6 25 0.0996 21 0.0492 17 0.0205 14 0.0094
9 I0 7 23 0.0905 19 0.0428 16 0.0218 13 0.0096
9 I0 8 21 0.0852 18 0.0481 15 0.0241 II 0.0076
9 I0 9 20 0.0985 16 0.0457 13 0.0223 I0 0.0090
9 I0 I0 17 0.0824 14 0.0448 Ii 0.0201 8 0.0087
9 I0 II 15 0.0878 12 0.0470 9 0.0202 7 0.0095
9 I0 12 13 0,0913 I0 0.0485 7 0.0238 4 0.0077
9 I0 13 II 0.0913 7 0.0495 5 0,0217 4 0.0077
9 10 14 7 0.0867 5 0,0217 5 0.0217 0 0.0
9 I0 15 6 0.0542 0 0.0 0 0.0 0 0.0
9 II 0 31 0.0877 27 0.0476 23 0.0232 18 0.0079
9 II 1 31 0.0877 27 0.0476 23 0.0232 18 0.0079
9 II 2 31 0.0908 27 0.0493 23 0.0240 18 0.0082
9 II 3 31 0.0974 26 0.0449 22 0.0212 18 0.0088
9 II 4 30 0.0940 25 0.0423 22 0.0240 18 0.0099
9 II 5 29 0.0938 25 0.0496 21 0.0231 17 0.0092
9 Ii 6 28 0,0970 23 0.0424 20 0.0234 16 0,0091
9 II 7 26 0,0894 22 0,0454 19 0.0248 15 0.0095
9 II 8 25 0,0990 20 0.0420 17 0.0221 13 0.0079
9 ii 9 23 0.0974 19 0.0488 15 0.0205 12 0.0095
9 II I0 21 0.0997 17 0.0485 13 0.0195 I0 0.0089
9 II II 18 0.0874 14 0.0406 II 0.0210 8 0.0082
9 II 12 15 0.0815 12 0.0437 9 0.0190 5 0.0072
9 II 13 14 0.0988 9 0.0426 6 0.0196 4 0.0043

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 93

TABLE 13--Halperin's U~ statistic (continued).

I~ha U Aa lg het bl ha
Mu N~ R~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 ii 14 9 0.0834 7 0.0477 5 0.0119 0 0.0

9 ii 15 7 0.0681 6 0.0298 0 0.0 0 0.0
9 ii 16 7 0.06~i 0 0.0 0 0.0 0 0.0
9 12 0 35 0~ 30 0.0477 26 0.0245 21 0.0092
9 12 i B5 0.0965 30 0.0477 26 0,0245 21 0.0092
9 12 2 35 0.0998 30 0.0494 25 0.0212 21 0.0095
9 12 3 34 0.093R 29 0.0455 25 0.0228 20 0.0082
9 12 4 33 0.0909 28 0.0433 24 0.0213 20 0.0093
9 12 5 32 0.0911 27 0.0428 23 0.0207 19 0.00~7
9 12 6 31 0.0942 26 0.0441 22 0.0211 18 0.0087
9 12 7 29 0.0877 25 0.0472 21 0.0224 17 0.0091
9 12 8 28 0.0970 23 0.0441 19 0,0202 15 0.0079
9 12 9 26 0.0960 21 0.0428 18 0.0238 14 0.0096
9 12 i0 24 0.0980 19 0.0426 16 0.0239 12 0.0090
9 12 ii 21 0.0880 17 0.0446 14 0.0238 10 0.0092
9 12 12 19 0.0949 15 0.0482 11 0.0196 8 0.0092
9 12 13 16 0.0900 13 0.0491 9 0.0243 5 0.0068
9 12 14 14 0.0900 9 0.0477 6 0.0170 5 0.0068
9 12 15 9 0.0827 7 0.0389 6 0.0170 0 0.0
9 12 16 8 0.0827 7 0.0389 0 0.0 0 0.0
9 12 17 8 0.0827 0 0.0 0 0.0 0 0.0
9 13 0 38 0.0929 33 0.0478 28 0.0217 23 0.0085
9 13 I 38 0.0929 33 0.0478 28 0.0217 23 0.0085
9 13 2 38 0.0960 33 0.0495 28 0.0225 23 0.0088
9 13 3 37 0.0906 32 0.0459 28 0.0242 23 0.0095
9 13 4 37 0.0998 31 0.0440 27 0.0228 22 0.0087
9 13 5 35 0.0885 30 0.0438 26 0.0224 21 0.0083
9 13 6 34 0.0917 29 0.0452 25 0.0230 20 0.0083
9 13 7 33 0.0977 28 0,0484 24 0.0246 19 0.0088
9 13 8 31 0.0944 26 0.0459 22 0.0228 18 0.0099
9 13 9 29 0.0939 24 0.0448 20 0.0219 16 0.0092
9 13 i0 27 0.0954 22 0.0455 18 0.0219 14 0.0088
9 13 Ii 24 0.0881 20 0.0471 16 0.0222 if 0.0094
9 13 12 22 0,0929 17 0.0422 14 0,0247 10 0.0090
9 13 13 19 0.0937 15 0.0454 Ii 0.0195 7 0.0092
9 13 14 17 0.0937 12 0.0463 8 0.0221 5 0.0040
9 13 15 12 0.0877 9 0.0489 7 0.0230 0 0.0
9 13 16 9 0.0747 8 0.0489 7 0.0230 0 0.0
9 13 17 9 0,0977 8 0,0489 0 0.0 0 0,0
9 13 18 9 0.0977 0 0.0 0 0.0 0 0.0
9 14 0 41 0.0897 36 0.0478 31 0.0228 26 0.0096
9 14 i 41 0.0897 36 0.0478 31 0.0228 26 0.0096
9 14 2 41 0,0926 36 0.0495 31 0,0236 26 0.0099
9 14 3 41 0.0985 35 0.0462 30 0.0216 25 0.0088
9 14 4 40 0.0963 34 0.0446 30 0.0242 25 0.0099
9 14 5 39 0.0969 33 0.0445 29 0.0240 24 0.0097
9 14 6 37 0.0891 32 0.0460 28 0.0247 23 0.0098
9 14 7 36 0.0949 31 0.0492 26 0.0223 21 0.0085
9 14 8 34 0.0921 29 0.0470 25 0.0248 20 0.0095

TABLE 13--Halperin's Ua statistic (continued).

U hct U het U hot l t hct

My Nx Rn 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 14 9 32 0.0916 27 0.0463 23 0 . 0 2 4 0 iS 0 . 0 0 8 9
9 14 I0 30 0.0955 25 0.0470 21 0 . 0 2 4 1 16 0 . 0 0 8 8
9 14 Ii 28 0.0981 23 0.0487 18 0 . 0 2 1 0 14 0 . 0 0 9 0
9 14 12 25 0.0914 20 0,0454 16 0 . 0 2 2 4 12 0 . 0 0 9 6
9 14 13 22 0.0901 17 0.0418 13 0 . 0 1 9 2 9 0.00~7
9 14 14 19 0.0859 15 0.0439 ii 0 . 0 2 4 5 7 0.0098
9 14 15 17 0.0849 ii 0.0455 8 0.0219 6 0.0061
9 14 16 12 0.0927 9 0.0455 7 0o0140 0 0~
9 14 17 i0 0.0892 8 0.0297 0 0.0 0 0.0
9 14 18 9 0.0595 0 0.0 00.0 00.o
9 15 0 45 0.0968 39 0.0478 34 0 . 0 2 3 8 28 0 . 0 0 8 9
9 15 i 45 0.0968 39 0.0478 34 0 . 0 2 3 8 28 0 . 0 0 8 9
9 15 2 45 0.0998 39 0,0494 34 0 . 0 2 4 6 28 0 . 0 0 9 2
9 15 3 44 0.0951 38 0.0464 33 0 . 0 2 2 8 28 0 . 0 0 9 9
9 15 4 43 0.0932 37 0.0450 32 0 . 0 2 1 8 27 0 . 0 0 9 3
9 15 5 42 0.0938 36 0.0450 31 0 . 0 2 1 6 26 O . O O W l
9 15 6 41 0.0970 35 0.0466 30 0 . 0 2 2 3 25 0 . 0 0 9 3
9 15 7 39 0.0921 34 0.0498 29 0 . 0 2 3 8 24 0 . 0 0 9 9
9 15 37 0.0896 32 0.0479 27 0 . 0 2 2 5 22 O . O O q l
9 15 9 35 0.0892 30 0.0474 25 0 . 0 2 2 0 20 0 . 0 0 8 7
9 15 I0 33 0.0911 28 0.0481 23 0 . 0 2 2 1 18 0 . 0 0 8 7
9 15 Ii 31 0.0949 25 0.0435 21 0 . 0 2 3 2 16 0 . 0 0 8 7
9 15 12 28 0.0907 23 0.0463 19 0 . 0 2 4 3 14 0 . 0 0 9 8
9 15 13 26 0.0967 20 0.0442 16 0 . 0 2 3 2 12 0 . 0 0 9 7
9 15 14 22 0.0901 18 0.0481 14 0 . 0 2 4 8 8 0.0084
9 15 15 20 0.0924 15 0.0481 10 0 . 0 2 3 5 7 0.0087
9 15 16 1R 0.0962 ii 0.0481 8 0.0186 7 0.0087
9 15 17 12 0 . 0 9 3 0 9 0.0372 8 0.0186 0 0.0
9 15 18 10 0.0707 9 0.0372 0 0.0 0 0.0
9 15 19 10 0 . 0 7 0 7 0 0.0 0 0.0 0 0.0
I0 7 0 21 0 . 0 9 6 6 17 0 . 0 4 3 9 14 0 . 0 2 1 5 Ii 0 . 0 0 9 3
I0 7 1 21 0 . 0 9 6 6 I7 0.0439 14 0 . 0 2 1 5 ii 0 . 0 0 9 3
i0 7 2 21 0 . 0 9 9 4 17 0 . 0 4 4 9 14 0 . 0 2 1 9 11 0 . 0 0 9 4
i0 7 3 20 0.08~2 17 0.0475 14 0 . 0 2 3 0 ii 0 . 0 0 9 7
I0 7 4 20 0.0974 16 0.0418 13 0 . 0 1 9 2 I0 0 . 0 0 7 5
I0 7 5 19 0.0927 16 0.0484 13 0 . 0 2 1 9 i0 0 . 0 0 8 5
i0 7 6 18 0.0923 15 0.0462 12 0 . 0 2 0 1 9 0.0072
I0 7 7 17 0.0947 14 0.0467 II 0 . 0 1 9 1 9 0.0090
I0 7 8 15 0.0831 15 0.0499 I0 0 . 0 1 9 6 ~ 0.0093
i0 7 9 14 0.0936 ii 0.0411 9 0.0222 7 0,00q7
i0 7 I0 12 0.0908 9 0.0376 8 0.0245 5 0.0073
I0 7 Ii i0 0.0939 7 0.0345 6 0.0232 4 0.0090
I0 7 12 R 0.0854 6 0.0486 3 0.0147 2 0.0034
iO 7 i3 5 0.0882 3 0.0147 3 0.0147 0 0.0
I0 7 14 4 0.0515 0 0.0 0 0.0 0 0.0
i0 8 0 24 0.0864 20 0.0416 17 0 . 0 2 1 7 13 0 . 0 0 7 8
I0 8 i 24 0.0864 20 0.0416 17 o . 0 2 1 7 13 0 . 0 0 7 8
i0 8 2 24 0.0~90 20 0.0426 17 0 . 0 2 2 2 15 0 . 0 0 7 9
i0 8 3 24 0.0949 20 0,0454 17 0 . 0 2 3 4 13 0 . 0 0 8 2

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 95

TABLE 13--Halperin's U. statistic (continued).

U I~ U h~ lt h a a hot
Mu N. Rn 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

i0 8 4 23 0.0884 19 0.0407 16 0 . 0 2 0 3 13 0.0090

i0 8 5 22 0.0855 19 0.0471 16 0 . 0 2 3 5 t2 0.0076
I0 8 6 21 0.0857 18 0.0465 15 0 . 0 2 2 2 t2 0.0093
i0 8 7 20 0.0900 17 0.0474 14 0 . 0 2 2 3 11 0.00~8
10 8 8 19 0.0977 15 0.0406 13 0 . 0 2 3 4 I0 0.0090
i0 R 9 17 0.0923 14 0.0461 ii 0 . 0 1 8 9 8 0.0088
i0 8 I0 15 0.0899 12 0.0437 10 0 . 0 2 3 8 7 0.0078
I0 8 Ii 13 0.0882 10 0.0430 R 0.0241 5 0.0053
i0 8 12 ii 0.0995 8 0.0430 6 0.0216 3 0.0065
i0 8 13 9 0.0R82 5 0.0392 4 0.0229 3 0.0065
i0 8 14 5 0.0686 4 0.0229 4 0.0229 0 0.0
I0 8 15 5 0.0686 0 0.0 00.0 00.0
10 9 0 28 0.0912 24 0.0474 20 0 . 0 2 1 7 16 0 . 0 0 8 6
i0 9 i 28 0.0912 24 0.0474 20 0 . 0 2 1 7 16 0 . 0 0 8 6
i0 9 2 28 0.0940 24 0.0487 20 0.0223 16 0 . 0 0 8 8
i0 9 3 27 0.0859 23 0.0432 20 0.0237 16 0 . 0 0 9 2
I0 9 4 27 0.0950 23 0.0480 19 0.0210 15 0 . 0 0 7 8
i0 9 5 26 0.0931 22 0.0459 19 0.0244 15 0. 0 0 9 0
i0 9 6 25 0.0946 21 0.0457 18 0.0238 14 0 . 0 0 8 3
i0 9 7 23 0.0841 20 0.0477 17 0.0242 13 0 . 0 0 8 1
i0 9 8 22 0.0924 18 0.0419 15 0.0204 12 0 . 0 0 8 6
i0 9 9 20 0.0891 17 0.0480 14 0.0231 ii 0 . 0 0 9 2
i0 9 i0 18 0.0874 15 0.0465 12 0.0218 9 0.0081
i0 9 ii 16 0.0906 13 0.0460 i0 0.0210 7 O. 0 0 7 3
i0 9 12 14 0.0978 ii 0.0495 8 0.0204 5 0.0079
i0 9 13 ii 0.0774 9 0.0495 5 0.0186 3 0.0031
I0 9 14 8 0.0975 5 0.0325 4 0.0108 0 0.0
i0 9 15 6 0.0867 5 0.0325 0 0.0 0 0.0
I0 9 16 6 0,0867 0 0.0 0 0.0 0 0.0
i0 I0 0 32 0.0952 27 0.0446 23 0.0216 19 O. 0 0 9 3
i0 I0 i 32 0.0952 27 0.0446 23 0.0216 19 0 . 0 0 9 3
I0 i0 2 32 0.0981 27 0.0460 23 0.0222 t 9 O. 0 0 9 5
i0 i0 3 31 0.0909 27 0.0490 23 0.0237 18 0 . 0 0 7 9
I0 I0 4 30 0.0868 26 0.0458 22 0.0215 18 0 . 0 0 8 7
I0 i0 5 30 0.0992 25 0.0443 22 0.0250 17 0 . 0 0 7 9
10 10 6 28 0.0875 24 0.0447 21 0.0247 17 0 . 0 0 9 7
I0 i0 7 27 0.0925 23 0.0469 19 0.0205 16 0 . 0 0 9 8
I0 I0 8 25 0.0870 21 0.0425 18 0.0223 14 O. 0 0 7 8
I0 I0 9 24 0.0991 20 0.0491 16 0.0204 13 0 . 0 0 8 9
i0 i0 i0 21 0.0839 18 0.0480 15 0.0243 ii 0 . 0 0 8 0
I0 i0 ii 19 0.0876 16 0.0492 13 0.0242 9 0.0078
i0 i0 12 17 0.0913 13 0.0420 i0 0.0206 70. 0063
i0 i0 13 14 0.0913 ii 0.0426 9 0.0248 5 0.0093
i0 i0 14 12 0.0867 8 0.0488 5 0.0163 4 O. 0 0 5 4
i0 i0 15 8 0.0975 6 0,()433 5 0.0163 0 0.0
i0 i0 16 6 0.0433 6 0.0433 0 0.0 0 0.0
I0 II 0 36 0.0986 31 0.0493 26 0.0215 22 0 . 0 0 9 8
i0 Ii 1 36 0.09~a6 31 0.0493 26 0.0215 22 0 . 0 0 9 8
i0 ii 2 35 0.0894 30 0.0436 26 0.0221 21 0 . 0 0 8 1

TABLE 13--Halperin's U . statistic (continued).

Rh= Uh~ Uha gh~

Mu Nx R~, 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

I0 II 3 35 0.0951 30 0.0465 26 0.0236 21 0 . 0 0 8 6

I0 II 4 34 0.0916 29 0.0439 25 0.0217 21 0 . 0 0 9 6
i0 ii 5 33 0.0910 28 0.0429 24 0.0208 20 0 . 0 0 8 9
i0 ii 6 32 0.0934 27 0.0435 23 0.0207 19 0 . 0 0 8 6
I0 ii 7 31 0.0991 26 0.0460 22 0.0217 18 0 . 0 0 8 8
I0 ii 8 29 0.0944 24 0.0424 21 0.0237 17 0 . 0 0 9 5
I0 Ii 9 27 0.0926 23 0.0484 19 0.0219 15 0 . 0 0 8 4
i0 Ii I0 25 0.0941 21 0.0482 17 0.0213 13 0 . 0 0 7 9
i0 II II 23 0.0973 19 0.0498 15 0.0217 11 0 . 0 0 7 2
10 ii 12 20 0,0894 16 0.0440 13 0.0226 9 0.0074
i0 II 13 17 0.0864 14 0.0499 iI 0.0243 7 0.0089
i0 Ii 14 15 0.0937 Ii 0.0426 7 0.0199 5 0 . 0085
I0 II 15 13 0.0957 7 0.0369 6 0.0227 5 0.0085
I0 Ii 16 8 0.0876 6 0.0227 6 0.0227 D 0.0
i0 Ii 17 7 0.0551 0 0.0 0 0.0 0 0.0
i0 12 0 39 0.0901 34 0.0465 29 0.0213 24 0 . 0 0 8 4
i0 12 I 39 0.0901 34 0.0465 29 0.0213 24 0 . 0 0 8 4
10 12 2 39 0.0928 34 0.0480 29 0.0219 24 0 . 0 0 8 7
I0 12 3 39 0.0986 33 0.0442 29 0.0234 24 0 . 0 0 9 2
I0 12 4 38 0.0956 33 0.0489 28 0.0219 23 0 . 0 0 8 3
I0 12 5 37 0.0955 32 0.0483 27 0.0212 23 0 . 0 0 9 7
I0 12 6 36 0.0983 31 0.0493 26 0.0213 22 0 . 0 0 9 6
I0 12 7 34 0.091R 29 0.0447 25 0.0224 21 0 . 0 0 9 9
tO 12 R 32 0.0881 28 0.0492 24 0.0247 19 0 . 0 0 8 6
i0 12 9 31 0.0992 26 0.0478 22 0.0234 18 0 . 0 1 0 0
IO I2 tO 28 0.0884 24 0.0480 20 0.o231 16 O . O 0 9 b
tO 12 It 26 0.0919 22 0.0500 18 0.0237 14 0 . 0 0 9 4
iO 12 12 23 0.0862 19 0.0445 15 0.0200 ii 0 . 0 0 7 8
I0 12 13 21 0.0927 16 0.0422 13 0.0223 9 0.0071
i0 12 14 17 0.0820 14 0.0449 ii 0.0232 6 0.0077
i0 12 15 16 0.0986 10 0.0433 7 0.0201 5 0.0046
I0 12 IA i0 0.0831 8 0.0478 6 0.0124 0 0.0
I0 12 17 8 0.0677 7 0.0301 0 0.0 0 0.0
i0 12 18 8 0.0677 0 0.0 0 0.0 0 0.0
i0 13 0 43 0.0951 37 0.0441 33 0.0247 27 0 . 0 0 8 9
i0 13 i 43 0.0931 37 0.0441 33 0.0247 27 0 . 0 0 8 9
i0 13 2 43 0.0959 37 0.0455 32 0.0217 27 0 . 0 0 9 1
i0 13 3 42 0.0907 37 0.0485 32 0.0232 27 0 . 0 0 9 8
i0 13 4 42 0.0990 36 0.0466 31 0.0219 26 0 . 0 0 9 0
i0 13 5 41 0.0992 35 0.0462 30 0.0214 25 0 . 0 0 8 5
i0 13 6 39 0.0909 34 0.0474 29 0.0217 24 0 . 0 0 8 5
I0 13 7 38 0.0964 32 0.0435 28 0.0229 23 0 . 0 0 8 9
10 13 R 36 0.0932 31 0,0477 26 0.0211 22 0 . 0 0 9 7
I0 13 9 34 0.0925 29 0.0466 25 0.0242 20 0 . 0 0 9 0
i0 13 i0 32 0.0942 27 0.0471 23 0.0241 18 0 . 0 0 8 7
10 13 ii 30 0.0989 25 0.0490 21 0.0249 16 0 . 0 0 9 0
I0 13 12 27 0.0932 22 0.0448 18 0.0222 14 0 . 0 0 9 1
I0 13 13 24 0.0894 20 0.0496 16 0.0245 ii 0 . 0 0 7 5
10 13 14 21 0.0905 17 0.0481 13 0.0219 9 0.0096

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 97

TABLE 13--Halperin's UH statistic (continued).

U h~ l t hc~ lg h a U ha
Mu N~ R~t 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

I0 13 15 18 0.0882 15 0.0489 10 0.0245 6 0.0070

i0 13 16 17 0.0995 i0 0.0470 7 0.0170 4 0.0070
I0 13 17 I0 0.0807 R 0.0382 7 0.0170 0 0.0
i0 13 18 9 0.0807 R 0.0382 0 0.0 0 0.0
I0 13 19 9 0.0807 0 0.0 0 0.0 0 0.0
i0 14 0 47 0.0958 41 0.0478 36 0.0242 30 0.0093
I0 14 i 47 0.0958 41 0.0478 36 0.0242 30 0.0093
I0 14 2 47 0.0985 41 0.0493 36 0.0250 30 0.0096
i0 14 3 46 0.0937 40 0.0462 35 0.0230 29 0.0085
i0 14 4 45 0.0915 39 0.0445 34 0.0218 29 0.0095
i0 14 5 44 0.0918 38 0.0445 33 0.0215 28 0.0092
i0 14 6 43 0.0946 37 0.0455 32 0.0219 27 0.0092
I0 14 7 41 0.0895 36 0.0483 31 0.0232 26 0.0097
I0 14 40 0.0973 34 0.0462 29 0.0217 24 0.00R8
i0 14 9 38 0.0970 32 0.0455 28 0.0248 22 0.00s
10 14 i0 36 0.0990 30 0.0461 26 0.0249 21 0.0099
i0 14 ii 33 0.0922 28 0.0480 23 0.0216 18 0.0083
I0 14 12 31 0.0981 25 0,0446 21 0.0233 16 0.00R5
i0 14 13 28 0.0971 23 0.0483 18 0.0214 13 0.0077
10 14 14 25 0.0927 20 0.0490 16 0.0245 11 0.0077
I0 14 ib 21 0.0901 17 0.0442 13 0.0219 8 0.0092
i0 14 16 19 0.0901 13 0.0449 9 0.0216 7 0.0099
i0 14 17 17 0.0967 I0 0.0471 8 0.0223 7 0.0099
I0 14 18 ii 0.0967 9 0.0471 R 0.0223 0 0.0
i0 14 19 i0 0.0942 9 0.0471 0 0.0 0 0.0
i0 14 20 i0 0.0942 () 0.0 0 0.0 0 0.0
10 15 0 51 0.0981 44 0.0455 39 0.0238 33 0.0096
10 15 1 51 0.0981 44 0.0455 39 0.0238 33 0.0096
i0 15 2 50 0.0912 44 0.0a69 39 0.0245 33 0.0099
10 15 3 50 0.0963 44 0.0498 38 0.0227 32 0.0090
10 t5 4 49 0.0944 43 0.0483 37 0.0217 32 0.0099
10 15 5 48 0.0948 42 0.0483 37 0.0248 31 0.0097
i0 15 6 47 0.0977 41 0.0497 35 0.0220 30 0.0098
i0 15 7 45 0.093(I 39 0.0465 34 0.0235 28 0.0087
i0 15 8 43 0.0907 37 0.0447 32 0.0220 27 0.0095
10 15 9 41 0.0904 35 0.0441 30 0.0215 25 0,0091
10 15 10 39 0.0923 33 0.0448 28 0.0216 23 0.0090
10 15 11 57 0.0963 31 0.0467 26 0.022b 21 0.0092
10 15 12 34 0.0921 28 0.0439 24 0.0240 19 0.0099
I0 15 13 31 0.0901 26 0.0479 21 0.0229 16 0.0090
10 lb 14 29 0.0997 23 0.0471 18 0.0217 13 0.0077
10 15 15 25 0.0910 20 0.0477 16 0.0241 I0 0,0086
i0 15 16 22 0.0969 18 0.0497 12 0.0234 8 0.0094
i0 15 17 20 0.0952 12 0.0431 9 0.0208 7 0.0059
i0 15 IR 13 0.087R i0 0.0431 8 0.0134 0 0.0
10 15 19 ii 0.0848 9 0.0283 0 0.0 0 0.0
10 15 20 10 0.0565 0 0.0 0 0.0 0 0.0
II 8 0 27 0.0887 23 0.0454 19 0.0204 15 0.0079
11 8 1 27 0.0887 23 0.0454 19 0.0204 15 0.0079

TABLE 1 3 - - H a l p e r i n ' s U n statistic ( c o n t i n u e d ) .

Hha aha Uha Uha

My N~ Rs 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob
ii 8 2 27 0.0909 23 0.0465 19 0 . 0 2 0 8 15 O . O O R O
Ii 8 3 27 0.0961 23 0.049(} 19 0 . 0 2 1 8 15 0 . 0 0 8 3
Ii 8 4 26 O.OR9R 22 0.0443 19 0 . 0 2 3 7 15 0 . 0 0 8 9
ii R 5 25 0.0863 21 0.0412 18 0 . 0 2 1 3 15 0 . 0 0 9 9
ii 8 6 24 0.0860 21 0.0490 17 0 . 0 1 9 9 14 0 . 0 0 8 9
II 8 7 23 0,0886 20 0.0497 16 0 . 0 1 9 4 13 0 . 0 0 8 3
ii 8 R 22 0.0955 18 0.0426 15 0 . 0 2 0 0 12 0 . 0 0 8 2
II 8 9 20 0.0895 17 0.0471 14 0 . 0 2 1 8 Ii 0.0086
II 8 I0 18 0.0866 15 0.0433 12 0 . 0 1 9 4 I0 0.0098
Ii 8 ii 16 0,0870 13 0.0427 Ii 0 . 0 2 3 7 R 0.0077
Ii 8 12 14 0.0878 ii 0.0412 9 0.0237 6 0.0074
Ii 8 13 ii 0.0805 9 0.0427 7 0.0200 4 0.0086
II 8 14 10 0.0975 6 0.0445 4 0.0181 3 0.0048
ii 8 15 6 0.0975 4 0.0181 4 0.0181 0 0.0
ii 8 16 5 0.0578 0 0.0 0 0.0 0 0.0
Ii 9 0 31 0.0877 27 0.0476 23 0 . 0 2 3 2 18 0 . 0 0 7 9
ii 9 I 31 0.0877 27 0~ 23 0 . 0 2 3 2 IR 0 . 0 0 7 9
Ii 9 2 31 0.0900 27 0.0488 23 0 . 0 2 3 7 18 0 . 0 0 8 1
II 9 3 31 0.0952 26 0.0435 23 0 . 0 2 5 0 18 0 . 0 0 8 4
II 9 4 30 0.0901 26 0.0476 22 0 . 0 2 2 3 18 0 . 0 0 9 1
11 9 5 29 0.0878 25 0.0453 21 0 . 0 2 0 6 17 0 . 0 0 8 0
Ii 9 6 28 0.0884 24 0.0448 21 0 . 0 2 4 5 17 0 . 0 0 9 5
Ii 9 7 27 0.0922 23 0,0459 20 0 , 0 2 4 7 16 0 . 0 0 9 2
ii 9 8 26 0.0997 22 0.0493 18 0 . 0 2 0 8 15 0 . 0 0 9 4
ii 9 9 24 0.0955 20 0.0453 17 0 . 0 2 3 0 13 0 . 0 0 7 6
II 9 10 22 0.0941 18 0.0435 15 0 . 0 2 0 9 12 0 . 0 0 8 9
Ii 9 ii 20 0.0959 16 0.0452 13 0 . 0 2 0 2 i0 0 . 0 0 7 8
Ii ~ 12 17 0.0857 14 0.0438 11 0 . 0 1 8 9 R 0.0081
Ii 9 13 15 0.0924 Ii 0.0379 9 0.0209 7 0.0090
II 9 14 12 0.0846 i0 0.0498 6 0.0200 4 0.0081
II 9 15 II 0.0975 6 0.0439 40.OORl 4 0.0081
ii 9 16 6 0.0737 5 0.0260 0 0.0 0 0.0
II 9 17 6 0.0737 0 0.0 0 0.0 0 0.0
II I0 0 36 0.0986 31 0.0493 26 0 . 0 2 1 5 22 0 . 0 0 9 8
ii I0 1 36 0.0986 31 0.0493 26 0 . 0 2 1 5 22 0 . 0 0 9 8
Ii i0 ? 35 0.0890 30 0.0434 26 0 . 0 2 2 0 210.OORI
II 10 3 35 0.0942 30 0.0459 26 0 . 0 2 3 2 21 0 . 0 0 8 5
ii i0 4 34 0.0900 29 0.042S 25 0 . 0 2 1 1 21 0 . 0 0 9 2
ii I0 5 33 0.0884 29 0.0486 25 0 . 0 2 4 0 20 0 . 0 0 8 3
Ii I0 6 32 0.0898 28 0.0486 24 0 . 0 2 3 5 20 0 . 0 1 0 0
ii I0 7 31 0.0940 26 0.0425 23 0 . 0 2 4 0 19 0 . 0 0 9 9
ii I0 8 29 0.0881 25 0.0458 21 0 . 0 2 0 8 17 0 . 0 0 7 9
ii I0 9 28 0.0989 23 0.0428 20 0 . 0 2 3 4 16 0 . 0 0 8 9
ii I0 i0 26 0.0992 21 0.0417 18 0 , 0 2 2 1 14 0 . 0 0 7 8
ii i0 ii 23 0.0873 19 0.0418 16 0 . 0 2 1 8 13 0 . 0 1 0 0
ii i0 12 21 0.0932 17 0.0441 14 0 . 0 2 2 3 ii 0 . 0 0 9 2
ii I0 13 18 0.0850 15 0.0489 12 0 . 0 2 5 0 9 0.0100
ii I0 14 15 0.0890 12 0.0436 9 0.0181 6 0.0095
II i0 15 13 0.0833 9 0.0464 6 0.0209 4 0.0039

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 9(2

TABLE 13--Halperin's UH statistic (continued).

U hct ~lh~ U h~ U ha
Mu N= RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

II IO i6 R 0.0805 6 0.0351 5 0.0124 0 0.0

iI IO 17 7 0.0902 6 0.0351 0 0.0 0 0.0
II iO 18 7 0.0902 0 0.0 0 0.0 0 0.0
ii iI 0 40 0.0966 34 0.0440 30 0.0237 25 0 . 0 0 9 6
II ii i 40 0.0966 34 0.0440 30 0,0237 25 0 . 0 0 9 6
ii Ii 2 40 0.0992 34 0~ 30 0.0243 25 0 . 0 0 9 8
ii II 3 39 0.0930 34 0.0478 29 0.0217 24 0 . 0 0 8 5
II Ii 4 38 0.0895 33 0.0451 29 0.0237 24 0 . 0 0 9 3
II II 5 37 0.0886 32 0.0440 28 0.0227 23 0 . 0 0 8 6
Ii ii 6 36 0.0903 31 0.0443 27 0.0225 22 0 . 0 0 8 2
ii ii 7 35 0.0949 30 0.0463 26 0.0232 21 0 . 0 0 8 3
ii ii 33 0.0901 28 0.0427 24 0.0207 20 0 . 0 0 8 7
ii ii 9 31 0.0@82 27 0.0480 23 0.0232 19 0 . 0 0 9 7
ii ii I0 29 0.08~5 25 0.0474 21 0.0224 i7 0 . 0 0 9 0
ii ii ii 27 0.0920 23 0.0487 I9 0.0224 15 0 . 0 0 ~ 6
Ii Ii 12 25 0.0976 20 0.0436 17 0,0234 i3 0 . 0 0 8 6
Ii iI 13 22 0.0950 18 0.0469 14 0.0200 II 0 . 0 0 9 8
ii II 14 19 0.08~5 i5 0.0459 12 0.0232 9 0.0090
Ii ii i5 16 0.0956 13 0.0459 9 0.0232 5 0.0062
II II 16 14 0.0905 8 0.0402 6 0.0175 5 0.0062
iI II 17 8 0.0727 7 0.0451 6 0.0175 0 0.0
ii iI 18 7 0.0451 7 0.0451 0 0.0 0 0.0
II 12 0 44 0.0948 38 0,0454 33 0.0219 28 0 . 0 0 9 4
ii I2 I 44 0.0948 3~ 0.0454 33 0.0219 28 0 . 0 0 9 4
ii 12 2 44 0.0974 38 0.0467 33 0.0225 28 0 . 0 0 9 6
ii 12 3 43 0.0918 38 0.0494 33 0.0239 27 0 . 0 0 8 4
II 12 4 43 0.0996 37 0.0471 32 0.0223 27 0 . 0 0 9 2
ii 12 5 42 0.0993 36 0.0463 31 0.0215 26 0 . 0 0 8 7
ii 12 6 40 0.0904 35 0.0470 30 0.0215 25 0 . 0 0 8 5
II 12 7 39 0.0951 34 0.0493 29 0.0223 24 0 . 0 0 8 6
Ii 12 8 37 0.0912 32 0.0462 28 0.0242 23 0 . 0 0 9 2
ii 12 9 35 0.0899 30 0.0447 26 0.0228 21 0 . 0 0 8 3
ii 12 I0 33 0.0910 28 0.0446 24 0.0224 20 0 . 0 0 9 9
II 12 II 31 0.0946 26 0.0461 22 0.0227 I8 0.0097
ii 12 12 28 0.0890 24 0.0491 20 0.0241 15 0 . 0 0 8 0
ii 12 13 26 0.0977 21 0.0457 17 0.0213 13 0 . 0 0 8 5
ii 12 14 23 0.0943 18 0.0447 15 0.0247 Ii 0 . 0 0 9 2
II 12 15 19 0.0891 15 0.0410 12 0.0225 8 0.0099
II 12 16 17 0.0955 13 0.0447 8 0.0210 6 0.0092
II 12 17 I5 0.0955 8 0.0379 7 0.0235 6 0.0092
ii 12 18 9 0.0883 7 0.0235 7 0.0235 0 0.0
Ii 12 19 8 0.0559 0 0.0 0 0.0 0 0,0
II 13 0 48 0.0931 42 0.0467 37 0.0237 3i 0.0092
ii 13 i 48 0.093I 42 0.0467 37 0.0237 31 0 . 0 0 9 2
II 13 2 48 0.0956 42 0.0480 37 0.0244 31 0 . 0 0 9 4
ii 13 3 47 0.0906 41 0.0447 36 0.0223 310.OIO0
Ii 13 4 47 0.098I 41 0.0487 36 0.0244 30 0 . 0 0 9 2
II 13 5 46 0.0980 40 0.0482 35 0.0238 29 0 . 0 0 8 7
ii 13 6 44 0.0901 39 0.0491 34 0.0240 28 0 . 0 0 8 6

TABLE ] 3 - - H a l p e r i n ' s U ~ statistic (continued).

U~ ~ha UAa ~ha

Mu Nx Rtt 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

II 13 7 43 0.0949 37 0.0452 32 0.0214 27 0.0089

ii 13 R 41 0.0916 36 0.0491 31 0.0233 26 0.0095
II 13 9 39 0.0908 34 0.0479 29 0.0222 24 0,0088
II 13 I0 37 0.0922 32 0,0483 27 0.0219 22 0.0085
II 13 ii 35 0.0961 29 0.0432 25 0.0226 20 0.0085
ii 13 12 32 0.0914 27 0.0460 23 0.0239 18 0,0089
ii 13 13 30 0.0995 24 0.0440 20 0.0219 16 0.0095
ii 13 14 26 0,0884 22 0,0487 17 0.0202 13 0.0090
ii 13 15 24 0.0976 18 0.0435 15 0.0234 Ii 0.0090
II 13 16 19 0.0824 16 0.0457 12 0.0242 7 0.0082
ii 13 17 18 0,0985 ii 0.0439 8 0.0205 6 0.0050
ii 13 18 Ii 0.0829 9 0.0478 7 0.0127 0 0,0
Ii 13 19 9 0,0673 8 0.0303 0 0.0 0 0.0
Ii 13 20 9 0.0673 0 0.0 0 0.0 0 0.0
ii 14 0 52 0.0916 46 0.0477 40 0.0221 34 0.0090
II 14 1 52 0.0916 46 0.0477 40 0.0221 34 0.0090
II 14 2 52 0.0940 46 0.0490 40 0.0227 34 0.0092
ii 14 3 52 0.0989 45 0.0460 40 0.0241 34 0.0098
II 14 4 51 0.0966 44 0.0444 39 0.0229 33 0.0091
II 14 5 50 0.0968 44 0.0498 38 0.0225 32 0.0087
II 14 6 49 0.0993 42 0.0450 37 0.0228 31 0.0087
II 14 7 47 0.0943 41 0.0473 36 0.0239 30 0.0090
ii 14 R 45 0.0915 39 0.0452 34 0.0224 29 0.0097
Ii 14 9 43 0.0910 37 0.0444 32 0.0216 27 0.0092
ii 14 I0 41 0.0927 35 0.0448 30 0.0215 25 0.0089
ii 14 ii 39 0.0967 33 0.0465 28 0.0222 23 0.0090
ii 14 12 36 0.0925 31 0,0497 26 0.0236 21 0.0095
Ii 14 13 33 0.0900 28 0,0478 23 0.0218 18 0.0086
ii 14 14 30 0.0903 2b 0.0463 21 0,0248 16 0.0097
Ii 14 15 27 0.0897 22 0.0470 17 0.0205 13 0.0086
Ii 14 16 24 0,0970 Ig 0.0479 15 0.0223 I0 0.0100
II 14 17 20 0.0868 17 0.0486 ii 0.0246 7 0.0071
II 14 18 19 0.0977 ii 0.0464 8 0.0170 7 0.0071
II 14 19 12 0.0999 9 0.0377 8 0.0170 0 0.0
II 14 20 i0 0.0791 9 0.0377 0 0.0 0 0.0
II 14 21 I0 0.0791 0 0.0 0 0.0 0 0.0
II 15 0 57 0.0991 50 0.0486 44 0.0236 37 0.0088
II 15 1 57 0.0991 50 0.0486 44 0.0236 37 0.0088
II 15 2 56 0.0925 50 0.0500 44 0.0243 37 0.0090
II 15 3 56 0.0972 49 0.0471 43 0.0226 37 0.0095
II 15 4 55 0.0952 48 0.0457 43 0.0246 36 0.0089
II 15 5 54 0.0955 47 0.0455 42 0.0243 35 0.0087
II 15 6 53 0.0980 46 0.0466 41 0.0248 34 0.0087
II 15 7 51 0.0935 45 0.0491 39 0,0227 33 0,0091
II 15 8 49 0.0912 43 0.0472 38 0.0247 32 0.0099
II 15 9 47 0.0909 41 0.0466 36 0.0241 30 0.0094
II 15 I0 45 0.0927 39 0.0473 34 0.0242 28 0.0093
Ii 15 ii 43 0.0966 37 0.0492 31 0.0216 26 0.0095
Ii 15 12 40 0.0928 34 0.0464 29 0.0230 23 0.0083

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 101

TABLE 13--Halperin's U , statistic (continued).

U hct U hct Ig hr U ha
Mu Nx R~ 0.10 Prob 0.05 Prob 0,025 Prob 0 . 0 1 Prob

ii lb 13 37 0.0911 31 0.0445 26 0.0217 21 0.0092

II 15 14 34 0 . 0 9 0 6 28 0.0443 24 0.0245 18 0.00~5
Ii 15 15 31 0.0934 26 0.0495 21 0.0247 15 O.OOg3
ii 15 16 2R 0.0929 22 0.0479 18 0.0243 13 0.0089
Ii 15 17 24 0.0998 19 0.0431 15 0.0243 9 0.0092
Ii 15 18 21 0.0872 17 0.0494 10 0.0211 8 0.0098
ii 15 19 19 0.0935 ii 0.0457 9 0.0217 8 0.0098
II 15 20 12 0.0935 I0 0.0457 9 0.0217 0 0.0
II Ib 21 ii 0.0913 I0 0.0457 0 0.0 0 0.0
ii 15 22 II 0.0913 0 0.0 0 0.0 0 D.o
12 9 0 35 0.0965 30 0.0477 26 0.0245 21 0.0092
12 9 1 35 0.0965 30 0.0477 26 0.0245 21 0.0092
12 9 2 35 0.0988 30 0.0488 25 0.0208 21 0.0093
12 9 3 34 0.0911 29 0.0437 25 0.0218 21 0.0097
12 9 4 34 0.0985 29 0.0473 25 0.0235 20 0.0083
12 9 5 33 0.0958 28 0.0449 24 0.0216 20 0.0092
12 9 6 32 0.0961 27 0.0439 23 0.0206 19 0.0084
12 9 7 31 0.0992 26 0.0446 22 0.0204 18 0.00:40
12 9 8 29 0.0915 25 0.0470 21 0.0211 17 0.0079
12 9 9 27 0.0865 23 0.0427 20 0.0229 16 0.0084
12 9 i0 26 0.0994 22 0.0493 18 0.0206 15 0.0094
12 9 ii 23 0.0846 20 0.0491 17 0.0249 13 0.0083
12 9 12 21 0.0897 18 0.0496 15 0.0248 ii 0.0079
12 9 13 19 0.0975 15 0.0425 12 0.0202 9 0.0080
12 9 14 16 0.0873 13 0.0498 i0 0.0226 7 0.0065
12 9 15 13 0.0907 Ii 0.0498 8 0.0248 4 0.0062
12 9 16 ii 0.0805 6 0.0359 5 0.0211 4 0.0062
12 9 17 6 0.0632 5 0.0211 5 0.0211 0 0.0
12 9 18 6 0.0632 0 0.0 0 0.0 0 0.0
12 I0 0 39 0.0901 34 0.0465 29 0.0213 24 0.00R4
12 i0 i 39 0.0901 34 0.0465 29 0.0213 24 0.0084
12 i0 2 39 0.0923 34 0.0476 29 0.0217 24 0.0086
12 i0 3 39 0.0970 33 0.0432 29 0.0228 24 0.0090
12 I0 4 38 0.0927 53 0.0468 29 0.0246 24 0.0096
12 i0 5 37 0.0909 32 0.0450 28 0.0232 23 0.0087
12 I0 6 36 0.0917 31 0.0447 27 0.0225 22 0.0082
12 i0 7 35 0.0952 30 0.0459 26 0.0228 22 0.0100
12 I0 8 33 0.0892 29 0.0488 25 0.0239 20 O.OORI
12 i0 9 32 0.0987 27 0.0456 23 0.0214 19 0.00R7
12 i0 I0 30 0.09R3 25 0.0442 22 0.0247 18 0.0099
12 I0 Ii 27 0.0866 23 0.0439 20 0.0243 16 0.0094
12 I0 12 25 0.0921 21 0.0458 18 0.0246 14 0.0090
12 I0 13 23 0.0984 19 0.0498 15 0.0208 12 0.0093
12 I0 14 20 0.0975 16 0.0449 13 0.0232 9 0.0072
12 i0 15 17 0.0944 13 0.0433 ii 0.0248 7 0.0091
12 i0 16 14 0.0929 II 0.0433 7 0.0231 5 0.0096
12 I0 17 9 0.0861 7 0.0478 5 0.0096 5 0.0096
12 i0 18 7 0.0779 6 0.0287 0 0.0 0 0.0
12 i0 19 7 0.0779 0 0.0 0 0.0 0 0.0

TABLE 13--Halperin's Un statistic (continued).

Uha Uha Uha Uha

M~ N~ Ru 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

12 Ii 0 44 0.0948 38 0.0454 33 0.0219 28 0.0094

12 ii i 44 0.0948 38 0.0454 33 0.0219 28 0.0094
12 ii 2 44 0.0971 38 0.0465 33 0.0224 28 0.0096
12 ii 3 43 0.0911 38 0.0489 33 0.0236 27 0.00~3
12 II 4 43 0.0983 37 0.0462 32 0.0218 27 0.0090
12 II 5 42 0.0972 36 0.0449 32 0.0244 26 0.0083
12 ii 6 41 0.0986 35 0.0450 31 0.0241 26 0.0096
12 11 7 39 0.0912 34 0.0465 30 0.0246 25 0.0096
12 Ii 8 38 0.0979 33 0.0498 28 0.0219 23 0.0081
12 II 9 36 0.0957 31 0.0474 27 0.0242 22 0.0088
12 II iO 34 0.0958 29 0.0465 25 0.0232 20 0.0081
12 Ii ii 32 0.0991 27 0.0473 23 0.0230 19 0.0096
12 II 12 29 0,0915 24 0.0420 21 0.0238 17 0.0097
12 iI 13 26 0.0871 22 0.0456 18 0.0210 14 O.OORO
i2 iI 14 24 0.0995 19 0.0425 16 0.0232 12 0.0092
I2 ii 15 21 0.0943 16 0.0439 13 0.0217 i0 0.0096
12 ii 16 17 0.0950 14 0.0477 ii 0.0240 6 0.0078
12 II 17 15 0.0883 9 0.0412 7 0.0229 5 0.0046
12 Ii 18 9 0.0843 7 0.0373 6 0.0137 0 0.0
12 11 19 8 0.0932 7 0,0373 0 0.0 0 0.0
12 11 20 8 0,0932 0 0.0 0 0.0 0 0.0
12 12 0 49 0.0989 42 0.0444 37 0.0225 31 0.0086
12 12 i 49 0.0989 42 0.0444 37 0.0225 31 0.0086
12 12 2 48 0.0913 42 0.0455 37 0.0230 31 0.0088
12 12 3 48 0.0958 42 0.0478 37 0.0242 31 0.0092
12 12 4 47 0.0927 41 0.0455 36 0.0226 30 0.0084
12 12 5 46 0.0920 40 0.0446 35 0.0217 30 0.0094
12 12 6 45 0.0937 39 0.0449 34 0.0216 29 0.0091
12 12 7 44 0.0979 38 0.0467 33 0.0222 28 0.0092
12 12 8 42 0.0939 36 0,0436 32 0.0237 27 0.0097
12 12 9 40 0.0923 35 0.0484 30 0.0223 25 0.0088
12 12 I0 38 0.0931 33 0.0481 28 0.0216 23 0.0082
12 12 Ii 36 0.0965 31 0.0493 26 0.0218 22 0.0099
12 12 12 33 0.0905 28 0.0449 24 0.0228 19 0.0081
12 12 13 31 0.0991 26 0.0489 22 0.0248 17 0.0087
12 12 14 28 0.0973 23 0.0473 19 0.0228 15 0.0095
12 12 15 24 0.0873 20 0.0454 16 0.0225 12 0.0092
12 12 16 22 0.0991 17 0.0491 13 0.0195 9 0.00~7
12 12 17 18 0.0991 15 0.0487 9 0.0206 6 0.0069
12 12 18 16 0.0937 9 0.0422 7 0.0186 6 0.0069
12 12 19 9 0.0745 8 0.0466 7 0.0186 0 0.0
12 12 20 8 0.0466 8 0.0466 0 0.0 0 0.0
12 13 0 53 0.0930 47 0.0488 41 0.0229 35 0.0094
12 13 I 53 0.0930 47 0.0488 41 0.0229 35 0.0094
12 13 2 53 0.0953 46 0.0445 41 0.0234 35 0.0096
12 13 3 53 0.0999 46 0.0468 41 0.0246 34 0.0086
12 13 4 52 0.0972 45 0.0448 40 0.0233 34 0.0093
12 13 5 51 0.0969 45 0.0499 39 0.0226 33 0.0088
12 13 6 50 0.0989 43 0.0447 38 0.0227 32 0.0087

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further re
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAl. TESTS 103

TABLE 13--Halperin's U . statistic (continued).

aha Uha Uha Uha

Mr Nx RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

12 13 7 48 0.0933 42 0.0466 37 0.0235 31 0.0088

12 13 8 46 0.0901 40 0.0441 35 0.0216 30 O. 0 0 9 4
12 13 9 45 0.0991 39 0.0488 34 0.0240 28 0.0086
12 13 i0 42 0.0901 37 0.0489 32 0.0237 27 0.0099
12 13 ii 40 0.0935 34 0.0440 30 0.0241 25 0.0099
12 13 12 38 0.0994 32 0.0466 27 0.0215 22 0.0084
12 13 13 35 0.0965 29 0.0441 25 0.0233 20 0.0090
12 13 14 32 0.0961 27 0.0497 22 0.0222 17 0.0081
12 13 15 29 0.0985 24 0.0497 19 0.0214 15 0.0096
12 13 16 26 0.0985 20 0,0449 16 0.0218 12 0.0086
12 13 17 21 0,0913 17 0.0427 14 0~ 8 0.0083
12 13 18 19 0.0969 15 0.0464 9 0.0219 7 0.0097
12 13 19 17 0.0969 9 0.0388 8 0.0242 7 0.0097
12 13 20 i0 0.0888 8 0.0242 8 0.0242 0 0.0
12 13 21 9 0.0565 0 0.0 0 0.0 0 0.0
12 14 0 58 0,0966 51 0.0475 45 0.0232 38 0.0087
12 14 I 58 0.0966 51 0.0475 45 0.0232 38 0.0087
12 14 2 58 0.09~8 51 0.0487 45 0.0238 38 O. 0 0 8 9
12 14 3 57 0.0943 50 0.0458 44 0.0220 38 0.0093
12 14 4 56 0.0921 50 0.0494 44 0.0238 37 0.0087
12 14 5 55 0.0919 49 (}.0490 43 0.0233 37 0.0097
12 14 6 54 0.0940 48 0.0498 42 0.0235 36 0.0097
12 14 7 53 0.0983 46 0.0464 41 0.0244 35 0.0099
12 14 8 51 0,0955 45 0.0498 39 0.0228 33 0.0090
12 14 9 49 0.0949 43 0.0489 37 0.0220 32 0.0100
12 14 I0 47 0.0964 41 0.0492 35 0.0218 30 0.0097
12 14 ii 44 0.0904 38 0.0449 33 0.0223 28 0.0098
12 14 12 42 0.0961 36 0.0478 31 0.0236 25 0.0085
12 14 13 39 0.0938 33 0.0458 28 0.0221 23 0.0092
12 14 14 36 0.0941 30 0.0447 26 0.0249 20 0.0085
12 14 15 33 0.0950 27 0.0453 23 0.0242 18 0.0099
12 14 16 29 0.0896 24 0.0453 19 0.0213 15 0.0097
12 14 17 26 0.0907 20 0.0444 17 0.0243 12 0.0096
12 14 18 21 0.0827 18 0.0464 13 0.0250 8 0.0085
12 14 19 20 0.0983 12 0.0443 9 0.0209 7 0.0052
12 14 20 13 0.1000 i0 0.0478 8 0.0130 0 0.0
12 14 21 i0 0,0670 9 0.0304 0 0.0 0 0.0
12 14 22 i0 0.0670 0 0.0 0 0.0 0 0.0
12 15 0 63 0,0998 55 0.0463 49 0.0234 42 0.0093
12 15 1 63 0.0998 55 0.0463 49 0.0234 42 0.0093
12 15 2 62 0.0935 55 0.0474 49 0.0240 42 0.0095
12 15 3 62 0,0978 55 0.0498 48 0.0224 41 0.0087
12 15 4 61 0.0958 54 0.0483 48 0.0242 41 0.0094
12 15 5 60 0,0958 53 0.0480 47 0.0238 40 0.0091
12 15 6 59 0.0981 52 0,0490 46 0.0242 39 0.0091
12 15 7 57 0,0936 50 0.0459 44 0.0221 38 0.0094
12 15 8 55 0,0913 4g 0.0494 43 0.0238 36 0.0086
12 15 9 54 0,0998 47 0.0487 41 0.0231 35 0.0096
12 15 I0 51 0.0925 45 0.0492 39 0.0231 33 0.0094

TABLE 13--Halperin's Un statistic (continued).

U hct U ha Ig hct lt h a
M~ Nx R• 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

12 15 11 49 0.0961 42 0,0454 37 0.0238 31 0.0095

12 15 12 46 0.0926 40 0.0483 34 0.0219 28 0.0084
12 15 13 43 0.0909 37 0.0466 32 0.0240 26 0.0092
12 15 14 40 0.0912 34 0.0461 29 0.0233 23 0.0086
12 15 15 37 0.0927 31 0.0466 26 0.0231 20 0.0083
12 15 16 34 0.0976 28 0.0477 23 0.0235 17 0.0084
12 15 17 30 0.0900 24 0.0458 19 0.0209 15 0.0097
12 15 18 26 0.0931 21 0.0477 17 0.0226 I0 0.0084
12 15 19 22 0.0855 19 0.0483 12 0.0246 8 0.0072
12 15 20 21 0.0961 12 0.0459 9 0.0169 8 0.0072
12 15 21 13 0.0981 i0 0.0572 9 0.0169 0 0,0
12 15 22 ii 0.0778 I0 0.0372 0 0,0 0 0.0
12 15 23 iI 0.0778 0 0.0 0 0.0 0 0.0
13 10 0 43 0.0931 37 0.0441 33 0.0247 27 0,0089
13 10 1 43 0.0931 37 0.0441 33 0.0247 27 0.0089
13 10 2 43 0.0951 37 0.0450 32 0.0215 27 0.0090
13 10 3 43 0.0994 37 0.0471 32 0.0224 27 0.0094
13 10 4 42 0.0951 36 0.0440 32 0.0240 27 0.0100
13 I0 5 41 0.0930 36 0,0486 31 0,0225 26 0,0090
13 i0 6 40 0,0934 35 0,0480 30 0,0217 25 0,0084
13 I0 7 39 0,0962 34 0,0489 29 0,0217 24 0.0082
13 I0 37 0,0900 32 0.0442 28 0.0224 23 0.0082
13 I0 9 36 0,0983 31 0,0480 27 0,0242 22 0,0086
13 10 I0 34 0,0973 29 0,0462 25 0,0225 21 0.0095
13 i0 ii 32 0,0989 27 0,0458 23 0,0216 19 0.00~7
13 I0 12 29 0,0897 25 0.0471 21 0,0217 17 0,0084
13 I0 13 27 0,0967 22 0,0423 19 0,0225 15 0,0083
13 I0 14 24 0,0927 20 0,0472 17 0,0249 13 0,0088
13 I0 15 21 0,0944 17 0,0424 14 0.0224 i0 0,0080
13 i0 16 18 0.0912 14 0.0460 ii 0.0193 8 0.0084
13 I0 17 15 0.0994 12 0.0454 8 0.0237 5 0.0075
13 I0 18 13 0.0886 7 0.0399 6 0,0237 5 0.0075
13 I0 19 7 0.0678 6 0,0237 6 0.0237 0 0.0
13 i0 2O 7 0.0678 0 0.0 0 0.0 0 0.0
13 II 0 48 0.0931 42 0.0467 37 0.0237 31 0.0092
13 Ii 1 48 0.0931 42 0.0467 37 0.0237 31 0.0092
13 ii 2 48 0.0952 42 0.0477 37 0,0242 31 0,0093
13 ii 3 48 0.0995 42 0.0499 36 0.0218 31 0.0097
13 ii 4 47 0.0958 41 0.0471 36 0,0234 30 0,0087
13 Ii 5 46 0,0944 40 0.0457 35 0,0223 30 0,0096
13 II 6 45 0,0953 39 0,0455 34 0,0218 29 0,0092
13 ii 7 44 0,0987 38 0,0467 33 0,0220 28 0,0090
13 II 8 42 0,0937 37 0.0494 32 0,0230 27 0,0093
13 II 9 40 0,0909 35 0.0468 30 0,0211 26 0,0099
13 Ii I0 38 0,0906 33 0.0456 29 0,023.8 24 0.0091
13 ii II 36 0.0928 31 0.0460 27 0.0235 22 0.0085
13 II 12 34 0.0976 29 0.0478 25 0.0240 20 0.0085
13 II 13 31 0.0927 26 0,0440 22 0,0211 18 0,0087
13 Ii 14 28 0,0907 24 0,0486 20 0,0235 16 0,0096

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 10.5

TABLE 13--Halperin's Un statistic (continued).

Uha Uha Uh~ Uha

M, N, RH 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

13 II 15 25 0.0903 21 0.0485 17 0 . 0 2 1 3 13 0 . 0 0 8 2
13 II 16 22 0.0903 18 0.0478 14 0 . 0 2 1 8 Ii 0 . 0 0 9 6
13 Ii 17 18 0.0927 15 0.0481 12 0 . 0 2 3 3 70.OOR4
13 ii iH 16 0.0991 13 O.04RI 70.OIR3 5 0.0034
13 ii 19 i0 0.0916 7 0.0311 6 0.0109 0 0.0
13 II 20 8 0.0815 7 0.0311 0 0.0 0 0.0
13 ii 21 8 0.0815 0 0.0 0 0.0 0 0.0
13 12 0 53 0.0930 47 0.0488 41 0 . 0 2 2 9 35 0 . 0 0 9 4
13 12 i 53 0.0930 47 0.0488 41 0 . 0 2 2 9 35 0 . 0 0 9 4
13 12 2 53 0.0951 47 0.0499 41 0 . 0 2 3 3 35 0 . 0 0 9 6
13 12 3 53 0.0993 46 0.0464 41 0 . 0 2 4 4 35 0 . 0 1 0 0
13 12 4 52 0.0962 46 0.0498 40 0 . 0 2 2 8 34 0 . 0 0 9 1
13 12 5 51 0.0953 45 0.0487 39 0 . 0 2 1 9 33 O . O O R 5
13 12 6 50 0.0966 44 0.0489 38 0 . 0 2 1 7 33 0 . 0 0 9 8
13 12 7 48 0.0903 42 0.0445 37 0 . 0 2 2 1 32 0 . 0 0 9 8
13 12 8 47 0.0961 41 0.0473 36 0 . 0 2 3 3 30 0 . 0 0 8 5
13 12 9 45 0.0942 39 0.0453 34 0 . 0 2 1 8 29 0 . 0 0 9 1
13 12 I0 43 0.0946 37 0.0447 33 0 . 0 2 4 6 27 0 . 0 0 8 5
13 12 II 41 0.0974 35 0.0455 31 0 . 0 2 4 7 25 O . O O R 2
13 12 12 38 0.0917 33 0.0476 2S 0 . 0 2 1 6 23 O . O O S 2
13 12 13 36 0.0995 30 0.0444 26 0 . 0 2 3 2 21 0 . 0 0 8 7
13 12 14 33 0.0980 28 0.0499 23 0 . 0 2 1 3 19 0 . 0 0 9 4
13 12 15 30 0.0987 25 0.0493 21 0 . 0 2 4 9 16 O . O O S 6
13 12 16 26 0.0920 22 0.0488 18 0 . 0 2 5 0 13 O . O O S 3
13 12 17 23 0.0S92 18 0.0481 15 0 . 0 2 4 3 ii O . O O S 8
13 12 18 18 0.0817 15 0.0389 ii 0 . 0 2 3 1 7 0.0088
13 12 19 17 0.0924 I0 0.0440 R 0.0246 6 0.0052
13 12 20 i0 0,0876 8 0,0391 7 0.0149 0 0.0
13 12 21 9 0.0957 8 0.0391 0 0.0 0 0.0
13 12 22 9 0.0957 0 0.0 0 0.0 0 0.0
13 13 0 58 0.0928 51 0.0454 45 0 . 0 2 2 1 39 0 . 0 0 9 5
13 13 i 58 0.092R 51 0.0454 45 0 . 0 2 2 1 39 0 . 0 0 9 5
13 13 2 58 0.0949 51 0.0464 45 0 . 0 2 2 5 39 0 . 0 0 9 7
13 13 3 58 0.0990 51 0.0486 45 0 . 0 2 3 6 38 0 . 0 0 8 7
13 13 4 57 0,0963 50 0.0466 44 0 . 0 2 2 3 38 0 . 0 0 9 4
13 13 5 56 0.0958 49 0.0458 44 0 . 0 2 4 6 37 0 . 0 0 8 9
13 13 6 55 0.0974 48 0.0462 43 0 . 0 2 4 6 36 O . O O S 7
13 13 7 53 0.0919 47 0.0479 41 0 . 0 2 2 0 35 0 . 0 0 8 8
13 15 8 52 0.0979 45 0.0452 40 0 . 0 2 3 4 34 0 . 0 0 9 2
13 13 9 50 0.0966 44 0.0495 38 0 . 0 2 2 1 32 0 . 0 0 8 4
13 13 I0 48 0.0975 42 0.0494 36 0 . 0 2 1 6 31 0 . 0 0 9 5
13 13 Ii 45 0.0907 39 0.0445 34 0 . 0 2 1 8 29 0 . 0 0 9 4
13 13 12 43 0.0959 37 0.0468 32 0 . 0 2 2 8 27 0 . 0 0 9 6
13 13 13 40 0.0932 34 0.0444 30 0 . 0 2 4 5 24 0 . 0 0 8 4
13 13 14 37 0.0923 32 0.0497 27 0 . 0 2 3 2 22 0 . 0 0 9 3
13 13 15 34 0.0940 29 0.0490 24 0 . 0 2 2 9 19 O . O O R S
13 13 16 31 0.0969 25 0.0442 21 0 . 0 2 2 3 16 0 . 0 0 8 5
13 13 17 27 0.0914 22 0.0450 18 0 . 0 2 4 6 13 O . O O R O
13 13 Is 24 0.0989 IS 0.0426 15 0 . 0 2 1 2 i0 0 . 0 0 9 7

TABLE 1 3 - - H a l p e r i n ' s U n statistic ( c o n t i n u e d ) .

u ha I,l hr lt h= Uha
Mu Nx Ru 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

13 13 19 19 0 . 0 8 0 1 16 0.0438 i0 0.0220 7 0.0075

13 13 20 18 0 . 0 9 6 3 I0 0.0438 8 0.0196 7 0.0075
13 13 21 I0 0 . 0 7 6 1 9 0.0478 8 0.0196 0 0.0
13 13 22 9 0.047R 9 0.0478 0 0.0 0 0.0
13 14 0 63 0 . 0 9 2 6 56 0.0472 50 0.0241 43 0 . 0 0 9 7
13 14 I 63 0 . 0 9 2 6 56 0.0472 50 0.0241 43 0 . 0 0 9 7
13 14 2 63 0 . 0 9 4 6 56 0.0483 50 0.0246 43 0 . 0 0 9 9
13 14 3 63 0 . 0 9 8 7 55 0.0455 49 0.0228 42 0 . 0 0 8 9
13 I4 4 62 0 . 0 9 6 3 b5 0.0488 49 0.0246 42 0 . 0 0 9 6
13 14 5 6i 0 . 0 9 6 0 54 0.0482 48 0.0240 41 0 . 0 0 9 2
13 14 6 60 0 . 0 9 7 8 53 0.0488 47 0.0241 40 O . O O g l
13 14 7 58 0 . 0 9 2 9 51 0.0454 46 0.0249 39 0 . 0 0 9 3
13 14 R 57 0 . 0 9 8 9 50 0.0484 44 0.0232 38 0 . 0 0 9 8
13 14 9 55 0 . 0 9 8 1 48 0.0473 42 0.0223 36 0 . 0 0 9 1
i3 14 iO 53 0 . 0 9 9 4 46 0.0475 40 0.0220 34 0 . 0 0 8 8
13 14 Ii 50 0 . 0 9 3 4 44 0.0488 38 0.0224 32 0 . 0 0 8 8
13 I4 12 48 0 . 0 9 8 8 41 0.0457 36 0.0235 30 0 . 0 0 9 1
i3 14 13 45 0 . 0 9 6 7 39 0.0496 33 0.0219 28 0 . 0 0 9 7
13 14 14 42 0 . 0 9 6 7 36 0.0487 3I 0.0245 25 0 . 0 0 9 0
13 14 15 39 0 . 0 9 H 6 33 0.0490 28 0.0242 22 0 . 0 0 8 6
13 14 16 35 0 . 0 9 1 4 29 0.0445 25 0.0245 19 0 . 0 0 8 4
13 14 17 310.OR91 26 0.0452 21 0.0221 i60.OOH8
13 14 18 28 0 . 0 8 9 9 22 0.0468 18 0.023O 13 [).0078
13 14 19 23 0 . 0 9 3 0 19 0.0441 15 0.0214 9 0.0089
13 14 20 21 0 . 0 9 8 1 17 0.0478 i0 0.0227 7 0.0039
13 14 21 19 0 . 0 9 8 1 i0 0.0395 9 0.0248 0 0.0
13 14 22 Ii 0 . 0 8 9 3 9 0.0248 9 0.0248 0 0.0
13 14 23 i0 0 . 0 5 7 0 0 0.0 0 0.0 0 0.0
13 15 0 68 0 . 0 9 2 3 61 0.0489 54 0.0232 47 0 . 0 0 9 7
13 15 i 68 0 . 0 9 2 3 61 0,0489 54 0.0232 47 0 . 0 0 9 7
13 15 2 68 0 . 0 9 4 2 61 0.0499 54 0.0237 47 O . O I O 0
13 15 3 68 0 . 0 9 8 2 60 0.0473 54 0.0249 46 0 . 0 0 9 1
13 15 4 67 0 . 0 9 6 2 59 0.0458 53 0.0238 46 0 . 0 0 9 8
13 15 5 66 0 . 0 9 6 1 58 0.0454 52 0,0234 45 0 . 0 0 9 5
13 15 6 65 0 . 0 9 ~ 0 57 0.0461 51 0.0235 44 0 . 0 0 9 4
13 15 7 63 0 . 0 9 3 6 56 0.0480 50 0.0244 43 0 . 0 0 9 7
13 15 62 0 . 0 9 9 5 54 0.0460 48 0.0230 41 0 . 0 0 8 8
13 15 9 60 0 . 0 9 9 1 52 0.0452 46 0,0223 40 0 . 0 0 9 7
13 15 10 57 0 . 0 9 2 0 50 0.0455 44 0.0221 38 0 . 0 0 9 5
13 15 Ii 55 0 . 0 9 5 2 48 0.0470 42 0.0227 36 0 . 0 0 9 5
13 15 12 52 0 . 0 9 1 9 46 0.0497 40 0.0239 34 0 . 0 1 0 0
13 15 13 50 0 . 0 9 9 1 43 0,0481 37 0,0226 31 0 . 0 0 9 1
13 15 14 47 0 . 0 9 9 4 40 0.0476 34 0.0220 28 0 . 0 0 8 6
13 15 15 43 0 . 0 9 2 2 37 0.0482 31 0.0217 26 0 . 0 0 9 9
13 15 16 40 0 . 0 9 6 4 34 0.0495 28 0.0225 23 0 . 0 0 9 9
13 15 17 36 0 . 0 9 2 3 30 0.0471 25 0.0228 19 0 . 0 0 8 5
13 15 18 32 0 . 0 9 0 2 27 0.0482 2I 0.0222 16 0 . 0 0 ~ 6
13 15 19 29 0 . 0 9 3 4 22 0.0451 19 0.0249 12 0 . 0 0 8 ~
13 15 20 23 O.OR2H 20 0.0469 I3 0.0223 9 0.00~8

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further re
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 107

TABLE 13--Halperin's Ua statistic (continued).

lt hr U hot ll ha ll ha
Mu N~ RH 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

13 15 21 22 0.0981 13 0.0447 I0 0.0211 8 0.0054

13 15 22 14 0.0996 Ii 0.0478 9 0.0133 0 0.0
13 I5 23 Ii 0.0667 IO 0.0306 0 0.0 0 0.0
13 15 24 ii 0.0667 0 0.0 0 0.0 0 0.0
14 ii 0 52 0.0916 46 0.0477 40 0.0221 34 0.0090
14 iI I 52 0.0916 46 0.0477 40 0.0221 34 0.0090
14 ii 2 52 0.0934 46 0.0486 40 0.0225 34 0.0091
14 Ii 3 52 0.0973 45 0.0449 40 0.0234 34 0.0044
14 ii 4 51 0.0936 45 0.0479 40 0.0249 33 O.OOa5
14 ii 5 50 0.0919 44 0,0464 39 0.0237 33 0.0092
14 ii A 49 0.0924 43 0.0460 38 0.0231 32 0.0088
14 ii 7 48 0.0951 42 0.0468 37 0.0232 Bl 0.0086
14 ii 8 46 0.0899 41 0.0491 36 0.0240 30 0.0087
14 Ii 9 45 0.0972 39 0.0463 34 0.0220 29 0.0091
14 II i0 43 0.0965 37 0.0449 33 0.0244 27 0.0082
14 ii ii 41 0.0982 35 0.0448 31 0.0239 26 0.0094
14 ii 12 38 0.0910 33 0.0461 29 0.0241 24 0.0092
14 ii 13 36 0.0977 31 0.0491 26 0.0211 22 0.0094
14 ii 14 33 0.0948 28 0.0461 24 0.0230 lq 0.0079
14 ii 15 30 0.0954 25 0.0442 21 0.0214 17 0.0088
14 ii 16 27 0.0951 22 0.0453 18 0.0203 14 0.0077
14 Ii 17 23 0.0877 19 0.0433 15 0.0211 12 0.0100
14 Ii 18 20 0.0933 15 0.0391 13 0.0225 N 0.0084
14 ii 19 16 0.0843 13 0.0391 8 0.0209 6 0.0087
14 ii 20 15 0.0957 8 0.0435 6 0.0087 6 0.0087
14 ii 21 8 0.0717 7 0.0261 0 0.0 0 0.0
14 ii 22 8 0.0717 0 0.0 0 0.0 0 0.0
14 12 0 58 0.0966 51 0.0475 45 0.0232 38 0.0087
14 12 I 58 0.0966 51 0.0475 45 0.0232 38 0.0087
14 12 2 58 0.0985 51 0.0484 45 0.0236 38 0.0088
14 12 3 57 0.0934 50 0.0451 45 0.0246 38 0.0091
14 12 4 57 0.0993 50 0.0481 44 0.0230 38 0.0097
14 12 5 56 0.0982 49 0.0469 43 0.0221 37 0.0091
14 12 6 55 0.0992 48 0.0449 43 0,0249 36 0.0088
14 12 7 53 0.0928 47 0.0481 41 0.0220 35 0.0087
14 12 8 52 0.0981 45 0,0448 40 0.0229 34 0.0089
14 12 9 50 0.0959 44 0.0485 39 0.0247 33 0.0095
14 12 i0 48 0.0959 42 0.0477 37 0.0238 31 0.0088
14 12 II 46 0.0982 40 0.0482 35 0.0236 29 0.0084
14 12 I2 43 0.0923 37 0.0437 33 0.0241 27 0.0083
i4 I2 13 4i 0.0990 35 0.0467 30 0.0218 25 0.00~6
14 12 14 38 0.0978 32 0.0447 28 0.0238 23 0.0093
14 12 15 35 0.0979 29 0.0439 25 0.0226 20 0.0085
14 I2 16 31 0.0891 26 0.0440 22 0.0227 18 O.OiO0
14 12 17 28 0.0966 23 0.0452 19 0.0222 15 0.0100
14 12 18 25 0.0975 20 0.0497 i6 0.0247 12 0.0093
14 12 19 20 0.0991 16 0.0417 13 0.0209 8 0.0096
14 12 20 17 0.0783 II 0.0441 8 0.0201 6 0.0040
14 12 21 ii 0.0963 8 0.0351 7 0.0120 0 0.0

TABLE 13--Halperin's U, statistic (continued).

gaol U h~ U h~ l,l h~
Mu N= R~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

14 12 22 9 0.0846 8 0.0331 0 0.0 0 0.0

14 12 23 9 0.0846 0 0.0 0 0.0 0 0.0
14 13 0 63 0.0926 56 0.0472 50 0.0241 43 0.0097
14 13 I 63 0.0926 56 0.0472 50 0.0241 43 0.0097
14 13 2 63 0.0944 56 0.0482 50 0.0246 43 0.0098
14 13 3 63 0.0982 55 0.0452 49 0.0227 42 0.0089
14 13 4 62 0.0955 55 0.0482 49 0,0242 42 0.0095
14 13 5 61 0.0947 54 0.0473 48 0.0234 41 0.0090
14 13 6 60 O.09hO 53 0.0475 47 0.0233 40 0.0087
14 13 7 59 0.0993 52 0.0489 46 0.0237 39 0.0087
14 13 8 57 0.0957 50 0.0461 45 0~ 38 0.0090
14 13 9 55 0.0942 48 0.0446 43 0,0236 37 0.0097
14 13 I0 53 0.0947 47 0.0497 41 0.0230 35 0.0091
14 13 II 5I 0.0973 44 0.0448 39 0,0230 33 0.0089
14 13 12 48 0.0925 42 0.0468 37 0.0238 31 0.0090
14 13 13 46 0.0995 39 0.0442 34 0.0219 29 0.0095
14 13 14 43 0.0987 37 0.0490 32 0.0240 26 0.0086
14 13 15 39 0.0895 34 0.0487 29 0,0234 24 0.0098
14 13 16 36 0.0934 3I 0.0497 26 0.0235 21 0.0095
14 13 17 33 0.09h7 27 0.0460 23 0.0237 18 0.0095
14 13 18 29 0.0976 24 0.0466 19 0.0225 15 0.0095
14 13 19 26 0.096~ 19 0.0415 16 0.0214 II 0.0087
14 13 20 20 0.0855 17 0.0415 II 0.0213 8 0.0097
14 13 21 19 0.0960 Ii ~.0464 8 0.0159 7 0.0058
14 13 22 II 0.0903 9 0.0407 8 0.0159 0 0.0
14 13 23 I0 0.0978 9 0.0407 0 0.0 0 0o0
14 I3 24 IO 0.0978 0 0.0 0 0.0 0 0.0
I4 14 0 69 0.0968 61 0.0469 55 0.0249 47 0.0093
14 14 1 69 0.0968 61 0.0469 55 0.0249 47 0.0093
14 14 2 69 0.0987 61 0.0479 54 0.0226 47 0.0095
14 14 3 68 0.0944 61 0,0499 54 0.0236 47 0.0098
14 14 4 67 0.0921 60 0.0481 53 0.0224 46 0.0092
14 14 5 67 0.0997 59 0.0475 53 0.0246 45 0.0088
14 14 6 65 0.0929 58 0.0479 52 0.0246 45 0.0099
14 14 7 64 0.0963 57 0.0495 50 0.0223 43 0.0087
14 14 8 62 0.0933 55 0.0471 49 0.0235 42 0.0091
14 14 9 60 0.0923 53 0.0459 47 0.0225 41 0.0098
14 14 I0 58 0.0931 51 0.0458 45 0.0221 39 0.0094
14 14 II 56 0.0959 49 0.0468 43 0.0223 37 0.0093
14 14 12 53 0.0918 47 0.0491 41 0.0232 35 0.0095
14 14 IB 51 0.0987 44 0.0471 39 0.0249 32 0.0085
14 14 14 48 0.0985 41 0.0461 36 0.0240 30 0.0094
14 14 15 44 0.0909 38 0.0463 33 0.0236 27 0.0089
14 14 16 41 0.0942 35 0.0477 30 0.0238 24 0.0089
14 14 17 37 0.0896 32 0.0498 26 0.0215 21 0.0088
14 14 18 34 0.0980 28 0.0476 23 0.0220 18 0.0096
14 14 19 30 0.0949 24 0.0459 19 0.0214 15 0.0089
14 14 20 26 0.0981 20 0.0447 17 0.0227 I0 0.0087
14 14 21 21 0.0824 18 0.0457 II 0.0232 8 0.0080

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 109

TABLE 13--Halperi.'; Ua statistic (continued).

Uh. Rh= UA~ Hh=

Mu N~ Rn 0.10 Prob 0.05 Prob 0.025 ['rob 0.01 Prob

14 14 22 20 0.0985 II 0.0452 9 0.0204 8 0.0080

14 14 23 ii 0.0774 iO 0.0489 9 0.0204 0 0.0
14 14 24 iO 0.0489 it) 0.0489 0 0.0 0 0.0
14 15 0 74 0.0930 66 0.0466 59 0.0229 51 0.0089
14 15 i 74 0.0930 66 0.0466 59 0.0229 51 0.0089
14 15 2 74 0.0948 66 0.0476 59 0.0234 51 0,009i
14 15 3 74 0.0985 66 0.0495 59 0.0244 51 0.0095
14 15 4 73 0,0964 65 0.0480 58 0,0234 50 0.00R9
14 15 5 72 0.0962 64 0,0475 57 0.0229 50 0.0098
14 15 6 71 0.0978 63 0.0481 56 0.0230 49 0.0097
14 15 7 69 0.0934 62 0.0498 55 0.0236 48 0.0099
14 15 8 68 0.0988 60 0.0477 53 0.0222 46 0.0090
14 15 9 66 0.0982 58 0.0468 52 0.0242 45 0.0098
14 15 i0 64 0.0994 56 0.0470 50 0.0240 43 0,0095
14 15 Ii 61 0.0941 54 0.0482 48 0.0244 41 0.0095
14 15 12 59 0.0991 51 0,0456 45 0.0225 39 0.0098
14 15 13 56 0.0975 49 0.0491 43 0,0242 36 0.0090
14 15 14 53 0.0976 46 0.0485 40 0.0235 34 0.0099
14 15 15 50 0.0998 43 0.0490 37 0.0233 31 0.0096
14 15 16 46 0.0944 39 0.0451 34 0.0239 28 0.0097
14 15 17 42 0.0915 36 0.0475 31 0.0250 25 0.0099
14 15 18 39 0.0970 32 0.0462 27 0.0229 21 0.0088
14 15 19 34 0.0914 29 0.0478 23 0.0225 18 0,0095
14 15 20 31 0.0918 24 0.0485 20 0.0241 15 0.0095
14 i5 21 25 0.0945 21 0.0452 17 0.0225 I0 0.0094
14 15 22 23 0,0990 19 0.0490 Ii 0.0234 8 0.0041
14 15 23 21 0.0990 II 0.0400 9 0.0105 0 0.0
14 15 24 12 0.0897 I0 0.0253 0 0.0 0 0.0
14 15 25 II 0.0575 0 0.0 0 0.0 0 0.0
15 12 0 63 0,0998 55 0.0463 49 0.0234 42 0.0093
15 12 1 63 0.0998 55 0.0463 49 0.0234 42 0.0093
15 12 2 62 0.0930 55 0.0471 49 0.0238 42 0.0094
15 12 3 62 0,0965 55 0,0489 49 0.0247 42 0.0098
15 12 4 61 0,0933 54 0,0466 48 0,0232 41 0.0089
15 12 5 60 0.0920 53 0.0454 47 0.0222 41 0.0097
15 12 6 59 0.0925 52 0.0451 47 0.0247 40 0.0093
15 12 7 58 0.0951 51 0.0459 46 0.0249 39 0.0092
15 12 8 57 0.0998 50 0.0480 44 0.0226 38 0.0093
15 12 9 55 0,0973 48 0.0457 43 0,0241 37 0.0098
15 12 I0 53 0.0969 46 0.0446 41 0.0230 35 0.0090
15 12 Ii 51 0.0987 44 0.0447 39 0.0226 33 0.0086
15 12 12 48 0,0927 42 0.0460 37 0.0229 31 0.0084
15 12 13 46 0.0987 40 0.0487 35 0.0240 29 0.0086
15 12 14 43 0.0968 37 0.0465 B2 0.0222 27 0.0091
15 12 15 40 0.0972 34 0.0455 30 0.0247 24 0.0082
15 12 16 37 0.0991 31 0.0458 27 0.0244 22 0.0095
15 12 17 33 0.0934 28 0.0468 23 0.0207 19 0.0092
15 12 18 30 0.0993 24 0.0439 21 0.0249 16 0.0095
15 12 19 26 0.0957 21 0.0459 17 0.0243 IB 0.0087

TABLE 13--Halperin's Un statistic (continued).

Uha Uha lth~ Uho~

M~ Nx Rn 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

15 12 20 21 0.0923 17 0.0433 14 0.0210 9 0.0098

15 12 21 18 0.0901 15 0.0433 9 0.0232 7 0.0098
15 12 22 II 0.0834 9 0.0466 7 0.0098 7 0.0098
15 12 23 9 0.0752 8 0.0282 0 0.0 0 0.0
15 12 24 9 0.0752 0 0.0 0 0.0 0 0.0
15 13 0 68 0.0923 61 0.0489 54 0.0232 47 0.0097
15 13 i 68 0.0923 61 0.0489 54 0.0232 47 0.0097
15 i3 2 68 0.0940 61 0.0498 54 0.0236 47 0.0099
15 I3 3 68 0.0974 60 0.0468 54 0.0245 46 0.0090
15 13 4 67 0.0947 60 0~ 53 0.0232 46 0.0095
15 13 5 66 0.0938 59 0.0486 52 0.0224 45 0.0090
15 13 6 65 0.0946 58 0.0487 52 0.0249 44 0.0087
15 13 7 64 0.0975 57 0.0498 50 0.0224 43 0.0087
15 13 8 62 0.0938 55 0,0470 49 0.0233 42 0.0089
15 13 9 60 0.0920 53 0,0452 48 0.0250 41 0.0094
15 13 I0 58 0.0921 52 0.0499 46 0.0242 39 0,0088
15 13 II 56 0.0941 49 0.0450 44 0.0241 37 0.0085
15 13 12 54 0.0983 47 0.0467 42 0.0247 35 0.0085
15 13 13 51 0.0953 45 0.0497 39 0.0227 33 0.0088
15 13 14 48 0.0942 42 0.0482 37 0.0247 31 0.0095
15 13 15 45 0.0949 39 0.0478 34 0~ 28 0.0088
15 13 16 42 0.0982 36 0.0485 31 0.0239 25 0.0085
15 13 17 38 0.0929 32 0.0444 28 0.0244 22 0~
15 13 18 35 0.0986 29 0.0475 24 0.0221 19 0.0085
15 13 19 30 0.0909 25 0.0431 21 0.0238 16 0.0094
15 13 20 27 0.0902 21 0.0480 17 0.0211 13 0.0087
15 13 21 21 0.0884 18 0.0449 15 0.0232 8 0.0077
15 13 22 19 0.0824 12 0,0473 9 0,0216 7 0.0046
15 13 23 II 0.0786 9 0,0349 8 0.0131 0 0.0
15 13 24 I0 0.0873 9 0.0349 0 0.0 0 0.0
15 13 25 I0 0.0873 0 0.0 0 0.0 0 0.0
15 14 0 74 0.0930 66 0.0466 59 0.0229 51 0.0089
15 14 I 74 0.0930 66 0.0466 59 0.0229 51 0.0089
15 14 2 74 0.0947 66 0.0475 59 0.0234 51 0.0091
15 14 3 74 0,0982 66 0.0493 59 0.0243 51 0.0094
15 14 4 73 0.0958 65 0.0475 58 0.0231 51 0.0100
15 14 5 72 0.0951 64 0,0468 57 0.0224 50 0.0096
15 14 6 71 0.0963 63 0.0470 57 0.0250 49 0.0093
15 14 7 70 0.0993 62 0.0483 55 0.0227 48 0.0094
15 14 8 68 0.0962 60 0.0459 54 0.0238 47 0.0097
15 14 9 66 0.0949 59 0.0494 52 0.0227 45 0,0090
15 14 I0 64 0.0955 57 0.0491 50 0.0222 44 0.0099
15 14 II 62 0.0980 54 0.0450 48 0.0222 42 0.0098
15 14 12 59 0.0938 52 0,0468 46 0.0230 40 0,0099
15 14 13 56 0.0916 50 0.0500 44 0.0244 37 0.0089
15 I4 i4 54 0.0999 47 0.0489 41 0.0233 35 0.0096
15 14 15 50 0.0923 44 0.0490 38 0.0229 32 0.0092
i5 L4 16 47 0.0955 40 0.0446 35 0.0230 29 0.0090
i5 14 L7 43 0.0910 37 0.0468 32 0.0239 26 0.0091

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 111

TABLE 13 Halperin's Un statistic (continued).

Uh~ Uha Uha Hha

My N~ R~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

15 14 IR 40 0.0994 34 0.0497 2R 0.0222 23 0 . 0 0 9 4

15 14 19 36 0.0966 30 0.0498 25 0.0230 19 0 . 0 0 8 5
15 14 20 31 0.0894 26 0.0452 21 0.0247 16 O . O O R 7
15 14 21 28 0.0937 21 0.0442 18 0.0232 12 0 . 0 0 9 7
15 14 22 22 0.0888 19 0.0439 12 0.0228 8 0.0063
15 14 23 21 0.0990 12 0.0485 9 0.0169 R 0.0063
15 14 24 12 0.0927 10 0.0421 9 0.0169 0 0.0
15 14 25 II 0.0996 i0 0.0421 0 0.0 0 0.0
15 14 26 ii 0.0996 (} 0.0 0 0.0 0 0.0
15 15 0 80 0.0936 72 0.0488 64 0.0227 56 0 . 0 0 9 3
15 15 1 80 0.0936 72 0.0488 64 0.0227 56 0 . 0 0 9 3
15 15 2 RO 0.0953 72 0.0497 6& 0.0231 56 0 . 0 0 9 4
15 15 3 80 0.09R7 71 0.0472 64 0.0240 56 0 . 0 0 9 8
15 15 4 79 0.0966 71 0.0500 63 0.0230 55 0 . 0 0 9 2
15 15 5 78 0.0962 70 0.0494 63 0,0249 54 O . O O R 9
15 15 6 77 0.0975 69 0.0498 62 0.0250 54 0 . 0 0 9 9
15 15 7 75 0.0932 67 0.0468 60 0.0230 52 0 . 0 0 R 8
15 15 8 74 o,ogsl 66 0.0493 59 0.0241 51 0 . 0 0 9 2
15 15 9 72 0.0972 64 0.04R2 57 0.0232 50 0 . 0 0 9 8
15 15 i0 70 0.09~i 62 0.0482 55 0.0229 48 0 . 0 0 9 5
15 15 Ii 67 0.0930 60 0.0493 53 0.0231 46 0 . 0 0 9 5
15 15 12 65 0.0974 57 0.0466 51 0.0240 44 0 ~
15 15 13 62 0.0956 55 0.049R 48 0.0227 41 0 . 0 0 8 8
15 15 14 59 0.0956 52 0.0491 46 0.0249 39 0 . 0 0 9 6
15 15 15 56 0.0974 49 0.0495 43 0.0247 36 0 . 0 0 9 3
15 15 16 52 0.0924 45 0.0457 39 0.0220 33 0 . 0 0 9 3
15 15 17 49 0.0977 42 O,04RO 36 0.0231 30 0 . 0 0 9 5
15 15 18 45 0.0961 38 0.0461 33 0.0243 26 0 . 0 0 8 5
15 15 19 41 0.0949 34 0.0452 29 0.0236 23 0 . 0 0 9 0
15 15 20 37 0.0995 30 0.0438 25 0.0222 19 0 . 0 0 R 6
15 15 21 33 0.0979 26 0.0465 21 0.0228 17 0 . 0 0 9 7
15 15 22 27 0.0946 22 0~ 19 0.0240 ii 0 ~
15 15 23 23 0.0843 20 0.0474 12 0.0242 90.OOS4
15 15 24 21 0.0843 12 0.0464 i0 0.0211 90.OOR4
15 15 25 12 0.0785 Ii 0.0498 I0 0.0211 0 0.0
15 15 26 ii 0.0498 Ii 0.0498 0 0.0 0 0.0
Note ~ r T a M e l 3 .
For
1=<My<15
[2,My - 3]m~x < N~ < 15
O < R n < N~ + Mu - 2
where
Utt = count o f all X, Y pairs in which a Y precedes an X, corrected for censoring
(see text for counting procedure); the smaller the value o f Un, the less
likely it would occur if Y = X,
My = size o f Y (test) sample,
N , = size o f X reference sample, and
RH = total number o f surviving specimens.
NOTE: When the statistic tabulated has no value such that a significance test is
possible at the desired level, this condition is stipulated by the probability entry 0.0.

2. Given N~ = 6, My = 5, and ry = 2, suppose the following data are

observed

Y: 192, 212 ( = Y2), (3 Y's 212 +) suspended

X: 112, 143, 1 5 1 , 1 7 7 , ( 2 X ' s 212 +) suspended

Consider the null hypothesis Y = X relative to the alternative hypothesis

Y>X.

Answer--Visual inspection does not suffice in this case. Thus we order

the entire collection of data, separating the suspended data from the life
data but retaining the X, Y identity of the individual datum

Life Data
112, 143, 151, 177, 192, 212 ( = y2)
X X X X Y

Suspended Tests
X's: 2
Y's: 3

Next, the test statistic used to refer to statistical tables must be defined for
these data.
Young's Ur statistic is appropriate for a test program terminated at the
occurrence of the ru th Y failure, (namely, for Type II censoring) Table 14.

Young's Uy Statistic
Young [27] devised a modification of Mann and Whitney's U statistic
which accounts for suspended specimens when the test program is termi-
nated early to speed the comparison. His statistic assumes the test program
is terminated at the occurrence of the r~th Y failure. When r~ = 1, the
test program usually is termed a precedence test [28]. The power o f
Young's statistic and precedence tests are discussed in Refs 21, 27, and 28.
Young's Uv statistic is defined as

Uy = UL(life data) + Uc(censored data)

in which UL* is Mann and Whitney's U statistic for the observed life data
(but based on the numbers of times an X precedes a Y in the combined
ordered samplO 9) ignoring the r~th Y failure, and Uc* is the term used to
account for the censored data. Ue* is computed as follows

Uc* = J~(M~ - ry + 1)

x9It is important to note that Young and Halperin define their statistics differently.
Young counts the X's that precede a Y, while Halperin counts the Y's that precede an X.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 1 13

where J~ is the number of failed X specimens, and (My - ry + 1) is the

number of Y specimens with lives exceeding the J~ failed X specimens,
including the ryth Y failure.
ExamplemGiven N~ = 6, My = 5, ry = 2. Suppose the test outcome is
X X X X Y y2 (2 X's, 3 Y's) suspended
The observed value of Ur, termed uy, is 20. (For this outcome, Jx = 4,
UL* = 4, and Ue* = 4(5 - 2 + 1) = 16.)
Example--Given N~ = 15, Mr = 4, and rv = 2. Suppose the test out-
come is
X X X X Y X X X X X X y2 (5 X's, 2 Y's) suspended
In this example, uy = 4 + 10(3) = 34.

Young's Tables
Table 14 consists of eleven columns. The first three columns identify Mr,
ru, and N~, respectively. The remaining eight columns consist of four pairs
of columns, each pair denoted uu and Prob. The four uy columns state the
critical significance values of (uy)~ for which (if the null hypothesis is true)
P r o b [ U r > (uy)~] < a
where a, the size of the Type I error, is equal to 0.10, 0.05, 0.025, and 0.01,
respectively. The four adjacent columns denoted Frob state the exact
size of the significance test for the corresponding values of (uy)~.
Given the tables and the values of ry and uy, the appropriate significance
test of size a is
reject the null hypothesis if uy > (uu)~, or
"accept" the null hypothesis if uy < (uy)~
Example--Given the outcome X X X X Y y.., (2 X's and 3 Y's) suspended,
consider the statistical significance at the 0.10 level.
Answer--For Mu = 5, N~ = 6, and ry = 2, examination of Table 14
indicates that a value of U r equal to 20 or larger would occur 0.0887 • 100
percent of the time by chance if the populations were indeed identical.
This probability is sufficiently small that we usually opt to reject the null
hypothesis and believe instead that Y is better than X.
Example--Given the outcome X X X X Y X X X X X X y2 (5 X's, 2 Y's)
suspended, consider the statistical significance at the 0.10 level.
Answer--For My = 4, N~ = 15, and ry = 2, examination of Table 14
indicates that a value of U r equal to 39 or larger is regarded as significant
statistically at the 0.10 level (exact size 0.0973). Thus, the observed value
uy = 34 does not support a decision to reject the null hypothesis.

TABLE 14---Young's Uy statistic, a

My r~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

1 I 9 9 0.1000 0 0.0 0 0.0 0 0.0

I 1 10 i0 0.0909 0 0.0 0 0.0 0 0.0
1 i II ii 0.0833 0 0.0 0 0.0 0 0.0
I 1 12 12 0.0769 0 0.0 0 0.0 0 0.0
1 I 13 13 0.0714 0 0.0 0 0.0 0 0.0
i 1 14 14 0.0667 0 0.0 0 0.0 0 0.0
1 1 15 15 0.0625 0 0.0 0 0.0 0 0.0
2 1 3 5 0.1000 0 0.0 0 0.0 0 0.0
2 2 3 6 0.1000 0 0.0 0 0.0 0 0.0
2 I 4 7 0.0667 0 0.0 0 0.0 0 0.0
2 2 4 R 0.0667 0 0.0 0 0.0 0 0.0
2 i 5 9 0.0676 9 0.0476 0 0.0 0 0,0
2 2 5 9 0.0952 i0 0.0476 0 0.0 0 0.0
2 i 6 II 0.0357 Ii 0.0557 0 0.0 0 0.0
2 2 6 Ii 0.0714 12 0.0557 0 0.0 0 0.0
2 I 7 II 0.0835 13 0.0278 0 0.0 0 0.0
2 2 7 13 0,0556 14 0.0278 0 0.0 0 0.0
2 i n 13 0.0667 15 0.0222 15 0.0222 0 0.0
2 2 8 14 0.0889 15 0,0444 16 0.0222 0 0.0
2 I 9 15 0.0545 17 0.0182 17 0.0182 0 0.0
2 2 9 16 0.0727 17 0.0364 18 0.0182 0 0.0
2 I I0 15 0.0909 17 0.0455 19 0.0152 0 0.0
2 2 10 17 0.0909 19 0.0303 20 0.0152 0 0.0
2 I II 17 0.0769 19 0.0585 21 0.0128 0 0.0
2 2 ii 19 0.0769 21 0.0256 22 0.0128 0 0.0
2 I 12 19 0.0659 21 0.0330 23 0.0110 0 0.0
2 2 12 20 0.0989 22 0.0440 23 0.0220 0 0.0
2 I 13 19 0.0952 23 0.02R6 25 0.0095 25 0.0095
2 2 13 22 0.0857 24 0.0381 25 0.0190 26 0.0095
2 i 14 21 0.0833 23 0.0500 25 0.0250 27 0.0083
2 2 14 23 0, i 0 0 0 25 0.0500 27 0.0167 2R 0.0083
2 I 15 23 0.0735 25 0.0441 27 0.0221 29 0.0074
2 2 15 25 0,0882 27 0.0441 29 0.0147 30 0.0074
3 i 2 4 0.1000 0 0.0 0 0.0 0 0.0
3 2 2 6 0.I000 0 0.0 0 0.0 0 0.0
3 3 2 6 0.I000 0 0.0 0 0.0 0 0.0
3 i 3 7 0.0500 7 0.0500 0 0.0 0 0.0
3 2 3 8 0.1000 9 0.0500 0 0.0 0 0.0
3 3 3 8 0,I000 9 0.0500 0 0.0 0 0,0
3 I 4 I0 0.0286 I0 0.0286 0 0.0 0 0.0
3 2 4 I0 0.0857 12 0.0286 0 0.0 0 0,0
3 3 4 II 0.0571 12 0.0286 0 0.0 0 0.0
3 I 5 i0 0.0714 13 0.0179 13 0.0179 0 0,0
3 2 5 13 0.0536 14 0.0357 15 0.0179 0 0.0
3 3 5 13 0.0714 14 0.0357 15 0.0179 0 0,0
3 I 6 13 0.0476 13 0.0476 16 0.0119 0 0.0
3 2 6 15 0.0714 16 0.0357 17 0.0238 0 0,0
3 3 6 15 0.0833 16 0.0476 17 0.0238 0 0.0
3 I 7 13 0.0833 16 0.0333 19 0.0083 19 0.0083

~See n o t e at end o f table, p. 129.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 4 O N DISTRIBUTION-FREE STATISTICAL TESTS 1 15

TABLE 14--Young's UY statistic (continued).

Ilia ll~10t Uy a llltct

Mu ru N= 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

3 2 7 16 0.i000 18 0.0500 19 0.0250 21 0 . 0 0 8 3

3 3 7 17 0.0917 19 0.0333 20 0.0167 210.OOg3
3 I 8 16 0.0606 19 0.0242 19 0.0242 22 0 . 0 0 6 1
3 2 8 19 0.0727 21 0.0364 22 0.0182 24 0 . 0 0 6 1
3 3 8 19 0.0970 21 0.0424 22 0.0242 24 0 . 0 0 6 1
3 I 9 16 0.0909 19 0.0455 22 0.0182 25 0 . 0 0 4 5
3 2 9 21 0.0818 23 0.0409 25 0.0136 26 0 . 0 0 9 1
3 3 9 22 0.0727 23 0.0500 25 0.0182 26 0 . 0 0 9 1
3 i I0 19 0.0699 22 0.0350 25 0.0140 28 0 . 0 0 3 5
3 2 i0 23 0.0839 25 0.0420 27 0.0210 29 0 . 0 0 7 0
3 3 I0 24 0.0804 26 0.0385 27 0.0245 29 0.007[)
3 I Ii 19 0.09A2 25 0.0275 28 0.0110 31 0 . 0 0 2 7
3 2 Ii 25 0.0824 27 0.0495 29 0.0247 31 0 . 0 0 8 2
3 3 ii 26 0.0852 28 0.0440 30 0.0192 32 0 . 0 0 5 5
3 1 12 22 0.0769 25 0.0440 28 0.0220 31 0 . 0 0 8 8
3 2 12 27 0.0879 30 0.0396 32 0.0198 34 0 . 0 0 6 6
3 3 12 28 0.0901 31 0.0352 32 0.0242 34 0 . 0 0 8 8
3 I 13 22 0.1000 28 0.0357 31 0.0179 34 0 . 0 0 7 1
3 2 13 29 0.0893 32 0.0429 34 0.0214 37 0.0[)54
3 3 13 30 0.0946 33 0.0411 35 0.0196 37 0 . 0 0 7 1
3 1 14 25 0.0R24 31 0.0294 34 0.0147 37 0 . 0 0 5 9
3 2 14 31 0.0882 34 0.0441 37 0.0176 39 0 . 0 0 8 8
3 3 14 32 0.0985 35 0.0456 37 0.0235 40 0 . 0 0 5 9
3 i 15 28 0.0686 31 0.0429 34 0.0245 40 0 . 0 0 4 9
3 2 15 33 0.0919 36 0.0490 39 0.0221 42 0 . 0 0 7 4
3 3 15 35 0.0821 38 0.0380 40 0.0196 42 0 . 0 0 8 6
4 i 2 5 0.0667 0 0.0 0 0.0 0 0.0
4 2 2 8 0.0667 0 0.0 0 0.0 0 0.0
4 3 2 8 0.0667 0 0,0 0 0.0 0 0.0
4 4 2 8 0,0667 0 0.0 0 0.0 0 0.0
4 I 3 9 0.0286 9 0.0286 0 0.0 0 0.0
4 2 3 I0 0.0857 12 0.0286 0 0.0 0 0.0
4 3 3 ii 0.0571 12 0.0286 0 0.0 0 0.0
4 4 3 Ii 0.0571 12 0.0286 0 0.0 0 0.0
4 i 4 9 0.0714 13 0.0143 13 0.014~ 0 0.0
4 2 4 13 0.0571 14 0.0429 16 0.0143 0 0.0
4 3 4 13 0.0857 15 0.0286 16 0.0143 0 0.0
4 4 4 13 0.I000 15 0.0286 16 0o0143 0 0.0
4 I 5 13 0.0397 13 0.0397 17 0.0079 17 0 . 0 0 7 9
4 2 5 15 0.0952 17 0.0317 18 0.0238 20 0 . 0 0 7 9
4 3 5 16 0.0873 17 0.0476 19 0.0159 20 0 . 0 0 7 9
4 4 5 16 0.0952 18 0.0317 19 0.0159 20 0 . 0 0 7 9
4 i 6 13 0.0714 17 0.0238 17 0.0238 21 0 . 0 0 4 8
4 2 6 17 0.0905 20 0.0381 21 0.0190 23 0 . 0 0 9 5
4 3 6 19 0.0762 21 0.0286 22 0.0190 23 0 . 0 0 9 5
4 4 6 19 0.0857 21 0.0333 22 0.0190 23 0 . 0 0 9 5
4 I 7 17 0.0455 17 0.0455 21 0.0152 25 0 . 0 0 3 0
4 2 7 20 0.0879 22 0.0485 24 0.0242 26 0 . 0 0 9 1
4 3 7 21 0.0970 23 0.0485 25 0.0182 27 0 . 0 0 6 1

TABLE 14---Young's Uy statistic (continued).

UZta U~/a Uy a Uy~,

My ru N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

4 4 7 22 0.0818 24 0.0364 25 0.0212 27 0.0061

4 i 8 17 0.0707 21 0.0303 25 0.0101 29 0.0020
4 2 8 22 0.0970 25 0.0404 27 0.0242 29 O,OORI
4 3 8 24 0,0929 26 0.0485 28 0.0222 30 0.00~I
4 4 8 25 0.0768 27 0.0364 28 0.0242 30 0.0081
4 I 9 17 0,0979 21 0.0490 25 0.0210 29 0.0070
4 2 9 25 0.0811 28 0.0420 30 0.0224 33 0.0056
4 3 9 27 0.0839 29 0.0448 31 0.0224 33 0.0084
4 4 9 27 0.0993 30 0.0378 32 0.0168 33 0.0098
4 i i0 21 0.0699 25 0,0350 29 0.0150 33 0.0050
4 2 i0 27 0.0969 30 0.0500 33 0.0200 36 0.0080
4 3 i0 29 0.0989 32 0.0460 34 0.0240 37 0.0060
4 4 i0 30 0.0939 33 0.0380 35 0o0180 37 0.0070
4 1 II 21 0.0923 29 0.0256 33 0.0110 37 0.0037
4 2 ii 29 0.0974 33 0.0440 36 0.0220 39 0.0088
4 3 ii 32 0.0945 35 0.0440 37 0.0234 40 0.0081
4 4 Ii 33 0.0886 36 0.0388 38 0.0198 40 0.0088
4 i 12 25 0.0692 29 0.0385 33 0.0192 37 0.0082
4 2 12 32 0.0923 36 0.0440 39 0.0220 42 0.0088
4 3 12 35 0.0874 38 0.0440 41 0.0176 43 0.0088
4 4 12 36 0.0852 39 0.0390 41 0.0209 43 0.0099
4 1 13 25 0.0882 33 0.0294 37 0.0147 41 0.0063
4 2 13 34 0.0983 38 0.0500 42 0.0210 45 0.0084
4 3 13 37 0.0996 41 0.0420 44 0.0193 47 0.0067
4 4 13 39 0.0815 42 0.0395 44 0.0223 47 0.0076
4 i I4 29 0.0686 33 0.0412 37 0.0229 45 0.0049
4 2 14 37 0.0879 41 0.0454 45 0.0196 48 0.0098
4 3 14 40 0.0954 44 0.0425 47 0.0196 50 0.0078
4 4 14 41 0.0961 45 0.0395 47 0.0232 50 0.0088
4 I 15 29 0.0851 57 0.0325 41 0.0181 45 0.0090
4 2 15 39 0.0973 44 0.0449 48 0.0206 52 0.0077
4 3 15 43 0.0895 47 0.0413 50 0.0206 53 0.0083
4 4 15 44 0.0924 48 0.0400 50 0.0243 53 0.0098
5 I 2 6 0.0476 6 0.0476 0 0.0 0 0.0
5 2 2 9 0.0952 i0 0.0476 0 0.0 0 0.0
5 3 2 9 0.0952 i0 0.0476 0 0.0 0 0.0
5 4 2 9 0.0952 I0 0.0476 0 0.0 0 0.0
5 5 2 9 0,0952 i0 0.0476 0 0.0 0 0,0
5 1 3 II 0.0179 II 0,0179 II 0.0179 0 0.0
5 2 3 11 0.0714 14 0.0357 15 0.0179 0 0.0
5 3 3 13 0.0714 14 0.0357 15 0.0179 0 0.0
5 4 3 13 0.0714 14 0.0357 15 0.0179 0 0.0
5 5 3 13 0.0714 14 0.0357 15 0.0179 0 0.0
5 I 4 Ii 0.0476 Ii 0.0476 16 0.0079 16 0.0079
5 2 4 15 0.0714 16 0.0397 IR 0.0238 20 0.0079
5 3 4 16 0.0714 17 0.0476 19 0.0159 20 0.0079
5 4 4 16 0.0873 18 0.0317 19 0.0159 20 0.0079
5 5 4 16 0.0952 IR 0.0317 19 0.0159 20 0.0079
5 i 5 ii 0.0833 16 0.0238 16 0.0238 21 0.0040

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 117

TABLE 14--Young's UY statistic (continued).

My a nila lly~ Uy a
My r~ Nx 0.I0 Prob 0.05 Prob 0.025 Prob 0.01 Prob

5 2 5 17 0.0873 20 0 . 0 3 9 7 21 0.0198 24 0.0079

5 3 5 19 0.0833 21 0 . 0 3 5 7 22 0.0238 24 0.0079
5 4 b 20 0.0714 21 0 . 0 4 3 7 23 0.0159 24 0.0079
5 5 5 20 0.0754 21 0 . 0 4 7 6 23 0.0159 24 0.0079
5 1 6 16 0.0455 16 0 . 0 4 5 5 21 0.0130 26 0.0022
5 2 6 20 0.0887 22 0 . 0 4 9 8 25 0.0216 27 0.0087
5 3 6 22 0.08A6 24 0 . 0 4 7 6 26 0.0195 28 0.0087
5 4 6 23 0.0823 25 0 . 0 3 9 0 26 0.0238 28 0.0087
5 5 6 23 0.0887 25 0 . 0 4 1 1 27 0.0152 28 0.0087
5 1 7 16 0.0707 21 0 . 0 2 6 5 26 0.0076 26 0.0076
5 2 7 22 0.0960 26 0 . 0 3 5 4 29 0.0189 31 0.0063
5 3 7 25 0.0909 28 0 . 0 4 0 4 30 0.0189 32 0.0076
5 4 7 26 0.0909 28 0 . 0 4 9 2 30 0.0227 32 0.0088
5 5 7 27 0.0745 29 0 . 0 3 6 6 30 0.0240 32 0.0088
5 1 8 16 0.0979 21 0 . 0 4 3 5 26 0.0163 31 0.0047
5 2 8 25 0.0940 29 0 . 0 4 4 3 31 0.0225 35 0.0078
5 3 8 28 0~ 31 0 . 0 4 3 5 33 0.0249 36 0,0070
5 4 R 29 0.0995 32 0 . 0 4 2 7 34 0.0210 36 0,0085
5 5 8 30 0.0855 32 0 . 0 4 6 6 34 0.0225 36 0.0093
5 1 9 21 0.0629 26 0 . 0 2 8 0 31 0.0105 36 0.0030
5 2 9 28 0.0929 31 0 . 0 4 8 0 35 0.0220 38 0.0100
5 3 9 31 0.0949 34 0 . 0 4 9 5 37 0.0215 40 0.0075
5 4 9 33 0.0844 36 0 . 0 3 7 5 38 0.0195 40 0.0090
5 5 9 33 0.0949 36 0 . 0 4 1 5 38 0.0210 40 0.0095
5 I i0 21 0.0839 26 0 . 0 4 2 0 31 0.0186 36 0.0070
5 2 i0 30 0.0986 35 0 . 0 4 3 6 38 0.0243 41 0.0100
5 3 i0 34 0.0976 38 0 . 0 4 3 3 41 0.0190 44 0.0073
5 4 i0 36 0.0909 39 0 . 0 4 4 6 41 0,0250 44 0.0090
5 5 I0 37 0.0823 39 0 . 0 4 9 6 42 0.0200 44 0.0097
5 1 11 26 0.0577 310.02RR 36 0.0128 41 0.0048
5 2 11 33 0.0964 38 0 . 0 4 4 9 41 0.0234 46 0.0069
5 3 ii 37 0.0989 41 0 . 0 4 6 2 44 0.0234 47 0.0098
5 4 ii 39 0.0968 43 0 . 0 4 0 1 45 0.0234 48 0.0089
5 5 ii 40 0.0902 43 0 . 0 4 4 9 46 0.0190 48 0.0096
5 I 12 26 0.0747 31 0 , 0 4 0 7 36 0.0204 41 0.0090
5 2 12 36 0.0882 41 0 . 0 4 2 2 45 0.0221 49 0.0097
5 3 12 41 0.0850 45 0 . 0 4 1 4 48 0.0215 51 0.0092
5 4 12 43 0.0850 46 0 . 0 4 5 7 49 0.0218 52 0.0089
5 5 12 43 0.0973 47 0 . 0 4 0 9 49 0.0242 52 0.0097
5 I 13 26 0.0924 36 0 . 0 2 9 4 41 0.0147 46 0.0065
5 2 13 38 0.0997 44 0 . 0 4 6 2 48 0.0240 53 0.0088
5 3 13 44 0.0877 48 0 . 0 4 5 2 51 0.0242 55 0.0091
5 4 13 46 0.0900 50 0 . 0 4 1 7 53 0.0205 56 0.0088
5 5 13 47 0.0867 50 0 ~ 53 0.0230 56 0.0097
5 1 14 31 0.0681 36 0 . 0 3 9 7 41 0.0217 51 0.0048
5 2 14 41 0.0921 46 0 . 0 4 9 5 51 0.0233 56 0.0090
5 3 14 47 0,0891 51 0 . 0 4 7 6 55 0.0224 59 0.0088
5 4 14 49 0.0947 53 0 . 0 4 6 7 56 0.0246 60 0.0088
5 5 14 50 0.0930 54 0 . 0 4 3 5 57 0.0218 60 0.0097

TABLE 14---Young's Uy statistic (continued).

U~'a ////a //1,'a U//a

Mr r~ N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

5 i 15 31 0.0830 41 0.0298 46 0.0163 51 0.0081

5 2 15 44 0.0939 50 0.0453 55 0.0220 60 0.0090
5 3 15 50 0.0904 55 0.0432 59 O.020R 63 0.0086
5 4 15 52 0.0989 57 0.0428 60 0.0232 64 0.0087
5 5 15 53 0,0987 57 0.0491 61 0.0209 64 0.0097
6 I 3 7 0.0R33 13 0.0119 13 0.0119 0 0.0
6 2 3 13 0.0476 13 0.0476 17 0.023R 0 O.O
6 3 3 14 0.0952 16 0.0476 17 0.0238 0 0.0
6 4 3 15 0.0833 16 0.0476 17 0.0238 0 0.0
6 5 3 15 0.0833 16 0.0476 17 0.0238 0 0.0
6 6 3 15 0.0833 16 0.0476 17 0.0238 0 0.0
6 1 4 13 0.0333 13 0.0333 19 0.0048 19 0.0048
6 2 4 16 0.0952 iR 0.0476 19 0.0238 23 0.0095
6 3 4 18 O.ORlO 20 0.0429 22 0.0190 23 0.0095
6 4 4 19 0.0714 21 0.0333 22 0.0190 23 0.0095
6 5 4 19 0.0810 21 0.0333 22 0.0190 23 0.0095
6 6 4 19 0.0857 21 0.0333 22 0.0190 23 0.0095
6 I 5 13 0.0606 19 0.0152 19 0.0152 25 0.0022
6 2 5 18 0.0996 22 0.0455 24 0.023R 27 0.0087
6 3 5 21 0.0952 24 0.0411 26 0.0195 28 0.0087
6 4 5 22 0.0996 25 0.0346 26 0.0238 28 0.0087
6 5 5 23 0,0844 25 0.0390 27 0.0152 28 O.OOR7
6 6 5 23 0.08R7 25 0.0411 27 0.0152 28 0.0087
6 i 6 13 0.0909 19 ~.0303 25 0,0076 25 0.0076
6 2 6 22 0.0887 25 0.0400 28 0.()238 31 0.0065
6 3 6 25 0.0812 28 0.0411 30 0.0216 32 0.0097
6 4 6 26 0.0942 29 0.0390 31 0.0173 33 0.0076
6 5 6 27 0.0844 29 0.0444 31 0.0195 33 0.0076
6 6 6 27 0.0898 29 0.0465 31 0,0206 33 0.0076
6 1 7 19 0.0490 19 0,0490 25 0.0163 31 0.0041
6 2 7 24 0,0979 29 0.0425 31 0.0221 36 0.0070
6 3 7 28 0.0956 31 0.0466 34 0.0233 37 0.0070
6 4 7 30 0.0903 33 0.0414 35 0.0216 37 0.0093
6 5 7 31 0.0833 33 0.0478 35 0.0245 38 0.0070
6 6 7 3I 0.0903 34 0.0367 36 0.0175 38 0.0070
6 1 8 i9 0.0699 25 0.0280 31 0.0093 31 0.0093
6 2 S 27 0.0979 31 0.0463 36 0.0196 39 0.0097
6 3 8 31 0.0976 35 0.0470 38 0.0230 41 0.0093
6 4 8 34 0.0R56 37 0.0413 39 0.0243 42 O.OOR7
6 5 8 35 0.0836 37 0.0500 40 0.0203 42 0.0097
6 6 8 35 0.0~06 38 0.0406 40 0.0213 42 0.0100
6 I 9 19 0.0923 25 0.0420 31 0.0168 37 0.0056
6 2 9 30 0.0979 35 0.0490 39 0.0230 43 0.0080
6 3 9 35 0.0917 39 0.0454 42 0.0234 46 0~
6 4 9 37 O.09Rl 41 0.0430 44 0.0198 47 0.0074
6 5 9 39 0.0829 42 0.0410 44 0.0234 47 0.0084
6 6 9 39 0.0905 42 0.0440 44 0.024~ 47 0.0088
6 l 10 25 0.0577 31 0.0262 37 0.0105 43 0.0035
6 2 i0 33 0.0944 38 0.0463 43 0.0189 48 0.0076

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS I 19

T A B L E 14--Young's Uy statistic (continued).

uu= Uw, uu= uv,,

Mu rv N, 0.10 Prob 0.05 Prob 0.025 Prob 0.01 P rob

6 3 10 38 0.0984 43 0.0422 46 0.0242 50 0.0090

6 4 10 41 0.0943 45 0.0441 48 0.0217 51 0.0092
6 5 10 43 0.0820 46 0.0431 49 0.0195 52 0.0076
6 6 10 43 0.0903 46 0.0467 49 0.0210 52 0.0080
6 1 11 25 0.0747 31 0.0373 37 0.0170 43 0.0068
6 2 11 36 0.0928 42 0.0430 46 0.0248 51 0.0099
6 3 11 42 0.0904 47 0,0415 50 0.0242 54 0.0099
6 4 11 45 0.0909 49 0.0440 52 0.0232 56 0.0081
6 5 11 46 0.0978 50 0.0451 53 0,0221 56 0.0095
6 6 11 47 0.0901 50 0.0491 53 0.0238 57 0.0073
6 1 12 25 0.0924 31 0.0498 37 0.0249 49 0.0045
6 2 12 38 0.0995 45 0.0468 49 0.0241 55 0.0088
6 3 12 45 0.0958 50 0.04RI 54 0.0243 59 0.0083
6 4 12 49 0~ 53 0.0447 56 0.0247 60 0.0095
6 5 12 50 0.0960 54 0.0467 57 0.0243 61 0.0085
6 6 12 51 0.0899 55 0.0415 58 0.0207 61 0.0091
6 1 13 31 0.0632 37 0.0341 43 0.0170 49 0.0077
6 2 13 42 0.0898 48 0.0464 54 0.0212 59 0.0099
6 3 13 49 0.0883 54 0.0462 58 0.0246 63 0.0092
6 4 13 52 0.0970 57 0.0451 61 0.0210 65 0.0084
6 5 13 54 0.0943 58 0.04~i 62 0.0212 66 0.0078
6 6 13 55 0.0896 59 0.0437 62 0.0231 66 0.0084
6 i 14 31 0.0775 37 0.0443 43 0.0238 55 0.0054
6 2 14 44 0.0975 51 0.0491 57 0.0235 63 0.0097
6 3 14 52 0.0941 5~ 0.0449 62 0.0245 67 0.0096
6 4 14 56 0.0936 61 0.0451 65 0.0224 69 0.0096
6 5 14 58 0.0928 62 0.0493 66 0.0231 70 0.0093
6 6 14 59 0.0893 63 0.0457 67 0.0204 71 0.0077
6 1 15 31 0.0922 43 0.0316 49 0.0170 55 0.0085
6 2 15 47 0.0981 54 0.0495 61 0.0210 67 0.0089
6 3 15 55 0.0971 61 0.0486 66 0.0245 72 0.0087
6 4 15 60 0.0906 65 0.0455 69 0.0236 74 0.0087
6 5 15 62 0.0915 67 0.0426 70 0,0249 75 0.0084
6 6 15 63 0.0890 67 0.0474 71 0.0224 75 0.0092
7 1 4 15 0.0242 Ib 0.0242 15 0,0242 22 0.0030
7 2 4 15 0.0879 21 0.0333 22 0.0152 26 0.0091
7 3 4 20 0.0758 22 0.0394 25 0.0182 27 0.0061
7 4 4 21 0.0848 23 0.0455 25 0.0212 27 0.0061
7 5 4 22 0.0727 24 0.0364 25 0.0212 27 0.0061
7 1 5 15 0.0455 15 0.0455 22 0.0101 29 0.0013
7 2 5 20 0.0985 22 0.0455 27 0.0227 29 0.0076
7 3 5 23 0.0960 27 0.0366 29 0.0189 32 0.0076
7 4 5 25 0.0922 28 0.0404 30 0.0202 32 0.0088
7 5 5 26 0.0871 28 0.0480 30 0.0227 32 0.0088
7 1 6 15 0.0699 22 0.0210 22 0.0210 29 0.0047
7 2 6 22 0.0862 27 0.0495 31 0.0216 35 0.0076
7 3 6 27 0.0892 31 0.0420 34 0.0186 36 O. 0093
7 4 6 29 0.0938 32 0.0478 35 0.0198 37 0.0093
7 5 6 30 0.0979 33 0.0443 35 0.0233 38 O. 0 0 7 0

T A B L E 14---Young's UY statistic (continued).

Ill/~ 141tct Ulla l,l~la

My r~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

7 1 7 15 0.0962 22 0.0350 29 0.0105 36 0.0023

7 2 7 27 0.0839 31 0.0452 35 0.0207 39 0.0093
7 3 7 31 0.0892 35 0.0425 38 0.0224 41 0.0099
7 4 7 33 0.0970 37 0.0420 40 0.0192 43 0.0067
7 5 7 35 0.0857 38 0.0417 40 0.0233 43 0.0079
7 I 8 22 0.0513 29 0.0186 29 0.0186 36 0.0056
7 2 8 29 0.0839 34 0.0491 39 0.0219 43 0.0079
7 3 8 34 0.0971 39 0.0443 42 0,0246 46 0.0098
7 4 8 37 0.0985 41 0.0472 44 0.0238 48 0.0076
7 5 8 39 0,0929 42 0.0497 45 0.0233 48 0.0092
7 I 9 22 0.0692 29 0.0288 36 0.0105 43 0.0031
7 2 9 32 0.0986 37 0.0497 43 0.0184 48 0.0093
7 3 9 38 0.0940 43 0.0437 47 0.0226 51 0.0090
7 4 9 42 0.0862 46 0.0431 49 0.0226 53 0.0082
7 5 9 43 0.0988 47 0.0462 50 0.0228 53 0.0100
7 1 I0 22 0.0882 29 0.0407 36 0.0170 43 0.0062
7 2 10 35 0.0947 41 0.0494 46 0.0249 52 0.0090
7 3 I0 42 0.0903 47 0.0466 51 0.0237 56 0.0087
7 4 i0 46 0.0883 50 0.0444 54 0.0217 58 0.0084
7 5 i0 48 0.0892 52 0.0435 55 0.0227 59 0.0080
7 i 11 29 0.0539 36 0.0249 36 0.0249 50 0.0038
7 2 Ii 38 0.0962 44 0.0486 50 0.0214 56 0.0095
7 3 ii 45 0.0981 51 0.0461 56 0.0214 61 0.0087
7 4 ii 50 0.0892 55 0.0426 59 0.0208 63 0.0086
7 5 ii 52 0.0943 56 0.0492 60 0.0224 64 0.0085
7 1 12 29 0.0681 36 0.0341 43 0.0157 50 0.0065
7 2 12 41 0.0982 48 0.0499 54 0.0249 61 0.0092
7 3 12 49 0.0935 55 0.0467 60 0.0235 65 0.0100
7 4 12 54 0.0910 59 0.0461 63 0.0238 68 0.0088
7 5 12 56 0.0988 61 0.0463 65 0.0221 69 0.0091
7 i 13 29 0.0830 36 0.0443 43 0.0221 57 0.0043
7 2 13 43 0.0996 51 0.0481 57 0.0238 64 0.0095
7 3 13 53 0.0917 59 0.0478 64 0.0243 70 0.0095
7 4 13 58 0.0920 63 0.0489 68 0.0227 73 0.0090
7 5 13 61 0.0905 66 0.0439 70 0.0219 74 0.0096
7 1 14 29 0.0984 43 0.0295 50 0.0148 57 0.0068
7 2 14 47 0.0988 55 0.0483 62 0.0231 69 0.0097
7 3 14 56 0.0964 63 0.0474 &9 0.0225 75 0.0093
7 4 14 62 0.0928 68 0.0453 73 0.0218 78 0.0090
7 5 14 65 0.0945 70 0.0485 75 0.0216 79 0.0100
7 i [5 36 0,0671 43 0.0377 50 0.0201 64 0.0046
7 2 15 50 0.0913 58 0.0480 65 0.0243 73 0.0100
7 3 15 60 0.0945 67 0.0480 73 0.0240 80 0.0089
7 4 15 66 0.0938 72 0.0479 77 0.0245 83 0.0091
7 5 15 69 0.0981 75 0.0461 80 0.0213 85 0.0084
8 I 5 17 0.0350 17 0.0350 25 0.0070 25 0.0070
8 2 5 22 0.0971 25 0.0319 30 0.0210 33 0.0047
8 3 5 25 0.0816 29 0.0443 31 0.0249 35 0.0093
8 4 5 28 0.0831 31 0.0389 33 0.0218 36 0.0085

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTIONIFREE STATISTICAt TESTS 121

TABLE 14--Young's Uy statistic (continued).

llycl Uy a I,tZl~ Uya

My ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

8 5 5 29 0.0839 32 0.0389 34 0.0202 36 0.0093

8 1 6 17 0.0549 25 0.0150 25 0.0150 33 0.0030
8 2 6 24 0.0909 3[) 0.0443 33 0.0163 38 0.0093
8 3 6 29 0.0932 33 0.0433 37 0.0210 40 0.0093
8 4 6 32 0.0889 36 0.0410 38 0.0246 41 0.0100
8 5 6 33 0.0999 37 0.0413 39 0.0240 42 0.0090
8 1 7 17 0.0769 25 0.0256 33 0.0070 33 0.0070
8 2 7 29 0.0872 32 0.0480 38 0.0219 41 0.0089
8 3 7 33 0.0872 38 0.0429 41 0.0233 46 0.0084
8 4 7 36 0.0970 40 0.0488 44 0.0210 47 0.0093
8 5 7 3R 0.0967 42 0.0443 45 0.0210 48 0.0087
8 1 8 17 O. i O 0 0 25 0.0385 33 0.0128 41 0.0035
8 2 R 31 0.0921 37 0.0464 41 0.0203 47 0.0094
8 3 8 37 0.0904 42 0.0464 46 0.0240 51 0.0088
8 4 8 41 0.0881 45 0.0473 49 0.0217 53 0.0086
8 5 8 43 0.0929 47 0.0456 50 0.0239 54 0.0083
8 i 9 25 0.0529 33 0.0204 33 0.0204 41 0.0068
8 2 9 33 0.0882 40 0.0450 46 0.0226 51 0.0099
8 3 9 41 0.0882 46 0.0485 51 0.0227 56 0.0092
8 4 9 45 0.0940 50 0.0460 54 0.0230 58 0.0099
8 5 9 48 0.0902 52 0.0471 56 0.0213 40 0.0081
8 1 10 25 0.0686 33 0.0294 41 0.0113 49 0.0038
8 2 10 37 0.0988 44 0.0470 49 0.0230 56 0.0095
8 3 i0 45 0.0889 51 0.0441 56 0.0221 61 0.0097
8 4 i0 49 0.0982 55 0.0448 59 0.0237 64 0.0089
8 5 10 52 0.I000 57 0.0475 61 0.0232 65 0.0096
8 1 11 25 0.0851 33 0.0397 41 0.0170 49 0.0065
8 2 Ii 40 0.0921 47 0.0493 54 0.0232 61 0.0093
8 3 ii 48 0.0982 55 0.0461 60 0.0249 66 0.0098
8 4 II 54 0.0921 60 0.0440 64 0.0241 70 0.00~2
8 5 ii 57 0.0958 62 0.0485 67 0.0209 71 0.0092
8 i 12 33 0.0511 41 0.0238 41 0.0238 57 0.0039
8 2 12 43 0.0982 51 0.0469 57 0.0232 65 O.OOa5
8 3 12 52 0.0969 59 0.0479 65 0.0228 71 0.0099
8 4 12 58 0.0958 64 0.0486 69 0.0247 75 0.0091
8 5 12 62 0.0931 67 0.0490 72 0.0225 77 0.0087
8 i 13 33 0.0632 41 0.0316 49 0.0148 57 0.0043
8 2 13 46 0.0991 55 0.0465 62 0.023~ 70 0.0098
8 3 13 56 0.0955 63 0.0494 70 0.0230 77 0.0089
8 4 13 62 0.0993 69 0.0474 75 0.0217 80 0,0099
8 5 13 67 0.0903 72 0.0494 77 0.0240 83 0.0083
8 i 14 33 0.0760 41 0.0402 49 0.0201 57 0.0094
8 2 14 49 0.0953 57 0.0494 65 0.0234 73 0.0099
8 3 14 60 0.0949 68 0.0468 74 0.0247 ~2 0.0090
8 4 14 67 0.0936 74 0.0462 80 0.0221 86 0.0091
8 5 14 71 0.0974 77 0.0496 83 0.0218 88 0.0096
8 1 15 33 0.0892 41 0.0496 57 0.0131 65 0.0061
8 2 15 52 0.0997 62 0.0481 70 0.0244 79 0.0094
8 3 15 64 0.0927 72 0.0474 79 0.0235 87 0.0091

TABLE 14---Young's Uy statistic (continued).

Ulta U~r~ ltVet Ulla

Mu ru Nx 0.10 Prob 0.05 Prob 0.025 ] ? r o b 0.01 Prob
8 4 15 71 0.0964 79 0.0450 85 0.0225 91 0.0098
8 5 15 76 0.0947 82 0.0500 R8 0.0231 94 0.0091
9 i 6 19 0.0440 19 0.0440 28 0.0110 37 0.0020
9 2 6 26 0.0949 28 0.0470 36 0.0182 41 0.0094
9 3 6 31 0.0965 35 0.0462 40 0.0224 44 0.0080
9 4 6 34 0.0923 38 0.0500 42 0.0220 46 0.0076
9 5 6 36 0.0983 40 0.0472 43 0.0232 46 0.0100
9 6 6 38 0.0869 41 0.0462 44 0.0214 47 0.0082
9 1 7 19 0.0625 28 0.0192 28 0.0192 37 0.0048
9 2 7 28 0.0769 35 0.0455 37 0.0245 45 0.0087
9 3 7 35 0.0864 40 0.0497 44 0.0240 50 0.0086
9 4 7 39 0.0928 43 0.0497 48 0.0212 52 0.0082
9 5 7 41 0.0998 46 0.0431 49 0.0233 53 0.0085
9 6 7 43 0.0940 47 0.0447 50 0.0226 53 0.0100
9 I 8 19 0.0824 28 0.0294 37 0.0090 37 0.0090
9 2 8 33 0.0973 37 0.0430 44 0.0232 51 0.0089
9 3 8 39 0.0963 44 0.0491 50 0.0223 55 0.0088
9 4 8 43 0.0981 49 0.0445 53 0.0229 58 0.0086
9 5 8 47 0,0879 51 0.0477 55 0.0228 59 0.0094
9 6 8 49 0.0859 53 0.0432 56 0.0234 60 0.0087
9 I 9 28 0.0412 28 0.0412 37 0.0147 46 0.0045
9 2 9 35 0.0982 43 0.0452 46 0.0249 55 0.0076
9 3 9 43 0.0923 49 0.0491 54 0.0249 61 0.0086
9 4 9 48 0.0963 54 0.0457 59 0.0215 64 0.0085
9 5 9 52 0.0892 57 0.0444 61 0.0224 65 0,0100
9 6 9 54 0.0915 58 0.0496 62 0.0239 66 0.0099
9 1 I0 28 0,0542 37 0.0217 37 0.0217 46 0.0077
9 2 I0 37 0.0913 45 0.0485 53 0.0220 60 0.0097
9 3 I0 47 0.0944 54 0.0453 60 0.0222 66 0,0094
9 4 i0 53 0.0915 59 0.0463 64 0.0232 70 0.0086
9 5 i0 57 0.0906 62 0.0478 67 0.0220 72 0.00R6
9 6 10 59 0,0960 64 0.0475 68 0.0242 73 0.0086
9 I II 28 0,0681 37 0.0298 46 0.0119 55 0,0043
9 2 II 42 0.0988 50 0.0485 55 0.0249 64 0.0092
9 3 Ii 51 0.0939 58 0.0489 64 0.0247 72 0.0086
9 4 II 57 0.0980 64 0.0467 70 0.0218 76 0.00~6
9 5 ii 62 0.0915 68 0.0447 73 0.0215 78 0.0090
9 6 Ii 64 0.0999 70 0.0457 74 0.0243 79 0.0095
9 1 12 28 0.0827 37 0.0389 46 0.0170 55 0.0068
9 2 12 45 0,0900 53 0.0491 61 0.0240 70 0.0092
9 3 12 55 0.0940 63 0.0465 70 0.0225 77 0.0095
9 4 12 62 0.0944 69 0.0475 75 0.0234 81 0.0100
9 5 12 67 0.0921 73 0.0473 78 0.0244 84 0.0094
9 6 12 70 0.0927 76 0.0441 80 0.0244 86 0.0085
9 1 13 28 0.0977 37 0.0489 46 0.0230 64 0.0040
9 2 13 49 0.0937 57 0.0489 64 0.0247 73 0.0097
9 3 13 59 0.0950 67 0.0489 74 0.0245 82 0.0096
9 4 13 66 0.0995 74 0,0476 80 0.0246 87 0~
9 5 13 72 0.0927 78 0.0699 84 0.0238 90 0,0097

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 123

TABLE 14--Young's UY statistic (continued).

My ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 6 13 75 0.0962 81 0.0485 86 0.0245 92 0.0092

9 I 14 37 0.0595 46 0.0297 55 0.0140 64 0.0061
9 2 14 52 0.0927 61 0.0481 70 0.0229 79 0.0098
9 3 14 63 0.0951 72 0.0467 79 0.0248 88 0.0093
9 4 14 71 0.0961 79 0.0479 86 0.0232 93 0.0097
9 5 14 77 0.0932 84 0.0469 90 0.0231 96 0.0099
9 6 14 80 0.0992 87 0.0467 92 0.0245 98 0.0098
9 i 15 37 0.0707 46 0.0372 55 0.0186 64 0.0087
9 2 15 54 0.0962 64 0.0481 73 0.0232 83 0.0096
9 3 15 67 0.0951 76 0.0489 84 0.0245 93 0.0098
9 4 15 76 0.0934 84 0.0484 91 0.0242 99 0.0095
9 5 15 82 0.0934 89 0.0490 96 0.0225 103 0.0088
9 6 15 86 0.0930 93 0.0452 98 0.0245 105 0.0088
I0 1 7 21 0.0515 31 0.0147 31 0.0147 41 0.0034
i0 2 7 30 0.0854 38 0,0430 41 0.0175 49 0.0083
i0 3 7 37 0.0893 42 0,0485 47 0.0245 53 0.0097
i0 4 7 41 0.0940 46 0.0497 51 0.0226 56 0.0089
I0 5 7 44 0.0983 49 0.0474 53 0.0239 58 0.0079
i0 6 7 46 0.0998 51 0.0449 55 0.0203 58 0.0098
i0 I 8 21 0.0686 31 0.0229 31 0.0229 41 0.0065
i0 2 8 31 0.0882 40 0.0430 47 0.0241 51 0.0090
i0 3 8 40 0.0958 47 0.0471 53 0.0232 59 0.0085
i0 4 8 46 0.0934 52 0,0466 57 0.0225 62 0.0098
io 5 8 50 0.0910 55 0.0468 59 0.0249 64 0.0097
i0 6 8 52 0.0965 57 0.0467 61 0.0230 65 0.0099
I0 I 9 21 0.0867 31 0.0325 41 0.0108 51 0.0031
i0 2 9 38 0.0913 41 0.0495 50 0.0227 59 0.0085
iO 3 9 45 0.097R 51 0.0497 58 0.0229 65 0.0091
i0 4 9 51 0.0925 57 0.0481 63 0.0225 69 0.0086
i0 5 9 55 0.0945 61 0.0459 66 0.0222 71 0.0091
I0 6 9 58 0.0933 63 0.0481 68 0.0214 72 0.0098
I0 i i0 31 0.0433 31 0.0433 41 0.0163 51 0.0054
i0 2 I0 40 0.0867 48 0.0495 56 0.0233 61 O. 0 0 9 9
i0 3 I0 49 0.0917 57 0 . 0 4 5 0 63 0.0245 70 0.0098
i0 4 I0 56 0.0915 63 0 . 0 4 5 5 68 0.0248 75 O. 0 0 9 2
i0 5 i0 60 0.0984 67 0 . 0 4 5 0 72 0.0231 78 0.0087
i0 6 I0 64 0.0906 69 0 . 0 4 9 0 74 0.0235 79 0.0097
I0 i ii 31 0.0551 41 0 . 0 2 2 7 41 0.0227 51 0.0085
i0 2 Ii 41 0.0957 51 0 . 0 4 2 6 59 0.0243 69 0.0086
I0 3 ii 53 0,0990 61 0 . 0 4 7 9 68 0.0243 76 0.0098
I0 4 ii 60 0.0988 68 0 . 0 4 6 9 74 0.0243 81 0.0096
i0 5 ii 66 0.0922 72 0 . 0 4 9 8 78 0.0258 84 0.0098
i0 6 ii 69 0.0977 75 0 . 0 4 9 7 81 0.0217 86 0.0095
I0 I 12 31 0.0677 41 0 . 0 3 0 1 51 0.0124 61 0.0046
i0 2 12 47 0.0986 56 0 . 0 4 9 5 64 0.0243 73 0.0096
IO 3 12 57 0.0967 66 0 . 0 4 8 0 74 0.0228 82 0 . 0 0 9 3
iO 4 12 65 0.0970 73 0 . 0 4 8 6 80 0.0236 87 0 . 0 1 0 0
I0 5 12 71 0.0948 78 0 . 0 4 8 5 84 0.0243 91 0.0092
I0 6 12 75 0.0947 82 0 . 0 4 4 7 87 0.0234 93 0.0093

TABLE 14---Young's Ur statistic (continued).

Uxta ~lcl UTla U~la

M~ r~ N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

I0 I 13 31 0.0807 41 0.0382 51 0.0170 61 0.0070

I0 2 13 49 0.0995 59 0.0489 68 0.0245 78 0.0099
i0 3 13 61 0.0076 71 0.0460 79 0.0227 88 0.0092
i0 4 13 70 0.0960 7R 0.0499 86 0.0230 94 0.0091
I0 5 13 76 0.0976 84 0.0474 90 0.0248 98 0.0087
i0 6 13 81 0.0920 88 0.0454 93 0.0249 i00 0.0091
i0 1 14 31 0.0942 41 0.0471 51 0.0223 61 0.0099
i0 2 14 51 0.0967 64 0.0475 71 0.0245 83 0.0094
I0 3 14 65 0.0997 75 0.0491 84 0.0255 93 0.0097
I0 4 14 75 0.0944 84 0.0471 91 0.0246 i00 0.0093
i0 5 14 81 0.0996 90 0.0464 97 0.0224 104 0.0095
i0 6 14 86 0.0977 94 0.0459 I00 0.0232 107 O.OOR9
I0 [ Ib 41 0.0565 51 0.0285 61 0.0134 71 0.0059
i0 2 15 57 0.0952 67 0.0497 77 0.0241 88 0.0098
i0 3 15 69 0.0993 RO 0.0481 89 0.0236 99 0.0094
I0 4 15 79 0.0997 89 0.0481 97 0.0241 106 0.0096
i0 5 15 87 0.0945 95 0.0496 103 0.0229 III 0.0090
I0 6 15 92 0.0950 i00 0.0464 106 0.0244 114 0.0088
II 1 8 23 0.0578 34 0.0181 34 0.0181 45 0.0048
ii 2 8 33 0.0975 43 0.0427 45 0.0237 55 0.0092
II 3 S 42 0.0938 50 0.0477 55 0.0231 63 0.0085
ii 4 8 48 0.0993 55 0.0455 61 0.0213 67 0~
II 5 8 52 0.09~6 59 0.0445 64 0.0217 69 0,0095
II 6 S 56 0.0902 61 0.0474 66 0.0219 71 0.0085
II 7 8 58 0.0911 63 0.0450 67 0.0228 72 0.0081
ii 1 9 23 0.0737 34 0.0260 45 0.0081 45 O.OOSl
Ii 2 9 34 0.0975 44 0.0498 54 0.0209 62 0.0098
ii 3 9 46 0.0988 54 0.0458 61 0.0242 68 0.0100
II 4 9 53 0.0949 60 0.0498 66 0.0248 73 0.0096
ii 5 9 58 0.0962 65 0.0453 70 0.0240 76 0.0096
II 6 9 62 0.0909 68 0.0448 73 0.0218 78 0.0093
11 7 9 64 0.0955 70 0.0441 74 0.0237 79 0.0094
II 1 I0 23 0.0902 34 0.0551 45 0.0124 56 0.0039
Ii 2 I0 42 0.0975 51 0.0493 56 0.0209 66 0.0087
Ii 3 i0 51 0.0963 60 0.0466 66 0.0237 74 0.0099
II 4 I0 58 0.0983 66 0.0472 72 0.0249 80 0.0093
Ii 5 i0 64 0.0921 71 0.0461 77 0.0227 83 0.0097
II 6 I0 68 0.0916 74 0.0477 80 0.0218 85 0.0099
Ii 7 i0 70 0.0989 76 0.0492 81 0.0244 87 0.0088
ii 1 II 34 0.0451 34 0.0451 45 0.0175 56 0.0062
II 2 Ii 44 0,0905 54 0.0459 63 0.0232 72 0.0098
II 3 ii 54 0,0998 64 0.0468 72 0.0254 81 0.0091
II 4 ii 63 0.0963 71 0.0499 79 0.0227 86 0.0099
Ii 5 ii 69 0.0972 77 0.0467 83 0.0242 90 0.0098
II 6 II 74 0.0918 81 0.0451 86 0.0247 93 0.0090
ii 7 ii 77 0.0923 83 0.0477 88 0.0248 94 0.0098
II 1 12 34 0.0559 45 0.0235 45 0.0235 56 0.0092
Ii 2 12 45 0.0955 56 0.0447 66 0.0225 76 0.0099
II 3 12 60 0.0958 69 0.0488 77 0.0236 86 0.0099

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 O N DISTRIBUTION-FREESTATISTICALTESTS 125

T A B L E 14--Young's Uy statistic (continued).

14~a Ny~ U~/a ~ya

Mu ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

Ii 4 12 68 0.0975 77 0.0478 85 0.0226 93 0.0094

ii 5 12 75 0.0936 83 0.0470 90 0.0229 97 0.0098
II 6 12 80 0.0921 87 0.0475 93 0.0242 i00 0.0094
II 7 12 83 0.0952 90 0.0464 96 0.0220 i02 0.0091
ii i 13 34 0.0673 45 0.0303 56 0.0127 67 0.0050
II 2 13 52 0.0985 63 0.0457 72 0.0234 82 0.0098
II 3 13 63 0.0990 74 0.0464 83 0.0228 92 0.0100
ii 4 i3 73 0.0965 82 0.0500 91 0.0228 i00 0.0089
Ii 5 13 80 0.0978 89 0,0474 96 0.0242 104 0.0098
II 6 13 85 0.0997 93 0.0497 i00 0.0238 107 0.0099
ii 7 13 89 0.0978 97 0.0452 i03 0.0224 109 0.0098
ii i 14 34 0.0791 45 0.0377 56 0.0170 67 0.0071
ii 2 14 54 0.0977 65 0.0486 75 0.0248 87 0.0093
Ii 3 14 68 0.0966 78 0.0495 88 0.0233 98 0.0097
ii 4 14 78 0.0967 88 0.0481 96 0.0250 106 0.0095
II 5 14 86 0.0945 95 0.0476 103 0.0230 iii 0.0097
ii 6 14 91 0.0993 i00 0.0471 107 0.0234 I15 0.0090
II 7 14 96 0.0922 103 0.0487 II0 0.0228 117 0.0092
II I 15 34 0.0913 45 0.0457 56 0.0217 67 0.0098
Ii 2 15 56 0.0935 67 0.0494 78 0.0243 92 0.0096
II 3 15 72 0.0976 83 0.0493 93 0.0243 104 0.0098
ii 4 15 83 0.0963 93 0.0498 102 0.0249 113 0.0091
II 5 15 91 0.0979 101 0.0477 i09 0.0240 118 0.0096
II 6 15 97 0.0989 106 0.0489 114 0.0230 122 0.0093
ii 7 15 102 0.0944 II0 0.0474 117 0.0230 124 0.0098
12 i 9 25 0.0632 37 0.0211 37 0.0211 49 0.0062
12 2 9 37 0.0805 47 0.0498 57 0.0226 61 0.0090
12 3 9 47 0.0941 57 0.0440 64 0.0249 71 0.0098
12 4 9 55 0.0977 63 0.0500 70 0.0229 78 0.0088
12 5 9 61 0.0940 68 0.0479 74 0.0246 81 0.0094
12 6 9 65 0.0947 72 0.0458 77 0.0245 84 0.0085
12 7 9 68 0.0948 74 0.0483 79 0.0245 85 0.0092
12 i i0 25 0.0779 37 0.0287 49 0.0096 49 0.0096
12 2 i0 45 0.0929 49 0.0433 59 0.0248 70 0.0093
12 3 i0 54 0.0944 61 0.0469 69 0.0237 79 0.0089
12 4 10 60 0.0996 69 0.0469 76 0.0247 84 0.0099
12 5 I0 67 0.0920 75 0.0455 81 0.0239 89 0.0088
12 6 IO 7i 0.0974 79 0.0449 85 0.0223 91 0.0097
i2 7 I0 75 0.0920 81 0.0491 87 0.0231 93 0.0093
i2 I iI 25 0.0932 37 0.0373 49 0.0137 61 0.0046
12 2 II 47 0.0883 57 0.0477 61 0.0240 73 0.0084
12 3 ii 57 0.0928 67 0.0469 75 0.0244 84 0.0097
12 4 Ii 66 0.0940 75 0.0474 82 0.0247 91 0.0097
12 5 ii 72 0.0988 81 0.0468 88 0.0237 96 0.0093
12 6 ii 77 0.0994 85 0.0489 92 0 . 0 2 3 1 99 0 . 0 0 9 4
12 7 ii 81 0.0977 88 0 . 0 4 9 8 94 0 . 0 2 4 9 101 0 . 0 0 9 4
12 1 12 37 0.0466 37 0.0466 49 0 . 0 1 8 6 61 0 . 0 0 6 9
12 2 12 48 0.0937 59 0.0487 70 0 . 0 2 2 5 81 0 . 0 0 9 2
12 3 12 61 0.0947 71 0.0477 80 0 . 0 2 4 7 91 0 . 0 0 9 2

T A B L E 14--Young's Uy statistic (continued).

Uvc~ Uua llv~ U~/a

M. r~ Nx 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

12 4 12 71 0.0943 80 0.0493 89 0.0233 98 0.0097

12 5 12 78 0.0967 87 0.0484 95 0.0232 i03 0.0097
12 6 12 84 0.0938 92 {).0478 99 0.0237 107 O.00ql
12 7 12 88 0.0948 96 0.0455 102 0.0235 109 0.0094
12 i 13 37 0.0565 49 0.0242 49 0.0242 61 0.0097
12 2 13 49 0.0969 61 0.0464 72 0.0240 84 0.0094
12 3 13 66 0.0985 77 0.0479 86 0.0241 97 0.0092
12 4 13 76 0.0963 86 0.0489 95 0.0237 105 0.0094
I2 5 13 84 0,0953 93 0.0494 101 0.0249 iii 0.0090
12 6 i3 90 0.0956 99 0.0469 106 0.0242 114 0.0099
12 7 13 94 0 . 0 9 9 5 i03 0.0460 109 0.0249 117 0.0094
I2 I 14 37 0.0670 49 0.0304 61 0.0130 73 0.0052
12 2 14 57 0.0983 69 0.0464 79 0.0243 91 0.0097
12 3 14 69 0.0998 RI 0.04~6 91 0 . 0 2 4 7 103 0.0095
12 4 14 81 0.0961 92 0.04~4 i01 0 . 0 2 4 3 112 0.0093
12 5 14 89 0 . 0 9 9 6 I00 0.0469 lOR 0 . 0 2 4 5 llR 0.0093
12 6 14 96 0 , 0 9 7 1 105 0,0499 I13 0 . 0 2 4 6 122 0.0095
12 7 14 I01 0 . 0 9 6 5 Ii0 0.0465 117 0 . 0 2 3 6 125 0.0094
12 I 15 37 0.0778 49 0.0372 61 0 , 0 1 6 9 73 0.0072
12 2 15 59 0.0961 71 0.0483 83 0 . 0 2 2 6 95 0.0097
12 3 15 74 0.0998 86 0.0499 97 0 , 0 2 3 6 109 0.0095
12 4 15 86 0.0975 97 0.0498 i07 0 . 0 2 4 6 118 0.0099
12 5 15 95 0 . 0 9 7 9 106 0.0479 115 0 . 0 2 3 9 125 0.0096
12 6 15 102 0 . 0 9 8 4 112 0.0488 120 0 . 0 2 5 0 130 0.0092
12 7 15 108 0 . 0 9 4 0 117 0.0469 124 0 . 0 2 4 7 133 0.0093
13 1 i0 27 0.0678 40 0.0237 40 0 . 0 2 3 7 53 0.0075
13 2 i0 40 O,ORR6 52 0.0454 63 0 . 0 2 3 1 74 0.0092
13 3 i0 52 0.0988 63 0.0472 73 0 . 0 2 2 7 82 0.0099
13 4 i0 62 0,0967 72 0.0472 80 0 . 0 2 3 6 88 0.0100
13 5 i0 69 0.0987 78 0.0478 85 0 . 0 2 4 3 94 0.0089
13 6 i0 74 0.0995 83 0.0457 89 0 . 0 2 4 5 97 0.0091
13 7 i0 78 0.0999 86 0.0473 92 0 . 0 2 4 2 99 0.0094
13 8 I0 81 0.0997 88 0.0496 94 0 . 0 2 4 1 I01 0.0087
13 1 ii 27 0.0815 40 0.0311 53 0 . 0 1 0 9 66 0.0034
13 2 ii 49 0.0991 53 0.0481 65 0 . 0 2 3 3 77 0.0096
13 3 ii 60 0.0936 70 0.0478 77 0 . 0 2 3 8 88 0.0093
13 4 ii 68 0.0984 78 0.0470 86 0 . 0 2 3 9 96 0.0094
13 5 ii 75 0.0973 84 0.0497 92 0 . 0 2 4 7 101 0.0096
13 6 II 81 0.0960 90 0.0462 97 0 . 0 2 3 4 105 0.0093
13 7 II 85 0.0991 93 0.0496 I00 0 . 0 2 4 0 108 0.0088
13 8 Ii 89 0.0936 96 0.0484 102 0 . 0 2 4 7 109 0.0097
13 i 12 27 0.0957 40 0.0391 53 0 . 0 1 4 9 66 0.0052
13 2 12 51 0.0924 63 0.0450 73 0 . 0 2 4 3 79 0.0100
i3 3 12 62 0.0991 74 0.0474 84 0 . 0 2 4 3 95 0.0096
I3 4 12 73 0.0959 84 0.0468 93 0 . 0 2 3 4 I03 0.0093
13 5 12 81 0,0977 91 0.0481 99 0 . 0 2 4 7 109 0.0093
13 6 12 87 0.0992 97 0.0465 104 0 . 0 2 4 8 113 0.0095
13 7 12 92 0 . 0 9 8 4 i01 0.0472 108 0 . 0 2 3 7 116 0.0093
13 8 12 96 0 . 0 9 5 8 104 0.0473 iii 0 . 0 2 2 4 118 0.0092

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER ~, ON DISTRIBUTION-FREE STATISTICAL TESTS 127

T A B L E 14--Young's UY statistic (continued).

Uya U11a Uva Uya

Mu ru N, 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

13 I 13 40 0.0478 40 0.0478 53 0.0196 66 0.0075

13 2 13 52 0.0963 65 0.0438 76 0.0243 89 0.0092
13 3 13 67 0.0989 78 0.0495 89 0,0235 LOl 0.0092
13 4 IB 78 0.0993 80 0.0494 99 0.0239 llO 0.0095
13 5 iB 87 0.0967 97 0.0498 106 0.024~ 116 0.0099
13 6 13 94 0.0958 104 0.0468 112 0.0236 121 0.0095
13 7 IB 99 0.0977 IOH 0.0490 116 0.0235 124 0.0098
13 8 13 i03 0.0977 112 0.0462 119 0.0229 127 0.0088
13 1 14 40 0.0570 53 0.0248 53 0.0248 79 0.0039
13 2 14 53 0.0981 66 0.0478 79 0.0214 92 O.00HR
13 3 14 73 0.0921 85 0.0465 95 0.0246 108 0.0093
13 4 14 84 0.0948 95 0.0496 105 0.0248 117 0.0005
13 5 14 93 0.0965 104 0.0482 113 0.0248 124 0.0095
13 6 14 i00 0.0986 ill 0.0469 119 0.0247 129 0.0096
13 7 14 106 0.0970 [16 0.0467 124 0.0232 133 0.0091
13 8 14 ii0 0.0993 119 0.0493 127 0.0233 135 0.0096
13 1 15 40 0.0667 55 0.0306 66 0.0133 79 0.0054
13 2 15 62 0.0981 75 0.0469 86 0.0249 i00 0.0094
13 3 15 76 0.0954 89 0.0476 i00 0.0245 113 0.0099
13 4 15 89 0.0958 i01 0.0486 112 0.0237 124 0.0095
13 5 15 99 0.0958 ii0 0.0497 120 0.0248 132 0.0092
13 6 15 107 0.0954 118 0.0469 127 0.0236 137 0.0096
13 7 15 113 0.0962 123 0.0482 132 0.0229 141 0.0095
13 8 15 118 0.0941 127 0.0481 135 0.0236 144 0.0091
14 I ii 29 0.0717 45 0.0241 57 0.0087 57 0.0087
14 2 ii 43 0.0957 57 0.0391 69 0.0225 Rl 0.0100
14 5 ii 61 0.0988 69 0.0488 80 0.0247 92 0.0097
14 4 Ii 69 0.0991 80 0.0481 90 0.0235 I00 0.0097
14 5 Ii 78 0.0948 88 0.0479 96 0.0248 106 0.0095
i4 6 ii 84 0.0979 94 0.0465 102 0.0230 Iii 0.0090
14 7 Ii 89 0.0978 98 0.0481 105 0.0248 114 0.0090
14 8 ii 93 0.0971 i01 0.0494 108 0.0243 116 0.0092
14 i 12 29 0.0846 43 0.0331 57 0.0120 71 0.0040
14 2 12 54 0.0913 66 0.0470 71 0.0209 84 0.0093
14 3 12 66 0.0909 77 0.0497 85 0.0249 97 0.0098
14 4 12 76 0.0949 87 0.0478 96 0.0241 107 0.0097
14 5 12 84 0.0969 95 0.0466 104 0.0234 114 0.0096
14 6 12 91 0.0956 I01 0.0477 109 0.0248 119 0.0094
14 7 12 96 0.0990 106 0.0471 114 0.0231 122 0.0100
14 8 12 I01 0.0941 109 0.0498 117 0.0232 125 0.0093
14 i 13 29 0.0978 43 0.0407 57 0.0159 71 0.0058
14 2 13 55 0.0960 68 0.0478 80 0.0240 93 0.0095
14 3 13 68 0.0968 81 0.0468 92 0.0250 105 0~
14 4 13 80 0.0982 92 0.0494 I03 0.0240 115 0.0094
14 5 13 90 0.0966 I01 0.0493 iii 0.0240 122 0.0095
14 6 13 97 0.0998 108 0.0487 117 0.0243 127 0.0098
14 7 13 103 0.0998 113 0.0499 122 0.0236 131 0.0097
14 8 13 108 0.0979 i18 0.0461 125 0.0246 134 0.0094
14 1 14 43 0.0489 43 0.0489 57 0.0204 71 0.0080

TABLE 14--Young's Uy statistic (continued).

u~ u~ u~ uya
My ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

14 2 14 56 0.0985 70 0.0457 83 0.0227 97 0.0089

14 3 14 75 0.0968 87 0.0484 97 0.0243 IiI 0.0093
14 4 14 86 0.0988 99 0.0473 109 0.0245 122 0.0094
14 5 14 96 0.0974 108 0.0482 i18 0.0244 130 0.0094
14 6 14 104 0.0977 115 0.0496 125 0.0239 136 0.0092
14 7 14 iii 0.0947 121 0.0488 130 0.0240 140 0.0095
14 8 14 116 0.0950 126 0.0465 134 0.0235 143 0.0095
14 I 15 43 0.0575 57 0.0253 71 0.0105 85 0.0041
14 2 15 57 0.0990 71 0.0490 85 0.0225 99 0.0095
14 3 15 79 0.0940 92 0.0485 104 0.0247 118 0.0096
14 4 15 91 0.09R6 104 0.0499 116 0.0242 129 0.0097
14 5 15 102 0.0976 115 0.0474 125 0.024R 138 0.0093
14 6 15 iii 0.0959 123 0.0472 133 0.0235 144 0.0095
14 7 15 i18 0.0956 129 0.0477 138 0.0244 149 0.0092
14 8 15 123 0.0982 134 0.0467 142 0.0246 152 0.0095
15 I 12 31 0.0752 46 0.0282 61 0.0098 61 0.0098
15 2 12 57 0.0901 61 0.0433 75 0.0210 89 0.0087
15 3 12 68 0.0986 76 0.0495 88 0.0236 I01 0.0098
15 4 12 78 0.0977 89 0.0479 99 0.0246 112 0.0092
15 5 12 86 0.0986 98 0.0476 108 0.0238 119 0.0095
15 6 12 94 0.0968 105 0.0479 114 0.0242 125 0.0091
15 7 12 I00 0.0974 II0 0.0494 119 0.0239 129 0.0091
15 8 12 105 0.0967 115 0.0466 123 0.0232 132 0.0091
15 9 12 109 0.0955 I18 0.0474 125 0.0248 134 0.0092
15 I ].3 31 0.0873 46 0.0349 61 0.0131 76 0.0046
15 2 13 58 0.0957 72 0.0449 76 0.0232 91 0.0087
15 3 13 71 0.0969 85 0.0451 96 0.0245 109 0.0096
15 4 13 83 0.0961 96 0.0475 106 0.0249 119 0.0095
15 5 13 93 0.0963 105 0.04R0 115 0.0245 127 0~
15 6 13 I01 0.0961 112 0.0493 122 0.0243 133 0.0097
15 7 13 107 0.0996 118 0.0494 127 0.0250 138 0.0093
15 8 13 113 0.0957 123 0.0482 132 0.0230 141 0.0096
15 9 13 117 0.0968 127 0.0465 135 0.0230 144 0.0090
15 1 14 31 0.0996 46 0.0421 61 0.0169 76 0.0063
15 2 14 59 0.0990 74 0.0439 87 0.0232 I01 0.0097
15 3 14 74 0.0937 88 0.0472 I01 0.0231 115 0.0094
15 4 14 88 0.0965 i01 0.0488 113 0.0240 126 0.0100
15 5 14 99 0.0960 112 0.0473 123 0.0236 135 0.0098
15 6 14 108 0.0954 120 0.0478 130 0,0244 142 0.0094
15 7 14 115 0.0961 126 0.0492 136 0.0240 147 0.0094
15 8 14 121 0.0947 131 0.0495 140 0.0248 151 0.0091
15 9 14 125 0.0979 135 0.0492 144 0.0234 153 0.0097
15 1 15 46 0.0498 46 0.0498 61 0.0211 76 0.0084
15 2 15 61 0.0843 75 0.0474 89 0.0240 104 0.0097
15 3 15 81 0.0998 95 0.0495 106 0.0244 121 0.0098
15 4 15 94 0.0982 108 0.0485 120 0.0240 134 0.0095
15 5 15 105 0.0980 I18 0.0499 130 0.0242 143 0.0098
15 6 15 iI4 0.0995 I27 0.0491 138 0.0244 150 O.OiO0
15 7 15 122 0.0980 134 0.0491 144 0.0249 156 0.0095

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREESTATISTICAL TESTS 129

TABLE 1 4 - - Y o u n g ' s UY statistic (continued).

Uyc~ Hy a Hycr I~iya

Mu ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob
15 8 15 128 0.0992 140 0.0474 149 0.0245 160 0.0095
15 9 15 133 0,0987 144 0.0480 153 0.0237 163 0.0094
Note for Table 14.
For
I-<M~<15
[2,Mu - 3].... =< Nx =< 15
1 _-< ru =< (Mu + 3)/2 when Mu > 6
where
Ur = count of all X, Ypairs in which an Xprecedes a Y, corrected for censoring
(see text for counting procedure); the larger the value of Ur, the less likely
it would occur if Y = X,
My = size of (test) sample,
N~ = size of reference sample, and
r u = r th failure of Y (the signal to terminate the overall test program).
NOTE: When the statistic tabulated has no value such that a sigmficance test is
possible at the desired level, this condition is stipulated by the probability entry 0.0.

APPENDIX I

Use of Precedence Tables to Compare Proportions Surviving a Fixed Test Duration

Table 15 may be used to compare proportions surviving a test program in which
all of the suspended tests pertain to the same fixed duration, say N* cycles. In this
analysis, the observed lives o f the failed specimens are considered to be of secondary
interest.

Tables
Table 15 has the same format as Table 14, except the test statistic Jx refers to the
number of X failures which precede the hypothetical r th Y failure. If the observed
value, ix, is larger than the tabulated value (jxL, we reject the null hypothesis
(of equal proportions failing) at the a level, that is
Prob [J~ > (ix)o] =<
The value of r~ is established as follows. Let Su equal the number of Y run-outs.
Then, r u = M u + 1 - S u.
E x a m p l e - - S u p p o s e M u = 10, Nx = 15, L = 14, and S~ = 6. Test the null
hypothesis Y = X at the 0.10 significance level versus the alternative hypothesis
Y > X.
A n s w e r - - W i t h r~ = (I0 + 1 - 6) = 5, examination of Table 15 indicates that
12 or more X failures is regarded as significant statistically at the 0.10 level. The
observed value, Jx = 14, is significant statistically even at the 0.01 level (exact size
0.0068). Thus, we opt to reject the null hypothesis and believe instead that Y is
better than X.

TABLE 15 Precedence tables. ~

M~ r~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

2 1 3 3 0.i000 0 0.0 0 0.0 0 0.0

2 1 4 4 0.0667 0 0.0 0 0.0 0 0.0
2 i 5 5 0.0476 5 0.0476 0 0.0 0 0.0
2 1 6 6 0.0357 6 0.0357 0 0.0 0 0.0
2 1 7 6 0.0833 7 0.0278 0 0.0 0 0.0
2 i 8 7 0.0667 8 0.0222 8 0.0222 0 0.0
2 1 9 8 0.0545 9 0.0182 9 0.0182 0 0.0
2 I i0 R 0.0909 9 0.0455 i0 0.0152 0 0.0
2 1 11 9 0.0769 10 0 . 0 3 8 5 11 O.O12R 0 0.0
2 i 12 I0 0.0659 Ii 0 . 0 3 3 0 12 0.0110 0 0.0
2 i 13 10 0.0952 12 0 . 0 2 8 6 13 0.0095 13 0.0095
2 I 14 ii 0.0833 12 0 . 0 5 0 0 13 0.0250 14 0.0083
2 I 15 12 0.0735 13 0 . 0 4 4 1 14 0.0221 15 0.0074
3 i 2 2 0.i000 0 0.0 0 0.0 0 0.0
3 1 3 3 0.0500 3 0.0500 0 0.0 0 0.0
3 1 4 4 0.0286 4 0.0286 0 0.0 0 0.0
3 i 5 4 0.0714 5 0.0179 5 0.0179 0 0.0
3 i 6 5 0.0476 5 0.0476 6 0.0119 0 0.0
3 2 6 6 0.0833 0 0.0 0 0.0 0 0.0
3 1 7 5 0.0833 6 0.0333 7 0.0083 7 0.0083
3 2 7 7 0,0667 0 0.0 0 0.0 0 0.0
3 i 8 6 0.0606 7 0.0242 7 0.0242 8 0.0061
3 2 8 8 0.0545 0 0.0 0 0.0 0 0.0
3 1 9 6 0.0909 7 0.0455 8 0.0182 9 0.0045
3 2 9 9 0.0455 9 0.0455 0 0.0 0 0.0
3 1 10 7 0.0699 8 0.0350 9 0.0140 I0 0.0035
3 2 10 i0 0.0385 i0 0 . 0 3 8 5 0 0.0 0 0.0
3 1 ii 7 0.0962 9 0.0275 I0 0.0110 ii 0.0027
3 2 ii I0 0.0934 ii 0 . 0 3 3 0 0 0.0 0 0.0
3 I 12 8 0.0769 9 0,0440 i0 0.0220 ii 0.0088
3 2 12 II 0.0813 12 0 . 0 2 8 6 0 0.0 0 0.0
3 i 13 8 0.I000 10 0 . 0 3 5 7 ii 0.0179 12 0.0071
3 2 13 12 0.0714 13 0 . 0 2 5 0 13 0.0250 0 0.0
3 1 14 9 0~ ii 0 . 0 2 9 4 12 0.0147 13 0.0059
3 2 14 13 0.0632 14 0 . 0 2 2 1 14 0.0221 0 0.0
3 1 15 I0 0.0686 Ii 0 . 0 4 2 9 12 0.0245 14 0.0049
3 2 15 14 0.0564 15 0 . 0 1 9 6 15 0.0196 0 0.0
4 i 2 2 0.0667 0 0.0 0 0.0 0 0.0
4 I 3 3 0.0286 3 0.0286 0 0.0 0 0.0
4 i 4 3 0.0714 4 0.0145 4 0.0143 0 0.0
4 2 4 4 0.0714 0 0.0 0 0.0 0 0.0
4 1 5 4 0.0397 4 0.0397 5 0.0079 5 0.0079
4 2 5 5 0,0476 5 0.0476 0 0.0 0 0.0
4 1 6 4 0.0714 5 0.0238 5 0.0238 6 0.0048
4 2 6 6 0.0333 6 0.0333 0 0.0 0 0.0
4 i 7 5 0.0455 5 0.0455 6 0.0152 7 0.0030
4 2 7 6 0~ 7 0.0242 7 0.0242 0 0.0
4 I 8 5 0.0707 6 0.0303 7 0.0101 8 0.0020
4 2 8 7 0.0667 8 0.0182 8 0.0182 0 0.0

"See note at end of table, p. 144.

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS ] 31

T A B L E 15--Precedence tables (continued).

J~ jx~ jx. j~
Mu ru N= 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

4 3 R 8 0.0~09 0 0.0 0 0.0 0 0.0

4 1 9 5 0.0979 6 0.0490 7 0.0210 8 0.0070
4 2 9 8 0.0517 9 0.0140 9 0.0140 0 0.0
4 3 9 9 0.0769 0 0.0 0 0,0 0 0.0
4 i 10 6 0.0699 7 0.0350 8 0.0150 9 0.0050
4 2 I0 8 0.0949 9 0.0410 I0 0.0110 0 0.0
4 3 10 I0 0.0659 0 0.0 0 0.0 0 0.0
4 i II 6 0.0923 R 0.0256 9 0.0110 i0 0.0037
4 2 Ii 9 0.0769 I0 0.0330 ii 0.0088 II 0.0088
4 3 ii 11 0.0571 0 0.0 0 0.0 0 0.0
4 1 12 7 0.0692 8 0.0385 9 0.0192 i0 0.0082
4 2 12 I0 0.0632 ii 0.0269 12 0.0071 12 0.0071
4 3 12 12 0.0500 12 0.0500 0 0.0 0 0.0
4 i 13 7 0.0882 9 0.0294 10 0.0147 11 0.0063
4 2 13 I0 0.0987 12 0.0223 12 0.0223 13 0.0059
4 3 13 13 0.0441 13 0.0441 0 0.0 0 0.0
4 1 14 8 0.06R6 9 0.0412 10 0.0229 12 0.0049
4 2 14 ii 0.0833 12 0.0441 13 0.0186 14 0.0049
4 3 14 14 0.0392 14 0.0392 0 0.0 0 0.0
4 I 15 8 0.0~51 I0 0.0325 ii 0.0181 12 0.0090
4 2 15 12 0.0709 13 0.0374 14 0.0157 15 0.0041
4 3 15 14 0.0970 15 0.0351 0 0.0 0 0.0
5 i 2 2 0.0476 2 0.0476 0 0.0 0 0.0
5 1 3 3 0.0179 3 0.0179 3 0.0179 0 0.0
5 2 3 3 0.0714 0 0.0 0 0.0 0 0.0
5 i 4 3 0.0476 3 0.0476 4 0.0079 4 0.0079
5 2 4 4 0.0397 4 0.0397 0 0.0 0 0.0
5 1 5 3 0.0833 4 0.0238 4 0.0238 5 0.0040
5 2 5 5 0.0238 5 0.0238 5 0.0238 0 0.0
5 3 5 5 0.0835 0 0.0 0 0.0 0 0.0
5 i 6 4 0.0455 4 0.0455 5 0.0130 6 0.0022
5 Z 6 5 0.0671 6 0.0152 6 0.0152 0 0.0
5 3 6 6 0.0606 0 0.0 0 0.0 0 0.0
5 1 7 4 0.0707 5 0.0265 6 0.0076 6 0.0076
5 2 7 6 0.0455 6 0.0455 7 0.0101 0 0.0
5 3 7 7 0.0455 7 0.0455 0 0.0 0 0.0
5 I 8 4 0.0979 5 0.0435 6 0.0163 7 0.0047
5 2 8 6 0.0862 7 0.0319 8 0.0070 8 0.0070
5 3 8 8 0.0350 8 0.0350 0 0.0 0 0.0
5 i 9 5 0.0629 6 0.0280 7 0.0105 8 0.0030
5 2 9 7 0.0629 8 0.0230 8 0.0230 9 0.0050
5 3 9 8 0.0949 9 0.0275 0 0.0 0 0.0
5 I I0 5 0.0839 6 0.0420 7 0.0186 8 0.0070
5 2 I0 8 0.0470 8 0.0470 9 0.0170 i0 0.0037
5 3 i0 9 0,0769 i0 0.0220 i0 0.0220 0 0.0
5 4 i0 i0 0.0952 0 0.0 0 0.0 0 0.0
5 1 Ii 6 0.0577 7 0.0288 8 0.0128 9 0.0048
5 2 ii 8 0.0769 9 0.0357 i0 0.0128 II 0.0027
5 3 ii I0 0.0632 ii 0.0179 ii 0.0179 0 0.0

T A B L E 15--Precedence tables (continued).

J~ J~ J=~ J~
M~ ~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

5 4 ii ii 0.0833 0 0.0 0 0.0 0 0.0

5 1 12 6 0.0747 7 0.0407 ~ 0.0204 9 0.0090
5 2 12 9 0.0600 I0 0.0276 Ii 0.0099 II 0.0099
5 3 12 ii 0.0525 12 0.0147 12 0.0147 0 0.0
5 4 12 12 0.0735 0 0,0 0 0.0 0 0.0
5 I 13 6 0.0924 8 0.0294 9 0.0147 i0 0.0065
5 2 13 9 0.0882 I0 0.0474 Ii 0.0217 12 0.0077
5 3 13 ii 0.0987 12 0.0441 13 0.0123 0 0.0
5 4 13 13 0.0654 0 0.0 0 0.0 0 0.0
5 i 14 7 0.0681 8 0.0397 9 0.0217 ii 0.0048
5 2 14 I0 0.0710 ii 0.0379 12 0.0173 13 0.0061
5 3 14 12 0.0844 13 0.0374 14 0.0103 0 0.0
5 4 14 14 0.05~5 0 0.0 0 0.0 0 0.0
5 i 15 7 0.0830 9 0.0298 i0 0.0163 II 0.0081
5 2 15 I0 0.0975 12 0,0307 13 0.0139 14 0.0049
5 3 15 13 0.0726 14 0.0320 15 0.0088 15 0.0088
5 4 15 15 0.0526 0 0.0 0 0.0 0 0.0
6 1 3 2 0.0833 3 0.0119 3 0.0119 0 0.0
6 2 3 3 0.0476 3 0.0476 0 0.0 0 0.0
6 I 4 3 0.0333 3 0.0333 4 0.0048 4 0.0048
6 2 4 4 0.0238 4 0.0238 4 0.0238 0 0.0
6 3 4 4 0.0714 0 0.0 0 0.0 0 0.0
6 I 5 3 0.0606 4 0.0152 4 0.0152 5 0.0022
6 2 5 4 0.0671 5 0.0130 5 0.0130 0 0.0
6 3 5 5 0.0455 5 0.0455 0 0.0 0 0.0
6 1 6 3 0.0909 4 0.0303 5 0.0076 5 0.0076
6 2 6 5 0,0400 5 0.0400 6 0.0076 6 0.0076
6 3 6 6 0.0303 6 0.0303 0 0.0 0 0.0
6 4 6 6 0.0909 0 0.0 0 0.0 0 0.0
6 I 7 4 0.0490 4 0.0490 5 0.0163 6 0.0041
6 2 7 5 0.0775 6 0.0251 7 0.0047 7 0.0047
6 3 7 6 0.0R62 7 0.0210 7 0.0210 0 0.0
6 4 7 7 0.0699 0 0.0 0 0.0 0 0.0
6 I 8 4 0.0699 5 0.0280 6 0.0093 6 0.0093
6 2 8 6 0.0513 7 0,0163 7 0.0163 8 0.0030
6 3 8 7 0.0629 8 0.0150 8 0.0150 0 0.0
6 4 8 8 0.0549 0 0.0 0 0.0 0 0.0
6 i 9 4 0.0923 5 0.0420 6 0.0168 7 0.0056
6 2 9 6 0.0839 7 0.0350 8 0.0110 9 0.0020
6 3 9 8 0.0470 8 0.0470 9 0o0110 0 0.0
6 4 9 9 0,0440 9 0.0440 0 0.0 0 0.0
6 I I0 5 0.0577 6 0.0262 7 0.0105 8 0.0035
6 2 I0 7 0.0594 8 0.0245 8 0.0245 9 0.0076
6 3 I0 8 0.0919 9 0.0357 I0 0.0082 I0 0.0082
6 4 i0 i0 0,0357 I0 0.0357 0 0.0 0 0.0
6 I ii 5 0.0747 6 0.0373 7 0.0170 8 0.0068
6 2 Ii 7 0.0882 8 0.0430 9 0.0175 I0 0.0054
6 3 Ii 9 0.0721 I0 0.0276 ii 0.0063 ii 0.0063
6 4 11 I0 0.0987 II 0.0294 0 0.0 0 0.0

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions auth
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 133

TABLE 15--Precedence tables (continued).

M~ r~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

6 i 12 5 0.0924 6 0.0498 7 0.0249 9 0.0045

6 2 12 8 0.0656 9 0.0317 I0 0.0128 ii 0 . 0 0 3 9
6 3 12 I0 0.0573 ii 0.0217 ii 0.0217 12 0 . 0 0 4 9
6 4 12 ii 0.0~33 12 0.0245 12 0.0245 0 0.0
6 5 12 12 0.09~0 0 0.0 0 0.0 0 0.0
6 1 13 6 0.0632 7 0.0341 8 0.0170 9 0.0077
6 2 13 8 0.0913 9 0.0495 i0 0.0237 II 0 . 0 0 9 5
6 3 13 I0 0.0967 Ii 0.0460 12 0.0173 13 0 . 0 0 3 9
6 4 13 12 0.0709 13 0.0206 13 0.0206 0 0.0
6 5 13 13 0.0877 0 0.0 0 0.0 0 0.0
6 i 14 6 0.0775 7 0.0443 8 0.0238 i0 0 . 0 0 5 4
6 2 14 9 0.0704 i0 0.0579 Ii 0.0181 12 0 . 0 0 7 2
6 3 14 II 0.0777 12 0.0374 13 0.0139 14 0 . 0 0 3 1
6 4 14 13 0.0609 14 0.0175 14 0.0175 0 0.0
6 5 14 14 0o07R9 0 0.0 0 0.0 0 0.0
6 1 15 6 0.0922 8 0.0316 9 0.0170 I00.OOS5
6 2 15 9 0.0937 Ii 0.0294 12 0.0139 13 0 . 0 0 5 5
6 3 15 12 0.0642 13 0.0307 14 0.0114 15 0 . 0 0 2 5
6 4 15 14 0.0526 15 0.0150 15 0.0150 0 0.0
6 5 15 15 0.0714 0 0.0 0 0.0 0 0.0
7 i 4 3 0.0242 3 0.0242 3 0.0242 4 0.0030
7 2 4 3 0.0879 4 0.0152 4 0.0152 0 0.0
7 3 4 4 0.0455 4 0.0455 0 0.0 0 0.0
7 I 5 3 0.0455 3 0.0455 4 0.0101 5 0.0013
7 2 5 4 0.0455 4 0.~D455 5 0.0076 5 0.0076
7 3 5 5 0.0265 5 0.0265 0 0.0 0 0.0
7 4 5 5 0.0707 0 0.0 0 0.0 0 0.0
7 1 6 3 0.0699 4 0.0210 4 0.0210 5 0.0047
7 2 6 4 0.0~62 5 0.0251 6 0.0041 6 0.0041
7 3 6 5 0.0775 6 0.0163 6 0.0163 0 0.0
7 4 6 6 0.0490 6 0.0490 0 0.0 0 0.0
7 I 7 3 0.0962 4 0.0350 5 0.0105 6 0.0023
7 2 7 5 0.0513 6 0.0146 6 0.0146 7 0.0023
7 3 7 6 0.0513 7 0.0105 7 0.0105 0 0.0
7 4 7 7 0.0350 7 0.0350 0 0.0 0 0.0
7 5 7 7 0.0962 0 0.0 0 0.0 0 0,0
7 1 8 4 0.0513 5 0.0186 5 0.0186 6 0.0056
7 2 8 5 0.0839 6 0.0317 7 0.0089 7 0.0089
7 3 S 7 0.0350 7 0.0350 8 0.0070 8 0.0070
7 4 8 8 0.0256 8 0.0256 0 0.0 0 0,0
7 5 8 8 0.0769 0 0.0 0 0.0 0 0,0
7 I 9 4 0.0692 5 0.0288 6 0.0105 7 0.0031
7 2 9 6 0.0545 7 0.0203 7 0.0203 8 0,0056
7 3 9 7 0.0717 8 0.0245 8 0.0245 9 0,0048
7 4 9 8 0.0769 9 0.0192 9 0.0192 0 0.0
7 5 9 9 0.0625 0 0.0 0 0.0 0 0,0
7 i i0 4 0.0882 5 0.0407 6 0.0170 7 0.0062
7 2 i0 6 0.0818 7 0.0364 a 0.0134 9 0.0037
7 3 i0 8 0.0522 9 0.0175 9 0.0175 I0 0 . 0 0 3 4

TABLE 15---Precedence tab~s (continued).

J~ J~ j~ J~
My r~ N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

7 4 I0 9 0.0600 i0 0.0147 I0 0.0147 0 0.0

7 5 I0 I0 0.0515 0 0.0 0 0.0 0 0.0
7 I II 5 0.0539 6 0.0249 6 0.0249 8 0.0038
7 2 II 7 0.0566 8 0.0249 8 0.0249 9 0.0090
7 3 ii 8 0.0882 9 0.0387 I0 0.0128 ii 0.0025
7 4 II i0 0.0474 I0 0.0474 II 0.0114 0 0.0
7 5 ii II 0.0429 II 0.0429 0 0.0 0 0.0
7 1 12 5 0.0681 6 0,0341 7 0.0157 8 0.0065
7 2 12 7 0,0799 8 0.0399 9 0,0174 i0 0.0063
7 3 12 9 0.0674 I0 0,0292 II 0.0095 II 0.0095
7 4 12 I0 0,0947 II 0,0379 12 0.0090 12 0.0090
7 5 12 12 0.0361 12 0.0361 0 0.0 0 0.0
7 I 13 5 0.0830 6 0.0443 7 0.0221 9 0.0043
7 2 13 8 0.0579 9 0,0286 I0 0.0124 II 0.0044
7 3 13 I0 0.0521 II 0.0223 Ii 0.0223 12 0.0072
7 4 13 II 0.0777 12 0.0307 13 0.0072 13 0.0072
7 5 13 13 0.0507 13 0.0307 0 0.0 0 0.0
7 I 14 5 0.0984 7 0,0295 8 0.0148 9 0.0068
7 2 14 8 0.0783 9 0.0426 I0 0.0209 II 0.0090
7 3 14 i0 0.0805 II 0.0408 12 0.0173 13 0.0055
7 4 14 12 0.0642 13 0,0251 14 0.0058 14 0.0058
7 5 14 13 0.0877 14 0.0265 0 0,0 0 0.0
7 6 14 14 0.I000 0 0.0 0 0.0 0 0.0
7 I 15 6 0.0671 7 0.0377 8 0.0201 i0 0.0046
7 2 15 9 0.0588 I0 0.0317 II 0.0155 12 0.0066
7 3 15 Ii 0.0642 12 0.0322 13 0.0136 14 0.0043
7 4 15 13 0.0536 14 0.0207 14 0.0207 15 0.0048
7 5 15 14 0.0766 15 0,0227 15 0.0227 0 0.0
7 6 15 15 0.0909 0 0.0 0 0.0 0 0.0
8 I 5 3 0.0350 3 0.0350 4 0.0070 4 0.0070
8 2 5 4 0,0319 4 0.0319 5 0.0047 5 0,0047
8 3 5 4 0.0862 5 0.0163 5 0.0163 0 0.0
8 4 5 5 0.0435 5 0.0435 0 0.0 0 0.0
8 5 5 5 0.0979 0 0.0 0 0.0 0 0o0
8 I 6 3 0,0549 4 0.0150 4 0.0150 5 0.0030
8 2 6 4 0.0629 5 0.0163 5 0.0163 6 0.0023
8 3 6 5 0.0513 6 0.0093 6 0.0093 6 0.0093
8 4 6 6 0.0280 6 0.0280 0 0.0 0 0.0
8 5 6 6 0,0699 0 0.0 0 0,0 0 0.0
8 I 7 3 0.0769 4 0.0256 5 0.0070 5 0.0070
8 2 7 5 0.0350 5 0.0350 6 0.0089 6 0.0089
8 3 7 6 0.0317 6 0.0317 7 0.0056 7 0.0056
8 4 7 6 0.0839 7 0.0186 7 0.0186 0 0.0
8 5 7 7 0.0513 0 0.0 0 0.0 0 0.0
8 1 8 3 0.I000 4 0.0385 5 0.0128 6 0.0035
8 2 8 5 0,0594 6 0.0203 6 0.0203 7 0.0051
8 3 8 6 0.0660 7 0.0203 7 0.0203 8 0.0035
8 4 8 7 0.059~ 8 0.0128 8 0.0128 0 0,0
8 5 8 8 0.0385 8 0.0385 0 0.0 0 0.0

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions autho
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 135

TABLE 15--Precedence tables (continued).

J~. Jx. j=. J~

M. ~ N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0.01 Prob

8 6 8 8 0.1000 0 0.0 0 0.0 0 0.0

8 I 9 4 0.0529 5 0.0204 5 0.0204 6 0.0068
8 2 9 5 0.0882 6 0.0364 7 0.0122 8 0.0030
8 3 9 7 0.0445 7 0.0445 8 0.0134 9 0.0023
8 4 9 8 0.0430 8 0.0430 9 0.0090 9 0.0090
8 5 9 9 0.0294 9 0.0294 0 0.0 0 0.0
8 6 9 9 0.0824 0 0.0 0 0.0 0 0.0
8 i i0 4 0.0686 5 0.0294 6 0.0113 7 0.0038
8 2 I0 6 0,0566 7 0.0230 7 0.0230 8 0.0076
8 3 I0 7 0.0767 8 0.0306 9 0,0090 9 0.0090
8 4 I0 8 0.0882 9 0.0317 I0 0.0065 I0 0.0065
8 5 I0 9 0.0882 I0 0.0229 I0 0.0229 0 0.0
8 6 I0 I0 0.0686 0 0,0 0 0o0 0 0.0
8 1 ii 4 0.0851 5 0.0397 6 0.0170 7 0.0065
8 2 II 6 0.0799 7 0.0371 R 0.0149 9 0.0049
8 3 ii 8 0.0549 9 0.0216 9 0.0216 I0 0.0063
8 4 ii 9 0.0674 I0 0.0237 I0 0.0237 II 0.0048
8 5 II I0 0.0710 II 0.0181 II 0.0181 0 0.0
8 6 II II 0.0578 0 0.0 0 0.0 0 0.0
8 I 12 5 0.0511 6 0.0238 6 0.0238 8 0.0039
8 2 12 7 0.0542 8 0,0249 8 0.0249 9 0.0099
8 3 12 8 0.0849 9 0.0399 I0 0.0154 ii 0.0044
8 4 12 I0 0.0521 Ii 0.0181 ii 0.0181 12 0.0036
8 5 12 II 0.0578 12 0.0144 12 0.0144 0 0.0
8 6 12 12 0.0491 12 0.0491 0 0.0 0 0.0
8 1 13 5 0.0632 6 0.0316 7 0.0148 8 0.0063
8 2 13 7 0.0738 8 0.0375 9 0.0170 i0 0.0067
8 3 13 9 0,0635 i0 0.0294 II 0.0112 12 0.0032
8 4 13 I0 0.0900 II 0.0408 12 0.0139 13 0.0028
8 5 13 12 0.0475 12 0.0475 13 0.0117 0 0.0
8 6 13 13 0.0421 13 0.0421 0 0.0 0 0.0
8 I 14 5 0.0760 6 0~ 7 0o0201 8 0.0094
8 2 14 7 0.0953 9 0.0263 I0 0.0119 ii 0.0046
8 3 14 9 0.0913 I0 0.0480 ii 0.0220 12 0.0083
8 4 14 II 0.0721 12 0.0322 13 0.0109 14 0.0021
8 5 14 12 0.0963 13 0.0393 14 0,0096 14 0.0096
8 6 14 14 0,0364 14 0.0364 0 0.0 0 0.0
8 1 15 5 0.0892 6 0.0496 8 0.0131 9 0.0061
8 2 15 8 0.0691 9 0.0376 I0 0.0188 II 0.0084
8 3 15 I0 0.0706 II 0.0367 12 0.0166 13 0.0062
8 4 15 12 0.0582 13 0.0257 14 0.0086 14 0.0086
8 5 15 13 0.0814 14 0.0329 15 0,0079 15 0.0079
8 6 15 15 0.0316 15 0.0316 0 0.0 0 0.0
9 I 6 3 0.0440 3 0.0440 4 0.0110 5 0.0020
9 2 6 4 0.0470 4 0.0470 5 0,0110 6 0,0014
9 3 6 5 0.0350 5 0.0350 6 0.0056 6 0.0056
9 4 6 5 0.0839 6 0.0168 6 0,0168 0 0,0
9 5 6 6 0.0420 6 0.0420 0 0.0 0 0.0
9 6 6 6 0.0923 0 0.0 0 0.0 0 0.0

TABLE 15--Precedence tables (continued).

M~ r~ N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

9 1 7 3 0.0625 4 0.0192 4 0.0192 5 0.0048

9 2 7 4 0.0769 5 0.0245 5 0.0245 6 0.0056
9 3 7 5 0.0717 6 0.0203 6 0.0203 7 0.0031
9 4 7 6 0.0545 7 0.0105 7 0.0105 0 0.0
9 5 7 7 0.0288 7 0.0288 0 0.0 0 0.0
9 6 7 7 0.0692 0 0.0 0 0.0 0 0.0
9 1 8 3 0.0824 4 0.0294 5 0.0090 5 0.0090
9 2 8 5 0.0430 5 0.0430 6 0.0134 7 0.0030
9 3 8 6 0,0445 6 0,0445 7 0.0122 8 0.0019
9 4 7 0.0364 7 0.0364 8 0.0068 8 0.0068
9 5 8 7 0.08S2 8 0.0204 8 0.0204 0 0.0
9 6 8 8 0.0529 0 0.0 0 0.0 0 0.0
9 I 9 4 0.0412 4 0.0412 5 0.0147 6 0.0045
9 2 9 5 0.0656 6 0.0249 6 0.0249 7 0.0076
9 3 9 6 0.0767 7 0.0283 8 0.0076 8 0.0076
9 4 9 7 0.0767 8 0.0249 8 0.0249 9 0.0045
9 5 9 8 0.0656 9 0.0147 9 0.0147 0 0.0
9 6 9 9 0.0412 9 0.0412 0 0.0 0 0.0
9 1 i0 4 0.0542 5 0.0217 5 0.0217 6 0,0077
9 2 i0 5 0.0913 6 0.0399 7 0.0149 8 0.0045
9 3 I0 7 0.0513 8 0.0185 8 0.0185 9 0.0049
9 4 I0 8 0.0549 9 0.0174 9 0.0174 I0 0.0031
9 5 I0 9 0.0495 9 0.0495 I0 0.0108 0 0.0
9 6 i0 I0 0.0325 I0 0.0325 0 0.0 0 0.0
9 7 i0 i0 0.0867 0 0.0 0 0.0 0 0.0
9 1 ii 4 0.0681 5 0.0298 6 0.0119 7 0.0043
9 2 Ii 6 0.0579 7 0.0249 7 0.0249 8 0.0092
9 3 ii 7 0.0799 8 0.0349 9 0.0124 i0 0.0032
9 4 11 8 0.0949 9 0.0399 i0 0.0124 ii 0.0022
9 5 ii i0 0.0379 i0 0.0579 II 0.0081 ii 0.0081
9 6 ii I0 0.0975 11 0.0260 0 0.0 0 0.0
9 7 ii Ii 0.0737 0 0.0 0 0.0 0 0.0
9 1 12 4 0.0827 5 0.0389 6 0.0170 7 0.0068
9 2 12 6 0.0783 7 0.0375 8 0.0159 9 0.0058
9 3 12 8 0.0563 9 0.0242 9 0.0242 10 0.0085
9 4 12 9 0.0713 I0 0.0294 II 0,0090 Ii 0.0090
9 5 12 I0 0.0805 II 0.0294 12 0.0062 12 0.0062
9 6 12 ii 0o0805 12 0.0211 12 0.0211 0 0.0
9 7 12 12 0.0632 0 0.0 0 0.0 0 0.0
9 1 13 4 0.0977 5 0.0489 6 0~ 8 0.0040
9 2 13 7 0.0523 8 0.0247 8 0.0247 i0 0,0038
9 3 13 8 0.0820 9 0.0402 i0 0.0170 ii 0.0059
9 4 13 I0 0.0542 II 0.0220 ii 0.0220 12 0.0066
9 5 13 Ii 0.0642 12 0.0231 12 0.0231 13 0.0048
9 6 13 12 0.0670 13 0.0172 13 0.0172 0 0.0
9 7 13 13 0.0545 0 0.0 0 0.0 0 0.0
9 1 14 5 0.0595 6 0.0297 7 0.0140 8 0.0061
9 2 14 7 0,0691 8 0.0355 9 0.0166 I0 0,0069
9 3 14 9 0.0602 I0 0.0291 II 0.0122 12 0.0042

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 137

T A B L E 15--Precedence tables (contmued).

M~ rv N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

9 4 14 i0 0.0857 II 0.0416 12 0.0166 13 0.0049

9 5 14 12 0.0517 13 0.0183 13 0.0183 14 0.0037
9 6 14 13 0.0562 14 0.0142 14 0.0142 0 0.0
9 7 14 14 0.0474 14 0.0474 0 0.0 0 0.0
9 i 15 5 0.0707 6 0.0372 7 0.0186 8 0.0087
9 2 15 7 0.0875 8 0.0481 9 0.0245 ii 0,0047
9 3 15 9 0.0836 I0 0.0447 II 0.0214 12 0.0088
9 4 15 II 0.0673 12 0.0322 13 0.0127 14 0.0037
9 5 15 12 0,0907 13 0.0420 14 0.0147 15 0.0030
9 6 15 14 0.0474 14 0.0474 15 0.0119 0 0.0
9 7 15 15 0.0415 15 0.0415 0 0.0 0 0.0
I0 1 7 3 0.0515 4 0.0147 4 0.0147 5 0.0034
i0 2 7 4 0.0600 5 0.0175 5 0.0175 6 0.0037
I0 3 7 5 0.0522 6 0.0134 6 0.0134 7 0.0019
i0 4 7 6 0.0364 6 0.0364 7 0.0062 7 0.0062
I0 5 7 6 0.0818 7 0.0170 7 0.0170 0 0.0
i0 6 7 7 0.0407 7 0.0407 0 0.0 0 0.0
I0 7 7 7 0.0882 0 0.0 0 0.0 0 0.0
i0 I 8 3 0.0686 4 0.0229 4 0.0229 5 0.0065
I0 2 8 4 0.0882 5 0.0317 6 0.0090 6 0.0090
i0 3 8 5 0.0882 6 0.0306 7 0.0076 7 0.0076
I0 4 8 6 0.0767 7 0.0250 7 0.0230 8 0.0038
i0 5 8 7 0.0566 8 0.0113 8 0.0113 0 0.0
i0 6 8 8 0.0294 8 0.0294 0 0.0 0 0.0
i0 7 8 8 0,0686 0 0.0 0 0.0 0 0.0
I0 1 9 3 0.0867 4 0.0325 5 0.0108 6 0.0031
i0 2 9 5 0~ 5 0.0495 6 0.0174 7 0.0049
I0 3 9 6 ().0549 7 0.0185 7 0.0185 8 0.0045
i0 4 9 7 0.0513 8 0.0149 8 0.0149 9 0.0024
10 5 9 8 0.0399 8 0.0399 9 0.0077 9 0.0077
i0 6 9 8 0.0913 9 0.0217 9 0.0217 0 0.0
I0 7 9 9 0.0542 0 0.0 0 0.0 0 0.0
I0 I 10 4 0.0433 4 0.0433 5 0.0163 6 0.0054
I0 2 i0 5 0.0704 6 0.0286 7 0.0099 7 0.0099
i0 3 i0 6 0.0849 7 0.0349 8 0.0115 9 0.0027
I0 4 I0 7 0.0894 8 0,0349 9 0.0099 9 0.0099
10 5 I0 8 0.0849 9 0.0286 i0 0.0054 I0 0.0054
i0 6 i0 9 0.0704 10 0,0163 I0 0,0163 0 0,0
i0 7 10 I0 0.0433 i0 0.0433 0 0.0 0 0.0
I0 1 II 4 0.0551 5 0.0227 5 0.0227 6 0.0085
I0 2 Ii 5 0.0937 6 0.0426 7 0.0170 8 0.0058
I0 3 ii 7 0.0563 8 0.0226 8 0.0226 9 0.0073
i0 4 ii 8 0.0635 9 0.0242 9 0.0242 I0 0.0067
10 5 ii 9 0.0635 I0 0.0209 10 0.0209 Ii 0.0039
I0 6 Ii I0 0.0550 II 0.0124 II 0.0124 0 0.0
i0 7 11 ii 0.0351 ii 0.0351 0 0.0 0 0.0
I0 8 II ii 0.0902 0 0.0 0 0.0 0 0.0
i0 I 12 4 0.0677 5 0.0301 6 0.0124 7 0.0046
10 2 12 6 0.0588 7 0.0263 8 0.0104 9 0.0035

T A B L E 15 Precedence tables (continued).

I.= J~= J,= A=

Mr r, N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

I0 3 12 7 0.0820 8 0.0380 9 0.0150 10 0.0048

I0 4 12 8 0.0992 9 0.0456 10 0.0170 11 0.0046
I0 5 12 10 0.0480 10 0.0480 11 0.0155 12 0.002~
I0 6 12 II 0.0433 II 0.0433 12 0.0096 12 0.0096
I0 7 12 12 0.0287 12 0.0287 0 0.0 0 0.0
I0 8 12 12 0.0779 0 0.0 0 0.0 0 0.0
I0 i 13 4 0.0807 5 0.0382 6 0.0170 7 0.0070
I0 2 13 6 0.0770 7 0.0376 8 0.0166 9 0.0065
I0 3 13 8 0.0571 9 0.0260 I0 0.0101 II 0.0032
I0 4 13 9 0.0736 I0 0.0332 II 0.0122 12 0.0033
i0 5 13 i0 0.0857 ii 0.0367 12 0.0116 13 0.0021
i0 6 13 II 0.0918 12 0.0345 13 0.0075 13 0.0075
I0 7 13 12 0.0886 13 0.0237 13 0.0237 0 0.0
10 8 13 13 0.0678 0 0.0 0 0.0 0 0.0
I0 1 14 4 0.0942 5 0.0471 6 0,0223 7 0.0099
I0 2 14 6 0.0967 8 0.0245 8 0.0245 I0 0.0042
I0 B 14 8 0.0796 9 0.0402 I0 O.OIAO ii 0.0069
10 4 14 I0 0.0551 II 0.0244 II 0.0244 12 0.0088
I0 5 14 ii 0.0673 12 0.0283 13 0.0088 13 0.0088
i0 6 14 12 0.0751 13 0.0278 14 0.0059 14 0.0059
I0 7 14 13 0.0751 14 0.0198 14 0.0198 0 0.0
i0 8 14 14 0.0593 0 0.0 0 0.0 0 0.0
10 I 15 5 0.0565 6 0.0283 7 0.0134 8 0.0059
I0 2 15 7 0.0654 8 0.0339 9 0.0162 i0 0.0070
i0 3 15 9 0.0576 i0 0.0287 II 0.0127 12 0.0048
I0 4 15 I0 0.0820 II 0.0416 12 0.0182 13 0.0065
I0 5 15 12 0.0533 13 0.0221 13 0.0221 14 0.0068
i0 6 15 13 0.0618 14 0.0225 14 0.0225 15 0.0047
i0 7 15 14 0.0640 15 0.0166 15 0.0166 0 0.0
I0 8 15 15 0.0522 0 0.0 0 0.0 0 0.0
II I 8 3 0.0578 4 0.0181 4 0.0181 5 0.0048
ii 2 8 4 0.0710 5 0.0237 5 0.0237 6 0.0063
II 3 8 5 0.0674 6 0.0216 6 0.0216 7 0.0049
II 4 8 6 0.0549 7 0.0149 7 0.0149 8 0.0022
II 5 8 7 0.0371 7 0.0371 8 0.0065 8 0.0065
II 6 8 7 0.0799 8 0.0170 8 0.0170 0 0.0
ii 7 8 8 0.0397 8 0.0397 0 0.0 0 0.0
ii 8 8 8 0.0851 0 0.0 0 0.0 0 0.0
II I 9 3 0.0737 4 0.0260 5 0.0081 5 0.0081
II 2 9 4 0.0975 5 0.0379 6 0o0124 7 0.0032
II 3 9 6 0.0399 6 0.0399 7 0.0124 8 0.0027
II 4 9 6 0.0949 7 0.0349 8 0.0092 R 0.0092
II 5 9 7 0,0799 8 0.0249 8 0.0249 9 0.0043
II 6 9 8 0.0579 9 0.0119 9 0.0119 0 0.0
ii 7 9 9 0.0298 9 0.0298 0 0.0 0 0.0
ii 8 9 9 0.0681 0 0.0 0 0.0 0 0.0
II I I0 3 0.0902 4 0.0351 5 0.0124 6 0.0039
II 2 I0 5 0.0550 6 0.0209 6 0.0209 7 0.0067
II 3 io 6 0.0635 7 0.0242 7 0.0242 8 0.0073

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions au
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 139

TABLE 15--Precedence tables (continued).

J~, j.~ J.,~ J.,~

My r~ N, 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

ii 4 i0 7 0.0635 8 0.0226 8 0.0226 9 0.0058

Ii 5 10 8 0.0563 9 0.0170 9 0.0170 IO 0.0028
ii 6 I0 9 0.0426 9 0.0426 I0 0.0085 I0 0.0085
Ii 7 10 9 0.0937 10 0.0227 10 0.0227 0 0.0
ii 8 i0 i0 0.0551 0 0.0 0 0.0 0 0.0
ii 1 ii 4 0.0451 4 0.0451 5 0.0175 6 0.0062
ii 2 11 5 0.0743 6 0.0317 7 0.0119 8 0.0038
ii 3 ii 6 0.0913 7 0.0402 8 0.0150 9 0.0045
ii 4 Ii 7 0.0992 8 0.0431 9 0.0150 I0 0.0038
ii 5 ii 8 0.0992 9 0.0402 10 0.0119 ii 0.0019
ii 6 ii 9 0.0913 i0 0.0317 ii 0.0062 Ii 0.0062
ii 7 II I0 0.0743 Ii 0.0175 ii 0.0175 0 0.0
ii 8 ii Ii 0.0451 ii 0.0451 0 0.0 0 0.0
ii 1 12 4 0.0559 5 0.0235 5 0.0235 6 0.0092
ii 2 12 5 0.0955 6 0.0447 7 0.0188 8 0.0069
Ii 3 12 7 0.0602 8 0.0260 9 0.0095 9 0.0095
Ii 4 12 8 0.0699 9 0.0296 i0 O.OlOl ii 0.0025
ii 5 12 9 0.0736 10 0.0291 11 0.0084 11 0.0084
ii 6 12 i0 0.0706 ii 0.0240 Ii 0.0240 12 0.0046
ii 7 12 ii 0.0595 12 0.0137 12 0.0137 0 0.0
ii 8 12 12 0.0373 12 0.0373 0 0.0 0 0.0
ii 9 12 12 0.0932 0 0.0 0 0.0 0 0.0
Ii 1 13 4 0.0673 5 0.0303 6 0.0127 7 0.0050
II 2 13 6 0.0595 7 0.0274 8 0.0114 9 0.0042
ii 3 13 7 0.0836 0.0402 9 0.0170 i0 0.0061
ii 4 13 9 0.0498 9 0.0498 i0 0.0207 ii 0.0069
ii 5 13 i0 0.0551 Ii 0.0214 ii 0.0214 12 0.0061
II 6 13 II 0.0551 12 0.0183 12 0.0183 13 0.0034
ii 7 13 12 0.0481 12 0.0481 13 0.0109 0 0.0
II 8 13 13 0.0311 13 0.0311 0 0.0 0 0.0
II 9 13 13 0.0815 0 0.0 0 0.0 0 0.0
II i 14 4 0.0791 5 0.0377 6 0.0170 7 0.0071
II 2 14 6 0.0759 7 0.0377 8 0.0171 9 0.0070
II 3 14 8 0.0576 9 0.0273 i0 0.0114 ii 0.0040
ii 4 14 9 0.0749 I0 0.0358 ii 0.0146 12 0.0048
II 5 14 i0 0.0887 ii 0.0416 12 0.0159 13 0.0044
II 6 14 ii 0.0982 12 0.0433 13 0,0141 14 0~
ii 7 14 13 0.0391 13 0.0391 14 0.0087 14 0.0087
ii 8 14 13 0.0957 14 0.0261 0 0.0 0 0.0
II 9 14 14 0.0717 0 0.0 0 0.0 0 0.0
ii 1 15 4 0.0913 5 0.0457 6 0.0217 7 0.0098
Ii 2 15 6 0.0935 7 0.0494 8 0.0243 I0 0.0045
II 3 15 8 0.0776 9 0.0401 10 0.0187 ii 0.0077
Ii 4 15 I0 0.0554 II 0.0260 12 0.0105 13 0.0034
Ii 5 15 ii 0.0688 12 0.0317 13 0.0119 14 0.0033
ii 6 15 12 0.0792 13 0.0343 14 0.0110 15 0.0020
ii 7 15 13 0.0848 14 0.0321 15 0.0070 15 0.0070
Ii 8 15 14 0.0823 15 0.0221 15 0.0221 0 0.0
II 9 15 15 0.0635 0 0.0 0 0.0 0 0.0

TABLE 15--Precedence tables (continued).

Mu ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

12 I 9 3 0.0632 4 0.0211 4 0.0211 5 0.0062

12 2 9 4 0.0805 5 0.0294 6 0.0090 6 0.0090
12 3 9 5 0.0805 6 0.0294 7 0.0085 7 0.00R5
12 4 9 6 0.0713 7 0.0262 7 0.0242 8 0.0058
12 5 9 7 0.0563 8 0.0159 8 0.0159 9 0.0024
12 6 9 8 0.0375 8 0.0375 9 0.0068 9 0.0068
12 7 9 8 0.0783 9 0.0170 9 0.0170 0 0.0
12 8 9 9 0.0389 9 0.0389 0 0.0 0 0.0
12 9 9 9 0.0827 0 0.0 0 0.0 0 0.0
12 I I0 3 0.0779 4 0.0287 5 0.0096 5 0.0096
12 2 i0 5 0.0433 5 0.0433 6 0.0155 7 0.0046
12 3 I0 6 0.0480 6 0.0480 7 0.0170 8 0.0048
12 4 I0 7 0.0456 7 0.0456 8 0.0150 9 0.0035
12 5 10 7 0,0992 8 0.0380 9 0.0104 I0 0,0015
12 6 I0 8 0.0820 9 0.0263 I0 0.0046 i0 0.0046
12 7 i0 9 0.0588 10 0.0124 10 0.0124 0 0.0
12 8 I0 I00.OBOI I0 0.0301 0 0.0 0 0.0
12 9 I0 I0 0 . 0 6 7 7 0 0.0 0 0.0 0 0.0
12 I Ii 3 0.0932 4 0.0373 5 0.0137 6 0.0046
12 2 Ii 5 0.0595 6 0.0240 6 0.0240 7 0.0084
12 3 ii 6 0.0706 7 0.0291 8 0.0101 9 0.0028
12 4 ii 7 0.0736 8 0.0296 9 0.0095 9 0.0095
12 5 ii 8 0,0699 9 0.0260 i0 0.0069 i0 0.0069
12 6 Ii 9 0.0602 I0 0.0188 10 0.0188 ii 0.0032
12 7 Ii i0 0 . 0 4 4 7 I0 0,0447 ii 0.0092 II 0.0092
12 8 11 i0 0 . 0 9 5 5 II 0.0235 II 0.0235 0 0.0
12 9 II ii 0 . 0 5 5 9 0 0.0 0 0.0 0 0.0
12 I 12 4 0.0466 4 0.0466 5 0.0186 6 0.0069
12 2 12 5 0.0775 6 0.0343 7 0.0136 8 0.0047
12 3 12 6 0.0965 7 0.0447 8 0.01~0 9 0.0061
]2 4 12 8 0.0498 8 0.0498 9 0.0196 i0 0.0061
12 5 12 9 0.0498 9 0.0498 i0 0.0180 ii 0.0047
12 6 12 i0 0 , 0 4 4 7 I0 0.0447 ii 0.0156 12 0.0025
12 7 12 I0 0 . 0 9 6 5 II 0.0543 12 0.0069 12 0.0069
12 8 12 Ii 0 . 0 7 7 5 12 0.0186 12 0.0186 0 0.0
12 9 12 12 0 . 0 4 6 6 12 0.0466 0 0.0 0 0.0
12 I 13 4 0.0565 5 0.0242 5 0.0242 6 0.0097
12 2 13 5 0.0969 6 0.0464 7 0.0202 8 0.0079
12 3 13 7 0.0633 8 0.0287 9 0.0114 i0 0.0058
]2 4 13 8 0.0749 9 0.0341 i0 0.0131 ii 0.0040
12 5 13 9 0.0812 I0 0.0358 II 0.0127 12 0.0033
12 6 13 i0 0 . 0 8 2 0 ii 0.0555 12 0.0100 12 0.0100
12 7 13 ii 0 . 0 7 6 6 12 0.0266 13 0.0052 13 0.0052
12 8 iB 12 0 . 0 6 3 4 13 0.0149 13 0.0149 0 0.0
12 9 13 13 0 . 0 3 9 1 13 0.0391 0 0.0 0 0.0
12 I0 13 13 0 . 0 9 5 7 0 0.0 0 0.0 0 0.0
12 1 14 4 0.0670 5 0.0304 6 0.0130 7 0.0052
12 2 14 6 0.0600 7 0.0283 8 0.0122 9 0.0047
12 3 14 7 0,0847 8 0.0420 9 0.0187 i0 0.0075

Copyright by ASTM Int'l (all rights reserved); Sun Jan 3 20:27:44 EST 2016
Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS ] 41

T A B L E 15---Precedence tables (continued).

M~ ry N~ 0.10 Prob 0 . 0 5 Prob 0.025 Prob 0.01 Prob

12 4 14 9 0.0529 I0 0.0236 i0 0.0236 Ii 0.0089

12 5 14 I0 0.0602 II 0.0260 12 0.0091 IZ 0.0091
12 6 14 Ii 0.0633 12 0.0254 13 0.0074 13 0.0074
12 7 14 12 0.0612 13 0.0209 13 0.0209 14 0.0040
12 8 14 13 0.0522 14 0.0120 14 0.0120 0 0.0
12 9 14 14 0.0331 14 0.0331 0 0.0 0 0.0
12 i0 14 14 0.0846 0 0.0 0 0.0 0 0.0
12 i 15 4 0.0778 5 0.0372 6 0.0169 7 0.0072
12 Z 15 6 0.0749 7 0.0377 8 0.0175 9 0.0075
12 3 15 8 0.057R 9 0.0282 10 0.0124 ii 0.()048
12 4 15 9 0.0757 i0 0.0377 ii 0.0165 12 0.0062
12 5 15 I0 0.0906 II 0.0450 12 0.0191 13 0.0066
12 6 15 12 0.0493 12 0.0493 13 0.0194 14 0.0056
12 7 15 13 0.0493 13 0.0493 14 0.0165 15 0.0031
12 8 15 14 0.0433 14 0.0433 15 0.0098 15 0.0098
12 9 15 15 0.0282 15 0.0282 0 0.0 0 0.0
12 10 15 15 0.0752 0 0.0 0 0.0 0 0.0
13 1 i0 3 0.0678 4 0.0257 4 0.0237 5 0.0075
13 2 I0 4 0.0886 5 0.0345 6 0.0116 7 0.0033
13 3 i0 5 0.0918 6 0.0367 7 0.0122 8 0.0032
13 4 i0 6 0.0857 7 0.0332 S 0.0101 9 0.0022
13 5 i0 7 0.0736 8 0.0260 9 0.0065 9 0.0065
13 6 I0 8 0.0571 9 0.0166 9 0.0166 I0 0.0026
13 7 I0 9 0.0376 9 0.0376 I0 0.0070 I0 0.0070
13 H i0 9 0.0770 I0 0.0170 I0 0.0170 0 0.0
13 9 I0 i0 0.0382 10 0.0382 0 0.0 0 0.0
13 i0 I0 i0 0.0807 0 0.0 0 0.0 0 0.0
13 1 ii 3 0.0~15 4 0.0311 5 0.0109 6 0.0034
13 2 II 5 0.0481 5 0.0481 6 0.0183 7 0.0061
13 3 Ii 6 0.0551 7 0.0214 7 0.0214 8 0.0069
13 4 ii 7 0.0551 8 0.0207 8 0.0207 9 0.0061
13 5 ii 8 0.0498 8 0.0498 9 0.0170 I0 0.0042
13 6 11 9 0.0402 9 0.0402 I0 0.0114 ii 0.0017
13 7 II 9 0.0836 i0 0.0274 ii 0.0050 II 0.0050
13 8 Ii i0 0.0595 ii 0.0127 ii 0.0127 0 0.0
13 9 ii 11 0.0305 ii 0.0303 0 0.0 0 0.0
13 i0 11 II 0.0673 0 0.0 0 0.0 0 0.0
13 I 12 3 0.0957 4 0.0591 5 0.0149 6 0.0052
13 2 12 5 0.0634 6 0.0266 7 0.0100 7 0.0100
13 3 12 6 0.0766 7 0.0335 8 0.0127 9 0.0040
13 4 12 7 0.0820 8 0.0358 9 0.0131 i0 0.0038
13 5 12 8 0.0812 9 0.0341 10 0.0114 11 0.0027
13 6 12 9 0.0749 10 0.0287 ii 0.0079 ii 0.0079
13 7 12 I0 0.0633 ii 0.0202 Ii 0.0202 12 0.0036
13 8 12 Ii 0.0464 ii 0.0464 12 0.0097 12 0.0097
13 9 12 Ii 0.0969 12 0.0242 12 0.0242 0 0.0
13 I0 12 12 0.0565 0 0.0 0 0.0 0 0.0
iB I 13 4 0.0478 4 0.0478 5 0.0196 6 0.0075
13 2 13 5 0.0801 6 0.0365 7 0.0151 8 0.0056
] 4.2 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

TABLE 15--Precedence tables (continued).

M~ rv N~ ~10 Prob 0 . 0 5 Prob 0.025 Prob 0 . 0 1 Prob

13 3 13 7 0.0484 7 0.0484 0.0207 9 0.0077

13 4 13 8 0.0554 9 0.0236 9 0.0236 I0 0 . 0 0 8 5
13 5 13 9 0.0576 I0 0.0236 i0 0 . 0 2 3 6 ii 0 . 0 0 7 7
13 6 13 I0 0.0654 11 0.0207 II 0 . 0 2 0 7 12 0 . 0 0 5 6
13 7 13 II 0.0484 11 0.0484 12 0 . 0 1 5 1 13 0 . 0 0 2 6
13 8 13 12 0.0365 12 0.0365 13 0 . 0 0 7 5 13 0 . 0 0 7 5
13 9 13 12 0.0801 13 0.0196 13 0 . 0 1 9 6 0 0.0
13 I0 13 13 0.0478 13 0.0478 0 0.0 0 0.0
13 I 14 4 0.0570 5 0.0248 5 0.0248 7 0.0039
13 2 14 5 0.0981 6 0.0478 7 0.0214 8 0.0088
13 3 14 7 0.0659 8 0.0310 9 0.0130 I0 0 . 0 0 4 8
13 4 14 g 0.0789 9 0.0377 I0 0 . 0 1 5 7 II 0 . 0 0 5 6
13 5 14 9 o.og71 IO 0.0412 Ii 0 . 0 1 6 5 12 0 . 0 0 5 3
13 6 14 IO 0.0906 II 0.0412 12 0 . 0 1 5 1 13 0 . 0 0 4 0
13 7 14 II 0.0891 12 0.0373 13 0 . 0 1 1 4 14 0 . 0 0 1 9
13 8 14 12 0.0817 13 0.0290 14 0 . 0 0 5 8 14 0 . 0 0 5 8
13 9 14 13 0.0667 14 0.0159 14 0 . 0 1 5 9 0 0.0
13 I0 14 14 0.0407 I4 0.0407 00.0 00.0
13 ii 14 14 0.0978 0 0.0 00.0 00.0
13 I 15 4 0.0667 5 0.0306 6 0.0133 7 0.0054
13 2 15 6 0.0604 7 0.0290 8 0.0128 9 0.0052
13 3 15 7 0.0855 8 0.0434 9 0.0201 I0 0 . 0 0 8 3
13 4 15 9 0.0554 I0 0.0259 Ii 0 . 0 1 0 7 12 0 . 0 0 3 7
13 5 15 i0 0.0642 ii 0.0298 12 0 . 0 1 1 7 13 0 . 0 0 3 7
13 6 15 II 0.0695 12 0.0310 13 0 . 0 1 1 1 14 0 , 0 0 2 9
13 7 15 12 0.0706 13 0.0290 14 0 . 0 0 8 7 14 0 . 0 0 8 7
13 8 15 13 0.0667 14 0.0232 14 0 . 0 2 3 2 15 0 . 0 0 4 6
13 9 15 14 0.0558 ib 0.0131 15 0 . 0 1 3 1 0 0.0
13 i0 15 15 0.0349 15 0.0349 0 0.0 0 0.0
13 Ii 15 15 0.0873 0 0.0 0 0.0 0 0.0
14 1 Ii 3 0.0717 4 0.0261 5 0.0087 5 0.0087
14 2 Ii 4 0.0957 5 0.0391 6 0.0141 7 0.0044
14 3 ii 6 0.0433 6 0.0433 7 0.0159 80. 0048
14 4 Ii 6 0.0982 7 0.0416 B 0.0146 9 0.0040
14 5 Ii 7 0.08~7 8 0.035~ 9 0.0114 i0 0 . 0 0 2 6
14 6 II 8 0.0749 9 0.0273 I0 0 . 0 0 7 0 i0 0 . 0 0 7 0
14 7 ii 9 0.0576 i0 0.0171 i0 0 . 0 1 7 1 ii 0 . 0 0 2 8
14 8 ii I0 0.0377 I0 0.0377 ii 0 . 0 0 7 1 II 0 . 0 0 7 1
14 9 II i0 0.0759 II 0.0170 ii 0 . 0 1 7 0 00.0
14 i0 II Ii 0.0577 ii 0.0577 0 0.0 00.0
14 ii Ii ii 0.0791 0 0.0 0 0.0 00.0
14 1 12 3 0.0846 4 0.0331 5 0.0120 6 0.0040
14 2 12 5 0.0522 6 0.0209 6 0.0209 7 0.0074
I4 3 12 6 0.0612 7 0.0254 8 0.0091 R 0.0091
I4 4 I2 7 0.0633 8 0.0260 9 0.0089 9 0.0089
14 5 12 R 0.0602 9 0.0236 9 0.0236 i0 0 . 0 0 7 3
14 6 12 9 0.0529 I0 0.0187 I0 0 . 0 1 8 7 ii 0 . 0 0 4 7
14 7 i2 i0 0.0420 I0 0.0420 ii 0 . 0 1 2 2 12 0 . 0 0 1 9
14 8 12 i0 0.0847 ii 0.0283 12 0 . 0 0 5 2 12 0 . 0 0 5 2
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 143

T A B L E 15---Precedence tables (continued).

J~ J~ J~ J~
Mu ru N~ 0.10 Prob 0.05 Prob 0.025 Prob 0 . 0 1 Prob

14 9 12 II 0 . 0 6 0 0 12 0.0130 12 0 . 0 1 3 0 0 0.0
14 i0 12 12 0 . 0 3 0 4 12 0.0304 0 0.0 0 0.0
14 11 12 12 0 . 0 6 7 0 0 0.0 0 0.0 0 0.0
14 1 13 3 0.0978 4 0.0407 5 0.0159 6 0.0058
14 2 13 5 0.0667 6 0.0290 7 0.0114 8 0.0040
14 3 13 60.OSI7 7 0.0373 8 0.0151 9 0.0053
14 4 13 7 0.0891 0.0412 9 0.0165 i0 0.0056
14 5 13 8 0.0906 9 0.0412 i0 0 . 0 1 5 7 II 0.0048
14 6 13 90.OS71 I0 0 . 0 3 7 7 Ii 0 . 0 1 3 0 12 0.0032
14 7 13 i0 0 . 0 7 8 9 ii 0 . 0 3 1 0 12 0 . 0 0 8 8 12 0.0088
14 8 13 II 0 . 0 6 5 9 12 0 . 0 2 1 4 12 0 . 0 2 1 4 13 0.0059
14 9 13 12 0 . 0 4 7 8 12 0 . 0 4 7 8 13 0 . 0 1 0 1 0 0.0
14 i0 13 12 0 . 0 9 8 1 13 0 . 0 2 4 8 13 0 . 0 2 4 8 0 0.0
14 ii 13 13 0 . 0 5 7 0 0 0.0 0 0.0 0 0.0
14 I 14 4 0.0489 4 0,0489 5 0.0204 6 0.0080
14 2 14 5 0.0824 6 0.0384 7 0.0164 8 0.0064
14 3 14 7 0.0516 8 0.0230 8 0.0230 9 0.0092
14 4 14 8 0.0601 9 0.0271 I0 0 . 0 1 0 7 ii 0.0035
14 5 14 9 0.0642 i0 0 . 0 2 8 5 ii 0 . 0 1 0 7 12 0.0032
14 6 14 i0 0 . 0 6 4 2 ii 0 . 0 2 7 1 12 0 . 0 0 9 2 12 0,0092
14 7 14 Ii 0 . 0 6 0 1 12 0 . 0 2 3 0 12 0 . 0 2 3 0 13 0.0064
14 8 14 12 0 . 0 5 1 6 13 0 . 0 1 6 4 13 0 . 0 1 6 4 14 0.0029
14 9 14 13 0 . 0 3 S 4 13 0 . 0 3 8 4 14 0 . 0 0 8 0 14 O. O O S O
14 10 14 13 0 . 0 8 2 4 14 0 . 0 2 0 4 14 0 . 0 2 0 4 0 0.0
14 Ii 14 14 0 . 0 4 8 9 14 0 . 0 4 8 9 0 0.0 0 0.0
14 I 15 4 0.0575 5 0.0253 6 0.0105 7 0.0041
14 2 15 5 0.0990 6 0.0490 7 0.0225 8 0.0095
14 3 15 70.06SO 8 0.0329 9 0.0144 I0 0.0056
14 4 15 O.0a?l 9 0.0407 10 O . O I R O ii 0.0070
14 5 15 9 0.0919 i0 0 . 0 4 5 7 Ii 0 . 0 1 9 9 12 0.0073
14 6 15 i0 0 . 0 9 7 4 11 0 . 0 4 7 6 12 0 . 0 1 9 7 13 0.0065
14 7 15 Ii 0 . 0 9 S 7 12 0 . 0 4 6 0 13 0 . 0 1 7 3 14 0.0047
14 8 15 12 0 . 0 9 5 3 13 0 . 0 4 0 7 14 0 . 0 1 2 7 15 0.0022
14 9 15 13 0 . 0 S 6 2 14 0 . 0 3 1 1 15 0 . 0 0 6 3 15 0.0063
14 i0 15 14 0 . 0 6 9 5 15 0 . 0 1 6 9 15 0 . 0 1 6 9 0 0.0
14 ii 15 15 0 . 0 4 2 1 15 0 . 0 4 2 1 0 0.0 0 0.0
14 12 15 15 0 , 0 9 9 6 0 0.0 0 0.0 0 0.0
15 1 12 3 0.0752 4 0.0282 5 0.0098 5 0.0098
15 2 12 5 0.0435 5 0.0433 6 0.0165 7 0.0056
15 3 12 6 0.0493 6 0.0493 7 0.0194 8 0.0066
15 4 12 7 0.0493 7 0.0493 8 0.0191 9 0.0062
15 5 12 8 0.0450 8 0.0450 9 0.0165 i0 0.0048
15 6 12 8 0.0906 9 0.0377 I0 0 . 0 1 2 4 ii 0.0029
15 7 12 9 0.0757 10 0 . 0 2 8 2 II 0 . 0 0 7 5 Ii 0.0075
15 8 12 i0 0 . 0 5 7 8 ii 0 . 0 1 7 5 II 0 . 0 1 7 5 12 0.0029
15 9 12 ii 0 . 0 3 7 7 ii 0 . 0 3 7 7 12 0 . 0 0 7 2 12 0.0072
15 i0 12 ii 0 . 0 7 4 9 12 0 . 0 1 6 9 12 0 . 0 1 6 9 0 0.0
15 11 12 12 0 . 0 5 7 2 12 0 . 0 3 7 2 0 0.0 0 0.0
15 12 12 12 0 . 0 7 7 8 00.0 00.0 0 0.0
| ~.4 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

TABLE 15--Precedence tables (continued).

j~, j~= J~ j~,

Mu r~ Nx 0.10 Prob 0.05 Prob 0.025 Prob 0.01 Prob

15 1 13 3 0.0873 4 0.0349 5 0.0131 6 0.0046

15 2 13 5 0.0558 6 0.0232 6 0.0232 7 0.0087
15 3 13 6 0.0667 7 0.0290 8 0.0111 9 0.0037
15 4 13 7 0.0706 8 0.0310 9 0.0117 i0 0.0037
15 5 13 8 0.0695 9 0.0298 i0 0.0107 ii 0.0030
15 6 13 9 0.0642 I0 0.0259 11 0.0083 ii 0.0083
15 7 13 I0 0.0554 ii 0.0201 Ii 0.0201 12 0.0052
15 8 13 II 0.0434 ii 0.0434 12 0.0128 13 0.0021
15 9 13 ii 0.0855 12 0.0290 13 0.0054 13 0.0054
15 I0 13 12 0.0604 13 0.0133 13 0.0133 0 0.0
15 Ii 13 13 0.0306 13 0.0306 0 0.0 0 0.0
15 12 13 13 0.0667 0 0.0 0 0.0 0 0.0
15 i 14 3 0.0996 4 0.0421 5 0.0169 6 0.0063
15 2 14 5 0.0695 6 0.0311 7 0.0127 8 0.0047
15 3 14 6 0.0862 7 0.0407 8 0.0173 9 0.0065
15 4 14 7 0.0953 8 0.0460 9 0.0197 I0 0.0073
15 5 14 8 0.0987 9 0.0476 i0 0.0199 ii 0.0070
15 6 14 9 0.0974 I0 0,0457 II 0.0180 12 0.0056
15 7 14 i0 0.0919 Ii 0.0407 12 0.0144 13 0.0036
15 8 14 ii O.OS21 12 0.0329 13 0.0095 13 0.0095
15 9 14 12 0.0680 13 0.0225 13 0,0225 14 0.0041
15 I0 14 13 0.0490 13 0.0490 14 0.0105 0 0.0
15 ii 14 13 0.0990 14 0.0253 0 0.0 0 0.0
15 12 14 14 0.0575 0 0.0 0 0.0 0 0.0
15 1 15 4 0.0498 4 0.0498 5 0.0211 6 0.0084
15 2 15 5 0.0843 6 0.0400 7 0.0176 8 0.0071
15 3 15 7 0.0543 8 0.0251 9 0.0105 I0 0.0039
15 4 15 8 0.0641 9 0.0302 10 0.0127 ii 0.0046
15 5 15 9 0.0697 I0 0.0328 ii 0.0134 12 0.0046
15 6 15 i0 0.0716 ii 0.0328 12 0.0127 13 0.0039
15 7 15 ii 0.0697 12 0.0302 13 0.0105 14 0.0026
15 8 15 12 0.0641 13 0.0251 14 0.0071 14 0.0071
15 9 15 13 0.0543 14 0.0176 14 0.0176 15 0.0032
15 i0 15 14 0.0400 14 0.0400 15 0.0084 15 0.0084
15 ii 15 14 0.0843 15 0.0211 15 0.0211 0 0.0
15 12 15 15 0,0498 15 0,0498 0 0.0 0 0.0

Note for Table 15.

For
2<M~<15
[2,My - 3]max =< N~ ~ 15
where
J~ = number of X failures that precede the r th Y failure; the larger the value of
Jx, the less likely it would occur if Y = X,
M~ = size of (test) sample, and
N~ = size of reference sample.
NOTE: When the statistic tabulated has no value such that a sigmficance test is
possible at the desired level, this condition is stipulated by the probability entry 0.0.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 145

A P P E N D I X II

Comparing Two Life Samples (RCB Test Program)

Complete (uncensored) Data
T h e s i g n e d - r a n k test statistic [5] is a p p r o p r i a t e for use in c o m p a r i n g two life
samples generated using a r a n d o m i z e d complete block (RCB), t h a t is, paired
c o m p a r i s o n , test p r o g r a m o r g a n i z a t i o n a l structure.

Procedure
1. R a n k the a b s o l u t e values o f the differences (y~ - x 0 , namely, I Y~ - x, [ = I d,I,
f r o m smallest to largest.
2. T h e n a t t a c h the sign to r yl - x~l = I d~l a c c o r d i n g to the sign o f (y~ - x~).
This two step p r o c e d u r e thus gives the signed r a n k s o f interest in this analysis.
3. C o m p u t e the s u m of the negative r a n k s st.
4. I f this s u m is very small ( t h a t is, if the respective (y~ - x,) are mostly positive)
t h e n it m a y be unlikely t h a t such a small value o f s~ would be o b s e r v e d by c h a n c e
u n d e r the null h y p o t h e s i s t h a t Y = X. Test the null hypothesis by c o m p a r i n g the
o b s e r v e d value sr to the a p p r o p r i a t e t a b u l a t e d value (srL, T a b l e 16, where ~, the

TABLE 16--Wilcoxon's signed rank statistic Sr~.

Sr~ Src~ Sra Sra

N 0.10 Prob 0.05 Prob 0. 025 Prob 0.01 Prob
4 0 0.0625 0 0.0 0 0.0 0 0.0
5 2 0.0938 0 0.0313 0 0.0 0 0.0
6 3 0.0781 2 0.0469 0 0.0156 0 0.0
7 5 0.0781 3 0.0391 2 0.0234 0 0.0078
8 8 0.0977 5 0.0391 3 0.0195 1 0.0078
9 10 0.0820 8 0.0488 5 0.0195 3 0.0098
10 14 0.0967 10 0.0420 8 0.0244 5 0.0098
I1 17 0.0874 13 0.0415 10 0.0210 7 0.0093
12 21 0.0881 17 0.0461 13 0.0212 9 0.0081
13 26 0.0955 21 0.0471 17 0.0239 12 0.0085
14 31 0.0969 25 0.0452 21 0.0247 15 0.0083
15 36 0.0938 30 0.0473 25 0.0240 19 0.0090
16 42 0.0964 35 0.0467 29 0.0222 23 0.0091
17 48 0.0950 41 0.0492 34 0.0224 27 0.0087
18 55 0.0982 47 0.0494 40 0.0241 32 0.0091
19 62 0.0978 53 0.0478 46 0.0247 37 0.0090
20 69 0.0947 60 0.0487 52 0.0242 43 0.0096

For
4<N=<20
where
SR = sum of the signed ranks for all pmred comparisons in which yi < x~ (see text for
counting procedure); the smaller the value of SR, the less likely it would occur
if Y = X, and
N = number of paired comparisons (blocks).
NOTE: When the statistic tabulated has no value such that a significance test is possible
at the desired level, this condition is stipulated by the probability entry 0.0.
146 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

Type I error, equals 0.10, 0.05, 0.025, and 0.01. IfSr < (Sr)a, reject the null hypothesis
that Y = X and opt instead for the alternative hypothesis Y > X.
Example--Given the following paired comparison fatigue life data in cycles.

Block X Y

1 128 1300 133 000

2 202 000 338 000
3 219 000 207 009
4 189 000 412 000
5 234 000 304 000
6 389 000 217 000
7 356 000 312 000
8 121 000 194 000

Test the null hypothesis Y = X with a Type I error equal 0.05.

Solution--First, construct the following table

Block d~ [d~J Rank Signed Rank

1 -}-5 000 5 000 1 +1

2 ,},136 000 136 000 7 -}-7
3 - 1 2 000 12 000 2 -2
4 -}-223 000 223 000 8 -}-8
5 -}-70 000 70 000 4 -}-4
6 .}.128 000 128 000 6 -}-6
7 - 4 4 000 44 000 3 -3
8 -}-73 000 73 000 5 -}-5

Accordingly, st, the sum of the negative ranks, equals 5 (2 + 3).

From Table 16, with ~ = 0.05 (and N --- 8), the critical value (st), is 5. Thus, we
reject the null hypothesis (with exact Type I error equal 0.0391) and believe instead
that Y > X.

Censored Data
When both tests within a block are suspended (at the same duration), the block
provides no information regarding the better material (processing, configuration)
and, therefore, may be neglected in analysis. However, if only one test is suspended
in one or more blocks, then the signed rank test is literally not applicable because
the exact ranks associated with the blocks with suspended tests are unknown.
Regardless of the number of blocks with one suspended test, the life data within
blocks provide as much information as two observed life values when working
with the simple sign statistic (that is, the sign test) [5] based on the well-known
binomial distribution with P equal 0.5. The sign statistic requires knowledge of
only which sample exhibits the greater life within a block, and requires no informa-
tion regarding the magnitude of the difference, tk. (The sign test may thus be used
to compare either censored and uncensored life samples generated by a paired
comparison test program. However, because the sign statistic in effect ignores
certain information, its statistical power is lower than the analogous power for the
signed rank statistic. Thus, the signed rank test is always preferred to the sign test
when both are applicable.)
REFERENCES 147

References
[1] Natrella, M. G., Experimental Statistics, National Bureau of Standards Handbook 91,
U.S. Government Office, Washington, D.C., 1963 (this reference should be in your
library).
[2] Finney, D. J., An Introduction to the Theory of Experimental Design, The Umversity of
Chicago Press, Chicago, 1960.
[3] Cox, D. R., Planning of Experiments, Wdey, New York, 1958.
[4] Youden, W. J., Statistical Design, Industrial and Engineering Chemistry Reprints,
American Chemical Society Applied Publications, Washington, D.C.
[5] Conover, W. J., Practical Nonparametric Statistics, Wiley, New York, 1971.
[6] Snedecor, G. W. and Cochran, W. G., Statistical Methods, 6th edition, Iowa State
University Press, Ames, Iowa, 1967.
[7] Brownlee, K. A., Statistical Theory and Methodology in Science and Engineering,
2nd edition, Wiley, New York, 1965.
[8] Hald, A., Statistical Theory with Engineering Applications, Wiley, New York, 1952.
[9] Little, R. E. and Jebe, E. H., Statistical Design of Fatigue Experiments, Applied
Science Publishers Ltd, London, 1975,
[10] Wilk, M. B. and Kempthorne, O., "Fixed, Mixed, and Random Models," Journal of
the American Statistical Association, Vol. 50, 1955, pp. 1144-1167.
[11] Little, R. E., "A Note on Selecting the Better Material in a Paired Comparison Test
Program," Journal of Testing and Evaluation, Vol. 2, 1974, pp. 235-239.
[12] Kastenbaum, M. A., Hoel, D. G., and Bowman, K. O., "Sample Size Requirements:
Randomized Block Designs," Biometrika, Vol. 57, Part 3, 1970, pp. 573-577.
[13] Bowker, A. H. and Lieberman, G. J., Engh~eering Statistics, Prentice-Hall, Inc.,
Englewood Chffs, N.J., second edition, 1972.
[14] Little, R. E. and Jebe, E. H., "A Note on the Gain in Precision for Optimal Allocation
in Regression as Applied to Extrapolation in S-N Fatigue Testing, "'Technometrics,
Vol. 11, 1969, pp. 389-392.
[15] Little, R. E., "Choosing the R~ght Fatigue Test," Machine Design, Vol. 39, No. 28,
7 Dec. 1967.
[16] Little, R. E., '% Rehabdlty Analysis of Fatigue L~mlts Based on Large Sample Quantal
Response Data," Mechantcal Behavior oJ Materials, Vol. V, The Society of Materials
Science, Japan, 1972, pp. 471-479.
[17] Little, R. E., "Estimating the Medmn Fatigue Limit for Very Small Up-and-Down
Quantal Response Tests and for S-N Data with Runouts," Probabdisric Aspects of
Fatigue, ASTM STP 511, American Society for Testing and Materials, 1972, pp. 29-42.
[18] Little, R. E., "Handbook for Up-and-Down Testmg with Small Samples," unpub-
lished.
[19] Dixon, W. J., "The Up-and-Down Method for Small Samples," Journal of the Ameri-
can Statistical Association, Vol. 60, 1965, pp. 967-978.
[20] Little, R. E., "The Up-and-Down Method for Small Samples with Extreme Value
Response Distributions," Journal of the American Statistical Association, Vol. 69,
1974, pp. 202-206.
[21] Young, D. H., "Distributions of Some Censored Rank Statistics Under Lehmann
Alternatives for the Two-Sample Case," Biometrika, Vol. 60, 1973, pp. 543-549.
[22] Basu, A. P., "On a Generalized Savage Statistic with Apphcations to Life Testing,"
Annals of Mathematical Statistics, Vol. 39, I968, pp. 1591-1604.
[23] Mann, H. B. and Whitney, D. R., "On a Test of Whether One of Two Random
Variables is Stochastically Larger Than the Other," Annals of Mathematical Statistics,
Vol. 18, 1947, pp. 50--60.
[24] Shorack, R. A., "Tables of the Distribution of the Mann-Whitney-WilcoxonU-Sta-
tistic Under Lehmann Alternatives," Technometrics, Vol. 9, 1967, pp. 666-677.
[25] Halperin, M., "Extension of the Wilcoxon-Mann-WhitneyTest to Samples Censored
at the Same Fixed Point, "Journal o] the American Statistical Association, Vol. 55,
1960, pp. 125-138.
[26] Shorack, R. A., "Recursive Generation of the Distribution of Several Non-Parametric
Test Statistics Under Censoring," Journal of the American Statistical Association,
Vol. 63, 1968, pp. 353-366.
148 MANUAL ON STATISTICAL PLANNING AND ANALYSIS

[27] Young, D. H., "Consideration of Power for Some Two Sample Tests with Censoring
Based on a Given Order Statistic," Biometrika, Vol. 57, 1970, pp. 595-604.
[28] Eilbott, J. and Nadler, J., "On Precedence Life Testing," Technometrics, Vol. 7, 1965,
pp. 359-377.
[29] Cochran, W. G. and Cox, G. M., Experimental Designs, 2nd edition, Wiley, New
York, 1957.
[30] Moses, L. E. and Oakford, R. V., Tables of Random Permutations, Stanford University
Press, Stanford, Calif., 1963.
CHAPTER 4 ON DISTRIBUTION-FREE STATISTICAL TESTS 149

Acknowledgments
I would especially like to acknowledge the continued efforts of H. S.
Reemsnyder and E. H. Jebe in bringing this manual to fruition. Dr.
Reemsnyder worked diligently over the last several years preparing work-
ing documents and kindling interest in planned experiments. Dr. Jebe
headed a design of experiments task group in ASTM Subcommittee
E09.06 charged with developing chapters on the statistical planning of
fatigue experiments for the proposed revision of A S T M STP 91A. The
first two chapters of the present manuscript evolved from the residue of
my work on that task group. Dr Jebe's continued assistance in reading
and criticising the entire manuscript have been invaluable. I would also
like to acknowledge the extensive and detailed comments made on the
first two chapters by W. S. Hyler, J. F. Throop, and H. S. Reemsnyder.
Finally, I would like to thank the Computing Center of the University
of Michigan, Dearborn Campus, for their financial support in developing
the computer programs used to generate the statistical tables presented
in Chapter 4.
R. E. Little
Dearborn, Mich.
8 Oct. 1975

STP 536-1973
No ratings yet
STP 536-1973
504 pages
The Impact of Gender Discrimination in The Workplace To Employees Emotional Well Being
No ratings yet
The Impact of Gender Discrimination in The Workplace To Employees Emotional Well Being
54 pages
Astm A480-A480m (2022)
100% (3)
Astm A480-A480m (2022)
25 pages
Astm E606 E606m 21
100% (1)
Astm E606 E606m 21
7 pages
Pantaloon Visual Merchandising
75% (8)
Pantaloon Visual Merchandising
45 pages
ISO 7870-4 - 2011 (E) - Cumulative Sum Charts
100% (2)
ISO 7870-4 - 2011 (E) - Cumulative Sum Charts
70 pages
Astm E606
75% (4)
Astm E606
16 pages
Fatigue
33% (3)
Fatigue
32 pages
DS5S2 - (1969) An Evaluation of The Yield, Tensilde, Creep and Rupture Strengths of Wrought 304, 316, 321, and 347 Stainless Steels at Elevated Temperatures
100% (1)
DS5S2 - (1969) An Evaluation of The Yield, Tensilde, Creep and Rupture Strengths of Wrought 304, 316, 321, and 347 Stainless Steels at Elevated Temperatures
74 pages
Astm 2017-D7678.40387
No ratings yet
Astm 2017-D7678.40387
10 pages
Iso 12107
No ratings yet
Iso 12107
1 page
DS7S1 - (1968) The Elevated-Temperature Properties of Selected Superalloys
No ratings yet
DS7S1 - (1968) The Elevated-Temperature Properties of Selected Superalloys
28 pages
Astm E739
No ratings yet
Astm E739
7 pages
ASTME466 07ConductingForceControlledConstantAmplitudeAxialFatigueTests
100% (1)
ASTME466 07ConductingForceControlledConstantAmplitudeAxialFatigueTests
5 pages
2.3 C1 Readings in Philippine History
No ratings yet
2.3 C1 Readings in Philippine History
38 pages
Conducting Force Controlled Constant Amplitude Axial Fatigue Tests of Metallic Materials
100% (1)
Conducting Force Controlled Constant Amplitude Axial Fatigue Tests of Metallic Materials
6 pages
The Stregth, Fracture Toughnees and Low Cycle 17-4PH PDF
No ratings yet
The Stregth, Fracture Toughnees and Low Cycle 17-4PH PDF
11 pages
How To Write Research Proposal
No ratings yet
How To Write Research Proposal
24 pages
NLM - Controlling
No ratings yet
NLM - Controlling
8 pages
0 Undergraduate Research Project 2018 2019
No ratings yet
0 Undergraduate Research Project 2018 2019
5 pages
E468 PDF
100% (1)
E468 PDF
6 pages
Masters Course of Transportation
No ratings yet
Masters Course of Transportation
43 pages
Experimental Method of Research
No ratings yet
Experimental Method of Research
21 pages
Ductile Fracture Initiation and Propagation Modeling Using Damage Plasticity Theory
No ratings yet
Ductile Fracture Initiation and Propagation Modeling Using Damage Plasticity Theory
18 pages
Khatib & Alami MAPA Construction & Trade Co. Inc.: Document Transmittal
No ratings yet
Khatib & Alami MAPA Construction & Trade Co. Inc.: Document Transmittal
5 pages
STP 595-1976
No ratings yet
STP 595-1976
344 pages
Azu TD 9502612 Sip1 M PDF
No ratings yet
Azu TD 9502612 Sip1 M PDF
276 pages
P 7.2-01 Prosedür - Müşteri İle İlişkili Prosesler Prosedürü - Eng
No ratings yet
P 7.2-01 Prosedür - Müşteri İle İlişkili Prosesler Prosedürü - Eng
4 pages
ASTM - STP 407 - Temper Embrittlement in Steel 1968 PDF
100% (2)
ASTM - STP 407 - Temper Embrittlement in Steel 1968 PDF
260 pages
Risk Management and Analysis in Hydro-Electric Projects in India
No ratings yet
Risk Management and Analysis in Hydro-Electric Projects in India
7 pages
STP 559-1974
No ratings yet
STP 559-1974
319 pages
Final Thesis Habtamu Wubneh Corrected
No ratings yet
Final Thesis Habtamu Wubneh Corrected
91 pages
Non Experimental Research
No ratings yet
Non Experimental Research
48 pages
STP 590-1976
No ratings yet
STP 590-1976
506 pages
Sistem Pentaksiran Baru STPM The New STPM Assessment System For Chemistry
No ratings yet
Sistem Pentaksiran Baru STPM The New STPM Assessment System For Chemistry
20 pages
BFC3042 BFC31802
No ratings yet
BFC3042 BFC31802
13 pages
Integrated Product Conceptualization Syllabus
No ratings yet
Integrated Product Conceptualization Syllabus
5 pages
STP 527-1973
No ratings yet
STP 527-1973
119 pages
Chapter Three 3.0 Methodology: 1. Development of Prototype System
No ratings yet
Chapter Three 3.0 Methodology: 1. Development of Prototype System
3 pages
For Students
No ratings yet
For Students
3 pages
DS63 - (1985) References and Conference Proceedings Towards The Understanding of Fracture Mechanics
100% (1)
DS63 - (1985) References and Conference Proceedings Towards The Understanding of Fracture Mechanics
58 pages
STP1303 Eb.1415051 1 PDF
100% (1)
STP1303 Eb.1415051 1 PDF
157 pages
STP 556-1974
No ratings yet
STP 556-1974
192 pages
Tiqa Student Template
No ratings yet
Tiqa Student Template
1 page
STP91A-EB.1595294-1 Fatiga
0% (1)
STP91A-EB.1595294-1 Fatiga
94 pages
Fatigue and Fracture Testing ASTM PDF
No ratings yet
Fatigue and Fracture Testing ASTM PDF
164 pages
MAS202 Final Project
No ratings yet
MAS202 Final Project
12 pages
Data Analysis and Evaluation
No ratings yet
Data Analysis and Evaluation
34 pages
1 s2.0 S0140197109001079 Main
No ratings yet
1 s2.0 S0140197109001079 Main
8 pages
Finnie's Notes On Fracture Mechanics - C.K.H. Dharan Et Al. (Springer, 2016) PDF
No ratings yet
Finnie's Notes On Fracture Mechanics - C.K.H. Dharan Et Al. (Springer, 2016) PDF
259 pages
CPD - Lesson Plan
No ratings yet
CPD - Lesson Plan
4 pages
Norma ASTM E606
100% (1)
Norma ASTM E606
16 pages
Seminar On Brittle and Ductile Fracture
100% (4)
Seminar On Brittle and Ductile Fracture
29 pages
Extending Steinberg
100% (1)
Extending Steinberg
35 pages
E606 Testing Method
100% (1)
E606 Testing Method
16 pages
Magnesium-Alloy Extruded Bars, Rods, Profiles, Tubes, and Wire
No ratings yet
Magnesium-Alloy Extruded Bars, Rods, Profiles, Tubes, and Wire
19 pages
Mechanical Properties and Microstructure of Forged Steel
No ratings yet
Mechanical Properties and Microstructure of Forged Steel
174 pages
QL R Titles That Can Be Done by Grade 11
No ratings yet
QL R Titles That Can Be Done by Grade 11
1 page
Little R E Mechanical Reliability Improvement Probability and Statistics For Experimental Testing Marcel Dekker 2001 PDF
No ratings yet
Little R E Mechanical Reliability Improvement Probability and Statistics For Experimental Testing Marcel Dekker 2001 PDF
595 pages
Dot-Faa-Ar-05 - 15 - Aug2005
No ratings yet
Dot-Faa-Ar-05 - 15 - Aug2005
126 pages
Sae j577 201509 實驗室跟108 一樣的衝擊測試
No ratings yet
Sae j577 201509 實驗室跟108 一樣的衝擊測試
27 pages
Reporting Uniaxial Strength Data and Estimating Weibull Distribution Parameters For Advanced Ceramics
No ratings yet
Reporting Uniaxial Strength Data and Estimating Weibull Distribution Parameters For Advanced Ceramics
17 pages
(Julie A. Bannantine) Fundamentals of Metal Fatigu (BookFi - Org) - 1
100% (1)
(Julie A. Bannantine) Fundamentals of Metal Fatigu (BookFi - Org) - 1
283 pages
Strain-Controlled Fatigue Testing: Standard Test Method For
No ratings yet
Strain-Controlled Fatigue Testing: Standard Test Method For
16 pages
Effect of Shot Peening On The Fatigue Life of 2024 Aluminum Alloy PDF
No ratings yet
Effect of Shot Peening On The Fatigue Life of 2024 Aluminum Alloy PDF
12 pages
E143 6837
No ratings yet
E143 6837
5 pages
A Study To Evaluate The Effectiveness of Structured Teaching Program On Knowledge Regarding
No ratings yet
A Study To Evaluate The Effectiveness of Structured Teaching Program On Knowledge Regarding
19 pages
Hexcel 8552 IM7 Prepreg Tests
No ratings yet
Hexcel 8552 IM7 Prepreg Tests
238 pages
Notch Chapter 7
No ratings yet
Notch Chapter 7
132 pages
Flexural Strength of Manufactured Carbon and Graphite Articles Using Three-Point Loading at Room Temperature
No ratings yet
Flexural Strength of Manufactured Carbon and Graphite Articles Using Three-Point Loading at Room Temperature
5 pages
Artigo - Course of SN-curves Especcially in The High-Cycle Fatigue Regime With Regard To Component Design and Safety PDF
No ratings yet
Artigo - Course of SN-curves Especcially in The High-Cycle Fatigue Regime With Regard To Component Design and Safety PDF
13 pages
Relating To Fatigue and Fracture Testing: Standard Terminology
100% (3)
Relating To Fatigue and Fracture Testing: Standard Terminology
25 pages
Ansi S2.26-2001 (R2006) - 1
No ratings yet
Ansi S2.26-2001 (R2006) - 1
22 pages
Static and Kinetic Friction Coefficients
No ratings yet
Static and Kinetic Friction Coefficients
6 pages
STP566 Foreword PDF
No ratings yet
STP566 Foreword PDF
11 pages
Curve Determination: Standard Test Method For
No ratings yet
Curve Determination: Standard Test Method For
16 pages
Assessment Learning 1 Chapter-3
No ratings yet
Assessment Learning 1 Chapter-3
69 pages
Abstract - Fatigue Behavior in Strain Cycling in The Low and Intermediate Cycle Range
No ratings yet
Abstract - Fatigue Behavior in Strain Cycling in The Low and Intermediate Cycle Range
55 pages
The Effect of Porosity On The Fatigue Life of Cast Aluminium-Silicon Alloys
No ratings yet
The Effect of Porosity On The Fatigue Life of Cast Aluminium-Silicon Alloys
12 pages
Evaluation of The VDA 238-100 Tight Radius Bending Test Using Digital Image Correlation Strain Measurement
No ratings yet
Evaluation of The VDA 238-100 Tight Radius Bending Test Using Digital Image Correlation Strain Measurement
9 pages
NX Nastran Advanced Nonlinear - Solution 601-701
No ratings yet
NX Nastran Advanced Nonlinear - Solution 601-701
3 pages
Power Analysis For Experimental Research A Practical Guide For The Biological Medical and Social Sciences 1st Edition R. Barker Bausell (Author)
No ratings yet
Power Analysis For Experimental Research A Practical Guide For The Biological Medical and Social Sciences 1st Edition R. Barker Bausell (Author)
56 pages
Stress Fatigue Principles
No ratings yet
Stress Fatigue Principles
12 pages
Engineering Failure Analysis-Crack Growth
No ratings yet
Engineering Failure Analysis-Crack Growth
27 pages
Effects of Participation in Student-Based School Clubs On Self-Esteem of Adolescents - 9208987196
No ratings yet
Effects of Participation in Student-Based School Clubs On Self-Esteem of Adolescents - 9208987196
19 pages
Nave Et Al 2018 Musical Preferences Predict Personality - Evidence From Active Listening and Facebook Likes
No ratings yet
Nave Et Al 2018 Musical Preferences Predict Personality - Evidence From Active Listening and Facebook Likes
20 pages
A Practical Method For Statistical Analysis of Strain-Life Fatigue Data
100% (1)
A Practical Method For Statistical Analysis of Strain-Life Fatigue Data
10 pages
Aluminium Embrittlement
No ratings yet
Aluminium Embrittlement
8 pages
Damage Tolerance of Wrought Alloy 718 Ni - Fe-Base Superalloy: M. Chang, A.K. Koul, P. Au, and T Terada
No ratings yet
Damage Tolerance of Wrought Alloy 718 Ni - Fe-Base Superalloy: M. Chang, A.K. Koul, P. Au, and T Terada
11 pages
ASTM - STP 1058 - Fatigue and Fracture Testing of Weldments
No ratings yet
ASTM - STP 1058 - Fatigue and Fracture Testing of Weldments
310 pages
Astm E739 91 1998
No ratings yet
Astm E739 91 1998
3 pages
Significance of Safety Monitoring
No ratings yet
Significance of Safety Monitoring
25 pages
My Work
No ratings yet
My Work
37 pages
API STD 6ACRA - E1 - Add - 1
No ratings yet
API STD 6ACRA - E1 - Add - 1
8 pages
Experimental Statistics
From Everand
Experimental Statistics
Mary Gibbons Natrella
4/5 (2)