Introduction to Design of Experiments with JMP

Examples Third Edition SAS Press Jacques Goupy

pdf download

https://ebookname.com/product/introduction-to-design-of-
experiments-with-jmp-examples-third-edition-sas-press-jacques-
goupy/

Get Instant Ebook Downloads – Browse at https://ebookname.com

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Design and Analysis of Experiments Introduction to

Experimental Design Volume 1 Second Edition Klaus
Hinkelmann

https://ebookname.com/product/design-and-analysis-of-experiments-
introduction-to-experimental-design-volume-1-second-edition-
klaus-hinkelmann/

JMP 8 Statistics and Graphics Guide 2nd Edition Sas

Publishing

https://ebookname.com/product/jmp-8-statistics-and-graphics-
guide-2nd-edition-sas-publishing/

SAS Guide to Report Writing Examples 2nd Edition

Michele M. Burlew

https://ebookname.com/product/sas-guide-to-report-writing-
examples-2nd-edition-michele-m-burlew/

Russian function words Meanings and use Conjunctions

interjections parenthetical words particles
prepositions 1st Edition Marina Rojavin

https://ebookname.com/product/russian-function-words-meanings-
and-use-conjunctions-interjections-parenthetical-words-particles-
prepositions-1st-edition-marina-rojavin/
Diagnosis for Physical Therapists A Symptom Based
Approach 1st Edition Todd E. Davenport

https://ebookname.com/product/diagnosis-for-physical-therapists-
a-symptom-based-approach-1st-edition-todd-e-davenport/

Crisis Decision Making 1st Edition Chien-Ta Bruce Ho

https://ebookname.com/product/crisis-decision-making-1st-edition-
chien-ta-bruce-ho/

Final freedom Michael Vorenberg

https://ebookname.com/product/final-freedom-michael-vorenberg/

O Tempora O Mores Marcus Tullius Cicero

https://ebookname.com/product/o-tempora-o-mores-marcus-tullius-
cicero/

Detective Fiction in Cuban Society and Culture 1st

Edition Stephen Wilkinson

https://ebookname.com/product/detective-fiction-in-cuban-society-
and-culture-1st-edition-stephen-wilkinson/
Rome s Gothic Wars From the Third Century to Alaric 1st
Edition Michael Kulikowski

https://ebookname.com/product/rome-s-gothic-wars-from-the-third-
century-to-alaric-1st-edition-michael-kulikowski/
Introduction to
Design of Experiments
with JMP Examples
®

Third Edition

Jacques Goupy
Lee Creighton
The correct bibliographic citation for this manual is as follows: Goupy, Jacques, and Lee Creighton. 2007.
Introduction to Design of Experiments with JMP® Examples, Third Edition. Cary, NC: SAS Institute Inc.

Introduction to Design of Experiments with JMP® Examples, Third Edition

Copyright © 2007, SAS Institute Inc., Cary, NC, USA

ISBN 978-1-59994-422-7

All rights reserved. Produced in the United States of America.

For a hard-copy book: No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, or otherwise, without the
prior written permission of the publisher, SAS Institute Inc.

For a Web download or e-book: Your use of this publication shall be governed by the terms established by
the vendor at the time you acquire this publication.

U.S. Government Restricted Rights Notice: Use, duplication, or disclosure of this software and related
documentation by the U.S. government is subject to the Agreement with SAS Institute and the restrictions set
forth in FAR 52.227-19, Commercial Computer Software-Restricted Rights (June 1987).
SAS Institute Inc., SAS Campus Drive, Cary, North Carolina 27513.

1st printing, October 2007

2nd printing, June 2008
®
SAS Publishing provides a complete selection of books and electronic products to help customers use SAS
software to its fullest potential. For more information about our e-books, e-learning products, CDs, and hard-
copy books, visit the SAS Publishing Web site at support.sas.com/publishing or call 1-800-727-3228.
®
SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS
Institute Inc. in the USA and other countries. ® indicates USA registration.

Other brand and product names are registered trademarks or trademarks of their respective companies.
Contents

Preface xiii

Chapter 1 Gaining Knowledge with Design of

Experiments 1
1.1 Introduction 2
1.2 The Process of Knowledge Acquisition 2
1.2.1 Choosing the Experimental Method 5
1.2.2 Analyzing the Results 5
1.2.3 Progressively Acquiring Knowledge 5
1.3 Studying a Phenomenon 5
1.4 Terminology 6
1.4.1 Factor Types 6
1.4.2 Experimental Space 7
1.4.3 Factor Domain 8
1.4.4 Study Domain 9
1.5 Centered and Scaled Variables 11
1.6 Experimental Points 13
1.7 Design of Experiments 14
1.7.1 Methodology of Designs without Constraints 14
1.7.2 Methodology of Designs with Constraints 15
1.7.3 The Response Surface 17
1.7.4 The a priori Mathematical Model of the Response 18

Chapter 2 Your First Designed Experiment 21

2.1 Introduction 21
2.2 Example 1: Control Your Car’s Gas Consumption 22
2.2.1 Preparing the Designed Experiment 22
2.2.2 Choosing the Design 24
iv Contents

2.2.3 Running the Experiment 27

2.2.4 Interpreting the Coefficients 28
2.2.5 Interpreting the Results 40

Chapter 3 A Three-Factor Designed Experiment 45

3.1 Introduction 45
3.2 Example 2: Gold Jewelry 46
3.2.1 Preparing the Designed Experiment 46
3.2.2 Choosing the Design 51
3.2.3 Running the Experiment 53
3.2.4 Interpreting the Results 54
3.2.5 Study Conclusion 67

Chapter 4 Four-Factor Full-Factorial Experiments 69

4.1 Introduction 69
4.2 Example 3: The Galette des Rois 70
4.2.1 Preparing the Designed Experiment 70
4.2.2 Choosing the Design 73
4.2.3 Running the Experiment 74
4.2.4 Interpreting the Results 76
4.2.5 Desirability Functions 83
4.2.6 Study Conclusion 91

Chapter 5 Statistical Concepts for Designed

Experiments 93
5.1 Introduction 94
5.2 Example 4: Lifespan of Disposable Cutting Tools 95
5.2.1 Preparing the Designed Experiment 95
5.2.2 Experimentation 97
5.2.3 Intepreting the Results of the Calculations 98
5.2.4 Statistical Concepts Used in DOE 99
5.2.5 Factors and Influential Interactions 113
 Contents v

5.2.6 Analysis of Variance (ANOVA) 115

5.2.7 Application 119
5.2.8 Residual Analysis 125
5.2.9 Study Conclusion 126

Chapter 6 Fractional Factorial and Screening

Designs 131
6.1 Introduction 132
6.2 Example 5: Measuring Tellurium Concentration 133
6.2.1 Preparing the Designed Experiment 133
6.2.2 Running the Experiment 135
6.2.3 Interpreting the Results 136
6.3 Alias Theory 137
6.3.1 Definition of Contrasts 138
6.3.2 New Interpretation Hypotheses 140
6.4 Box Calculations 141
6.4.1 Box Notation 141
6.4.2 Operations on the Column of Signs 142
6.4.3 Rules to Remember 144
6.5 Equivalence Relation 144
6.5.1 Basic Design 144
6.5.2 Equivalence Relation 146
6.6 Alias Generating Functions 147
6.6.1 Alias Generating Function of the Upper Half Design 147
6.6.2 Alias Generating Function of the Lower Half Design 148
6.6.3 Reading Aliases in Software 149
6.7 Practical Construction of a Fractional Factorial Design 149
3
6.7.1 2 Full Factorial Design 151
4-1
6.7.2 2 Fractional Factorial 152
5-2
6.7.3 2 Fractional Factorial 154
7-4
6.7.4 2 Fractional Factorial 156
6.8 Maximum Number of Factors from a Base Design 157
vi Contents

6.9 Aliasing Theory with the Tellurium Example 157

6.9.1 Confounding in the Tellurium Example 158
6.9.2 Application of the Interpretation Assumption to the
Tellurium Example 158
6.9.3 Study Conclusion 159

Chapter 7 Examples of Fractional Factorial Designs 161

7.1 Introduction 162
7.2 Example 6: Sulfonation 163
7.2.1 Preparing the Designed Experiment 163
7.2.2 Carrying Out the Experiment 165
7.2.3 Interpreting the Results 166
7.2.4 Study Conclusion 173
7.3 Example 7: Spectrofluorimetry 174
7.3.1 Preparing the Designed Experiment 174
7.3.2 Carrying Out the Experiment 180
7.3.3 Interpreting the Results 180
7.3.4 Building the Complementary Design 190
7.3.5 Study Conclusion 196
7.4 Example 8: Potato Chips 198
7.4.1 Preparing the Designed Experiment 198
7.4.2 Carrying Out the Experiment 201
7.4.3 Interpreting the Results 201
7.4.4 Study Conclusion 204

Chapter 8 Trial Order and Blocking 205

8.1 Introduction 206
8.2 The Nature of Errors 207
8.2.1 Blocking 208
8.2.2 Anti-Drift Designs 209
8.2.3 Random Variation 210
8.2.4 Small Systematic Variations 211
 Contents vii

8.3 Example 9: Penicillium Chrysogenum (Blocking Example) 212

8.3.1 Preparing the Designed Experiment 212
8.3.2 Constructing the Plan and Checking the Blocking
Advantage 214
8.3.3 Carrying Out the Trials 217
8.3.4 Interpreting the Results 218
8.3.5 Study Conclusion 224
8.4 Example 10: Yates’s Beans 225
8.4.1 Preparing the Designed Experiment 225
8.4.2 Running the Experiment 228
8.4.3 Interpreting the Results 230
8.4.4 Study Conclusion 232
8.5 Example 11: “The Crusher” (Example of an Anti-Drift
Design) 233
8.5.1 Introduction 233
8.5.2 Preparing the Designed Experiment 236
8.5.3 Carrying Out the Trials 237
8.5.4 Interpreting the Results 238
8.5.5 Study Conclusion 241
8.6 Advantages and Dangers of Randomization 242

Chapter 9 Response Surface Designs 245

9.1 Introduction 247
9.2 Composite Designs 247
9.3 Box-Behnken Designs 249
9.4 Doehlert Designs 250
9.5 Example 12: The Foreman’s Rectification (Example of a
Composite Design) 255
9.5.1 Preparing the Designed Experiment 255
9.5.2 Responses and Objectives of the Study 256
9.5.3 Factors 257
9.5.4 Study Domain 257
9.5.5 Choosing the Design 257
9.6 Experimentation (Factorial Design) 258
9.7 Interpreting the Factorial Design Results 258
viii Contents

9.7.1 Verification of the Linear Model 259

9.7.2 Positions of Star Points 260
9.8 Second-Degree Complementary Design 262
9.9 Interpreting the Results 263
9.9.1 Modeling 263
9.9.2 Graphical Representation of the Results 264
9.9.3 Confirming the Results 268
9.10 Study Conclusion 268
9.11 Example 13: Soft Yogurt (Box-Behnken Design Example) 269
9.11.1 Preparing the Design Experiment 269
9.11.2 Factors 270
9.11.3 Study Domain 271
9.11.4 Responses 271
9.11.5 Choosing the Design 271
9.12 Experimentation 271
9.13 Interpreting the Design 272
9.13.1 Model 272
9.13.2 Interpreting the Results 274
9.14 Study Conclusions 276
9.15 Example 14: Insecticide (Example of a Doehlert Design) 277
9.15.1 Preparing the Designed Experiment 277
9.15.2 Factors and Study Domain 277
9.15.3 Responses 278
9.15.4 The Designed Experiment 278
9.16 Experimentation 278
9.17 Interpreting the Results 279
9.17.1 Calculating the Coefficients 279
9.17.2 Modeling 281
9.17.3 Residual Analysis 282
9.17.4 Graphical Representation of the Results 282
9.18 Study Conclusion 286
 Contents ix

Chapter 10 Mixture Designs 287

10.1 Introduction 288
10.2 Fundamental Constraint of Mixtures 289
10.3 Geometric Representation of a Mixture 289
10.3.1 Two-Component Mixtures 289
10.3.2 Reading a Binary Mixture Plot 291
10.3.3 Reading a Ternary Mixture Plot 293
10.3.4 Four-Component Mixture 294
10.4 Classical Mixture Designs 295
10.4.1 Simplex-Lattice Designs 295
10.4.2 Simplex-Centroid Designs 297
10.4.3 Augmented Simplex-Centroid Designs 298
10.5 Mixture Design Mathematical Models 299
10.5.1 First-Degree Models 299
10.5.2 Second-Degree Models 300
10.5.3 Third-Degree Models 301
10.6 Example 15: Three Polymers 302
10.6.1 Preparing the Designed Experiment 302
10.6.2 Running the Experiment 303
10.6.3 Interpreting the Results 304
10.6.4 Study Conclusion 306

Chapter 11 The Concept of Optimal Designs 307

11.1 Introduction 308
11.2 Hotelling’s Example 308
11.3 Weighing and Experimental Design 311
11.3.1 Standard Method 312
11.3.2 Hotelling’s Method 313
11.4 Optimality 314
11.4.1 Maximum Determinant (D-Optimal) Criterion 315
11.4.2 Computing Optimal Designs 315
11.5 Optimal Designs with a Linear Model 319
11.6 When to Use Optimal Designs 322
11.7 Adaptability of Optimal Designs 323
x Contents

11.8 Example 16: Developing a Fissure Detector 326

11.8.1 Preparing the Designed Experiment 326
3
11.8.2 Using a 2 Factorial with Center Points 327
11.8.3 Using a D-Optimal Design with Repetitions 330
11.9 Example 17: Pharmaceutical Tablets 333
11.9.1 Preparing the Designed Experiment 333
11.9.2 The Design 335
11.9.3 Running the Experiment 336
11.9.4 Analysis of the 28-Run Design 337
11.9.5 Study Conclusion 340
11.10 Example 18: Self-Tanning Cream (Rescuing a Bad
Design) 343
11.10.1 Why a Design Can Be Bad 343
11.10.2 Preparing the Designed Experiment 344
11.10.3 Running the Experiment 345
11.10.4 Interpreting the Results 345
11.10.5 Study Conclusion 349

Chapter 12 Latin Squares and Similar Designs 351

12.1 Introduction 352
12.2 Example 19: Salaries and One Categorical Factor 353
12.2.1 Mathematical Model 353
12.2.2 Global Effect of a Factor 355
12.3 Example 19: Salaries and Two Categorical Factors 356
12.3.1 Preparing the Designed Experiment 356
12.3.2 Mathematical Model 357
12.3.3 Effect of the Factors 359
12.3.4 Analysis of Sums of Squares 361
12.3.5 Analysis of Variance 365
12.4 Latin Squares 365
12.5 Greco-Latin Squares 367
12.6 Youden Squares 368
 Contents xi

12.7 Example 20: Penetrometry Study 370

12.7.1 Preparing the Designed Experiment 370
12.7.2 Running the Experiment 370
12.7.3 Interpreting the Results 371
12.7 Study Conclusion 373

Chapter 13 Summary and Advice 375

13.1 Introduction 376
13.2 Choosing the Experimental Method 376
13.2.1 Posing the Problem Well 376
13.2.2 Preliminary Questions 378
13.2.3 Choosing a Design 381
13.3 Running the Experiment 382
13.4 Interpreting the Results 382
13.4.1 Critical Examination 382
13.4.2 What to Do Next 384
13.5 Progressively Acquiring Knowledge 385
13.6 Recommendations 386
13.7 What Experiments Cannot Do 386

Chapter 14 Using JMP to Design Experiments 389

14.1 Introduction 390
14.2 Designing an Experiment 390
14.2.1 Screening Designer (Classical Designs Only) 391
14.2.2 Custom Designer (Classical and Other
Optimal Designs) 393
14.3 The JMP Data Table 396
14.4 Choosing and Fitting a Model 398
14.4.1 The Screening Designer 401
14.5 Examining and Analyzing Results 402
14.6 Principal Graphics and Analysis Reports 404
14.7 Transferring and Saving Analysis Results 406
xii Contents

Appendix A Example Data Sources 409

Appendix B Comparing Two Independent Means 413
Appendix C Introduction to Matrix Calculations 415
Bibliography 431
Index 439
Preface
This book first appeared in 1988. Since then, the use of designed experiments (DOE) has
developed considerably in the industrial and academic worlds. Many software packages
dedicated to designed experiments have appeared, and these have simplified design
construction, removed difficulties in calculation, and allowed rapid visualization of
graphs that give unique insights into experimental results.

These developments compelled us to update the original Introduction aux Plans

d’Expériences in order to take these rapid changes into account. We also saw an
opportunity to share the concepts with an even broader audience. This is why SAS Press
in the USA and Dunod in France decided to offer a third edition in both English and
French. The first and second editions deal only with full- and fractional-factorial designs.
The third edition adds chapters on response surface designs, mixture designs, D-optimal
designs, and designs for discrete variables. This diversity would have not been possible
without modern software, which easily constructs all of these designs (and more), and
which carries out the most complex calculations easily and swiftly.

The examples and illustrations in this edition are carried out using JMP (pronounced
“jump”) software from SAS Institute Inc. Data tables are available for download from the
SAS Press companion Web site for this book (support.sas.com/goupycreighton).
Additionally, a blog to accompany the French-language version of this text is found at
www.plansdexperiences.com and www.chimiometrie.org, where you can find additional
explanations on our examples in English and French as well as a short French
introduction to JMP. You can contact us using any of these sites. Up-to-date
documentation is also available from www.jmp.com.

Publishing a book is always a group process, and this one is no different. We would like
to thank Michel Genetay, Rémi Longeray, Lalatiana Rakotozafy, and Jacques Vaillé, who
read the original French manuscript and who offered many helpful suggestions and sage
advice. Thanks also to Ann Lehman, Brad Jones, and Marissa Langford, who read early
versions of the English text, and whose comments undoubtedly made this a better book.
Thanks also to Ronite Tubiana and Martine Lemonnier from Dunod and to Stephenie
Joyner, Mary Beth Steinbach, Caroline Brickley, and Candy Farrell from SAS who
helped us in the editing process and who ensured that our manuscript would eventually
become a book.
xiv Preface

We advise all of you who want to deepen your understanding of DOE to augment your
reading by working out the examples with JMP itself. Using JMP deepens understanding
of the analyses and encourages proper interpretation of results. We have tried to furnish
all the tools necessary so that these methods are immediately applicable to your work. All
that remains is for us to wish you “Happy Experimenting!”

Jacques Goupy, Paris (France)

Lee Creighton, Cary (North Carolina, USA)
C h a p t e r 1
Gaining Knowledge with Design of
Experiments

1.1 Introduction 2
1.2 The Process of Knowledge Acquisition 2
1.2.1 Choosing the Experimental Method 5
1.2.2 Analyzing the Results 5
1.2.3 Progressively Acquiring Knowledge 5
1.3 Studying a Phenomenon 5
1.4 Terminology 6
1.4.1 Factor Types 6
1.4.2 Experimental Space 7
1.4.3 Factor Domain 8
1.4.4 Study Domain 9
1.5 Centered and Scaled Variables 11
1.6 Experimental Points 13
1.7 Design of Experiments 14
2 Introduction to Design of Experiments with JMP Examples

1.7.1 Methodology of Designs without Constraints 14

1.7.2 Methodology of Designs with Constraints 15
1.7.3 The Response Surface 17
1.7.4 The a priori Mathematical Model of the Response 18

1.1 Introduction
If you are reading this book, you probably do experiments that you would like to
organize well. Ideally, you want to conduct only those experiments whose results yield
the best possible information and insights.

This book is written to help you. It describes methods and tools that allow you to
consistently design useful experiments, ensuring that you extract the maximum amount
of information from each one. You will be able to efficiently solve problems and make
decisions while gaining experience using JMP software.

We initially look at how experimental design is integrated into the process of knowledge
acquisition. Then, we examine some basic concepts that make it possible to properly
define a study, and how to interpret a study’s results.

1.2 The Process of Knowledge Acquisition

The process of acquiring knowledge can be thought of as answering carefully posed
questions. For example, if a farmer wants to know how a fertilizer influences corn yield,
the following (and certainly more) questions are reasonable to ask:

ƒ Can the field produce ten more bushels of corn per acre if I increase the quantity
of fertilizer?
ƒ How does the amount of rain affect the effectiveness of fertilizer?
ƒ Will the quality of corn remain good if I use a certain fertilizer?
ƒ How much fertilizer should I apply to get the biggest harvest (the most bushels
of corn per acre)?
 Chapter 1: Gaining Knowledge with Design of Experiments 3

These questions frame both the problem and the solution. They also specify what work
should be carried out. It is therefore important to ask questions that are truly
representative of the problem.

Of course, before doing any experiments, it is good practice to verify that the question
hasn’t already been answered. To this end, you can review current literature from a
bibliography, consult subject-matter experts, conduct some theoretical calculations,
simulations, or do anything else that may answer your questions without
experimentation. If, after conducting this research, you find that the question is
answered, then there is no need to conduct experiments. More typically, however, these
investigations cause you to modify your questions or generate new ones. This is the point
when experiments—well-designed ones—are necessary to completely solve the problem.
Unquestionably, this preliminary work is part of the experimenter’s job. However, we do
not focus on these preliminaries in this book.

For questions that don’t have readily available answers, it is necessary to carry out
experiments. How can the experiments be designed so that they accomplish the following
goals?

ƒ Quickly arrive at the best possible results

ƒ Omit unnecessary trials
ƒ Give results with the best possible precision
ƒ Progress without failure
ƒ Establish a model for the studied phenomenon
ƒ Discover the optimal solution
4 Introduction to Design of Experiments with JMP Examples

The three essential aspects of the knowledge acquisition process are shown in Figure 1.1:

ƒ The choice of the experimental method

ƒ The analysis of the results
ƒ The progressive acquisition of knowledge

Figure 1.1 Design of experiments optimizes the three highlighted parts of

the knowledge acquisition process.

Let’s look at these three aspects in detail, remembering that experiments are organized to
facilitate the analysis of the results and to allow the progressive acquisition of
knowledge.
 Chapter 1: Gaining Knowledge with Design of Experiments 5

1.2.1 Choosing the Experimental Method

The experimental method must facilitate the interpretation of the results. However, it
should also minimize the number of runs without sacrificing quality. Using designed
experiments ensures the conditions that give the best possible precision with the smallest
number of runs. Designed experiments give maximum efficiency using the smallest
number of trials, and therefore the minimum cost.

1.2.2 Analyzing the Results

Analyzing the results of experiments is linked to the initial design choice. If the
experiments are well-prepared, the results are easy to interpret, and they are also rich in
information.

Thanks to computers and software, the construction of experimental designs and the
necessary analysis calculations have become simple. These tools also support graphical
representations that illustrate the results spectacularly and increase understanding of the
phenomenon.

1.2.3 Progressively Acquiring Knowledge

An experimenter who undertakes a study obviously does not know at the outset the final
results or what they reveal. Therefore, it is wise to advance gradually, to be able to adjust
the experimental runs based on initial results. Using an initial outline, for example,
makes it possible to direct the tests towards only the interesting aspects of the study and
to avoid dead ends.

An initial batch of experiments leads to preliminary, tentative conclusions. Using these

initial conclusions, the experimenter can carry out a new series of improved tests. Both
series of experiments are used to obtain precise results. In this way, the experimenter
accumulates only the results that are needed, and has the flexibility to stop when the
results are satisfactory.

1.3 Studying a Phenomenon

Studying a phenomenon is often thought of as focusing on a particular measurement: a
car’s gasoline consumption, the wholesale price of a chemical, or the corn yield per acre.
This measurement (consumption, price, or yield) depends on a great number of variables.
6 Introduction to Design of Experiments with JMP Examples

For example, gas consumption is related to the speed of the vehicle, the engine
horsepower, driving style, the direction and force of the wind, the inflation of the tires,
the presence (or not) of a luggage rack, the number of people in the car, the make of car,
and so on. The price of a chemical depends on the quality of the raw materials, the yield
of each manufacturing unit, external specifications, conditions of manufacture, and many
other quantities. Corn yield, too, depends on the quality of the soil, the quantity of
incorporated fertilizer, sun exposure, climate, corn variety, and so on.

Mathematically, we can write the measurement of interest as y, (which we will call the
response) as a function of several variables xi (which we will call factors) as

y = f ( x1 , x2 , x3 ,", xk )
The study of a phenomenon boils down to determining the function f that relates the
response to the factors x1, x2, …, xk.

To look at this approach in more detail, it is necessary to introduce a few special ideas
and also some terminology specific to designed experiments.

1.4 Terminology
This section describes terms related to factors and their representation.

1.4.1 Factor Types

The construction of designs and the interpretation of their results depend largely on the
types of factors involved in the study. Statisticians distinguish among several types of
factors. We discuss four types:

Continuous Factors
Pressure is an example of a continuous factor. For a given interval, any value in
the interval can be chosen. Other examples are wavelength, concentration, or
temperature. Values taken by continuous factors are therefore represented by
continuous numbers.

Discrete Factors
On the other hand, discrete factors can take only particular values. These values
are not necessarily numeric. The color of a product (say, blue, red, or yellow) is
an example of a discrete factor. A discrete factor can take values that are names,
 Chapter 1: Gaining Knowledge with Design of Experiments 7

letters, properties, or numbers. In the latter case, the number is really just a
numeric label, and does not represent a numeric quantity. It is merely a name or
a reference.

Ordinal Factors
Ordinal factors are discrete factors that can be placed in a logical order. For
example, size may be represented as large, medium, or small. Ranks, also, are
ordinal: first, second, third, and fourth.

Boolean Factors
Boolean factors are discrete factors which can take only two levels: high and
low, open or closed, black and white, −1 and 1, and so on.

The border is sometimes fuzzy among these various types of factors. Color (orange, red,
blue, etc.), apparently a discrete factor, can be transformed into an ordinal measurement,
and even a continuous measurement if the concept of wavelength is introduced. A
continuous factor, like speed, can be transformed into an ordinal or discrete factor: rapid
and slow, or speed A and speed B. This possibility is not a disadvantage—it is an
additional flexibility that the experimenter can use when interpreting results. In fact, the
choice sometimes makes it easier to highlight certain aspects of the study.

Changing the variable type is also a means of adapting the answer to the aim of the
study. For example, consider the age of the members of a population. If the study looks
at average age, the variable “age” is regarded as continuous. If, however, the study
examines the number of people reaching a certain age, the variable “age” is an ordinal
variable, since there are several age categories: young people, teenagers, adults, and
seniors. Finally, if we were studying only the proportion of young people younger than
18, the variable “age” is Boolean: younger than 18 and older than 18.

1.4.2 Experimental Space

Understanding experimental design also requires grasping the essential concept of
experimental space of the variables in the study. We now examine this fundamental
concept in detail using continuous factors, since they are the most frequently used.

To graphically illustrate an experimental space, we use a two-dimensional area. This

representation allows an easy extension to multidimensional spaces.

One continuous factor can be represented by a directed and graduated axis. If there is a
second continuous factor, it is represented by a similar axis. This second axis is drawn
orthogonally to the first (i.e., they form a 90° angle). Mathematically, this gives a
Cartesian plane that defines a Euclidean space in two dimensions. This area is called the
8 Introduction to Design of Experiments with JMP Examples

experimental space (Figure 1.2). The experimental space is composed of all the points of
the plane factor 1 × factor 2 where each point represents an experimental trial.

Figure 1.2 Each factor is represented by a graduated and oriented axis. The
factor axes are orthogonal to each other. The space thus defined
is the experimental space.

If there is a third factor, it too is represented by a directed and graduated axis, positioned
perpendicularly to the first two. With four factors or more, the same construction applies,
but it is not possible to represent the space geometrically. A purely mathematical
representation (a hypercube of four dimensions) of the experimental space is necessary.

1.4.3 Factor Domain

The value given to a factor while running an experimental trial is called a level.

Figure 1.3 The domain of variation for speed contains all the speeds
between 45 mph and 80 mph. The low level of the factor is
written as −1 and the high level as +1.
 Chapter 1: Gaining Knowledge with Design of Experiments 9

When we study the effect of a factor, in general, we restrict its variation to be between
two limits. The experimenter defines these two levels according to specifics of the study.
The lower limit is the low level. The upper limit is the high level. For example, to study
the effect of a vehicle’s speed on its gas usage, its speed is allowed to vary between 45
and 70 miles per hour. The speed of 45 mph is the low level and the speed of 70 mph is
the high level. The set containing all the values between the low and the high level that
the factor can take is called the factor’s domain of variation or, more simply, the factor’s
domain.

If there are more factors, each has its own domain. Since the different factors may have
different units of measurement, it useful to have a common representation for all of them.
In design of experiments (DOE), it is common to denote the low levels by –1 and the
high levels by +1. Here we designate the speed of 45 mph as the –1 level and 70 mph as
the +1 level.

The interior of a factor’s domain contains all the values that it can theoretically take.
Two, three, or more levels can therefore be chosen according to the needs of the study.
For example, if the study uses a second-degree (quadratic) model, three or four levels
should be chosen. That is, we should choose three or four different speeds.

1.4.4 Study Domain

In practice, the experimenter chooses a portion of the experimental space to carry out the
study. This special zone of the experimental space is the study domain (Figure 1.4). This
domain is defined by the high and low levels of all the factors and possibly by constraints
among the factors. Let’s suppose that the second factor is a vehicle’s additional weight,
defined as any additional mass aside from that of the vehicle and the driver. The lower
level of this additional weight is 0 lbs; the high level may be, say, 550 lbs. If there are no
constraints, the study domain is represented by all the points where additional weight lies
between 0 and 600 lbs and whose speeds lie between 45 and 70 mph.
10 Introduction to Design of Experiments with JMP Examples

Figure 1.4 The study domain is defined by the union of the domains from
the different factors. Here, there are no constraints.

There may be constraints on the study domain. For example, it might be impossible to
attain a speed of 70 mph with a lot of additional weight. Figure 1.5 illustrates this
possible reduction of the initial study domain.
Random documents with unrelated
content Scribd suggests to you:
back
back
back
back
back
back
back
back
back
back
back
back
back
back
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookname.com