I have a copy of the original version of the book that was published by Wiley. Often books like this one are added to the Wiley classic series and are reprinted in paperback. Although Wiley did not choose to do this, I am glad to see that SIAM decided to add it to its applied mathematics classic series.

Ingram Olkin was one of my professor's when I was a graduate student at Stanford. I met Milton Sobel through my colleague the late Ram Uppuluri, who I worked with at Oak Ridge.

Sobel was one of the originators of the ranking and selection approach to comparison of three or more populations. His work with Jack Kiefer and Robert Bechhofer was published in a text in the late 1960s. The basic idea was that many problems come up where several populations are being compared. An example would be the comparison of a placebo with two or more competitive drugs. The traditional analysis of variance tests equality of means versus a general alternative that at least one of the means is different from the rest. When the null hypothesis is rejected, the answer is that at least one mean is probably different, but it does not answer the basic questions. Which one or group of means is different and how large is the difference? Contrasts are then used and the method of multiple comparisons is used to identify populations with statistically significantly different means from rest.

The ranking and selection approach is different. It asks given the data on the distributions of these K populations, what is the probability that we can correctly rank them from worst to best? What is the probability that we can choose the best population (perhaps the one with the largest population mean)or at least the best M out of the K populations?

The idea is simple, the methodology is well developed but the approach has to this day not caught on as a basic component of statistical training and is not being applied in practice. This was the state of affairs that motivated Sobel to initiate the project of writing this book. The idea was that if the theory could be presented in an elementary way it might be better appreciated and more often used. Greater exposure of the methods could also stimulate further research.

The book provides the clear exposition. The fact that the techniques have not caught on remains a mystery. Milton Sobel discusses this issue in an interview that was just published in the May 2000 issue of the journal "Statistical Science".

The three authors each provide special talents that make this an excellent book. Olkin is thorough in his research and this is reflected in the completeness of the references (for that time). Gibbons is an excellent writer who must have had a strong influence on the clarity of exposition. Sobel is one of the founding fathers of the methodology who provides the knowledge of the theory and applications.

I will not duplicate what is in Vickie Kearn's review. She gives an accurate description of the book and its value. In my view the authors have successfully demonstrated the value of ranking and selection yet it has not caught on. Partially this is because everyone knows the standard ANOVA approach. This is what they are trained and it has consequently become their natural approach to such problems. It is unfortunate that many well trained statisticians do not even know of the existence of this large body of literature on ranking and selection. Sobel has noted in his experience that many younger statisticians rediscover the ranking and selection ideas. Until it becomes a part of the standard courses this will continue.

Another factor is software. These days procedures get used in practice only if they are included in some standard commercial package. The statisticians that invented ranking and selection did not see to it that it was incorporated in SAS or some other important package.

Another factor may be that the methodology might provide an answer like "the probability of correctly selecting the best population is 0.15" and this may not seem too spectacular an answer to the investigator.

Selecting and Ordering Populations: A New Statistical Methodology Jean Dickinson Gibbons, Ingram Olkin, and Milton Sobel Classics in Applied Mathematics 26

This SIAM Classics edition is an unabridged, corrected republication of the work first published in 1977. It provides a compendium of applied aspects of ordering and selection procedures and includes tables that permit the practitioner to carry out the experiment and draw statistically justified conclusions. These tables are not readily available in other texts. Although more than 1000 papers and several books on the general theory of ranking and selection have been published since this book first appeared, the methodology is presented in a more elementary fashion with numerous examples to help the reader apply it to a specific problem.

There is a dichotomy in modern statistics that distinguishes between analyses done before an experiment is completed and those done afterward. Ranking and selection methods are useful in both of these categories. The authors provide an alternative to the overused "testing the null hypothesis" when what the practitioner really needs is a method of ranking k given populations, selecting the t best populations, or some similar goal. That need and purpose is as important today as when the subject was first developed nearly 50 years ago.

Applied statisticians as well as researchers who use the basic methods of statistical analysis (psychologists, engineers, biologists, management scientists, etc.) will find this book a valuable reference. Readers should be familiar with standard first-year statistics; no knowledge of calculus is necessary.

Jean Dickinson Gibbons is the Thomas D. Russell Professor Emerita of Applied Statistics at the University of Alabama. She has published numerous articles and books on nonparametric statistics, both theoretical and applied. She has been a Fellow of the American Statistical Association (ASA) since 1972 and was elected to their Board of Directors for four different terms. Ingram Olkin is Professor of Statistics and Education at Stanford University in California and is a member of the SIAM Classics in Applied Mathematics editorial board. He has been a Fellow of the ASA since 1962. He was awarded the Samuel L. Wilks Memorial Medal by the ASA in 1991 and the Elizabeth L. Scott Award by the Committee of Presidents of Statistical Societies (COPSS) in 1998. Milton Sobel is Professor Emeritus of Statistics and Applied Probability at the University of California, Santa Barbara. He is a Fellow of both the ASA and the Institute of Mathematical Statistics (IMS) and was elected a member of the International Statistical Institute (ISI). He has authored several books, published well over 100 journal articles and reports, collaborated with many researchers, and served as advisor to numerous Ph.D. students in statistics.

Contents Chapter 1: The Philosophy of Selecting and Ordering Populations; Chapter 2: Selecting the One Best Population for Normal Distributions with Common Known Variance; Chapter 3: Selecting the One Best Population for Other Normal Distribution Models; Chapter 4: Selecting the One Best Population Bionomial (or Bernoulli) Distributions; Chapter 5: Selecting the One Normal Population with the Smallest Variance; Chapter 6: Selecting the One Best Category for the Multinomial Distribution; Chapter 7: Nonparametric Selection Procedures; Chapter 8: Selection Procedures for a Design with Paired Comparisons; Chapter 9: Selecting the Normal Population with the Best Regression Value; Chapter 10: Selecting Normal Populations Better than a Control; Chapter 11: Selecting the t Best Out of k Populations; Chapter 12: Complete Ordering of k Populations; Chapter 13: Subset Selection (or Elimination) Procedures; Chapter 14: Selecting the Best Gamma Population; Chapter 15: Selection Procedures for Multivariate Normal Distributions; Appendix A: Tables for Normal Means Selection Problems; Appendix B: Figures for Normal Means Selection Problems; Appendix C: Table of the Cumulative Standard Normal Distribution F(z); Appendix D: Table of Critical Values for the Chi-Square Distribution; Appendix E: Tables for Binomial Selection Problems; Appendix F: Figures for Binomial Selection Problems; Appendix G: Tables for Normal Variances Selection Problems; Appendix H: Tables for Multinomial Selection Problems; Appendix I: Curtailment Tables for the Multinomial Selection Problem; Appendix J: Tables of the Incomplete Beta Function; Appendix K: Tables for Nonparametric Selection Problems; Appendix L: Tables for Paired-Comparison Selection Problems; Appendix M: Tables for Selecting from k Normal Populations Those Better Than a Control ; Appendix N: Tables for Selecting the t Best Normal Populations; Appendix O: Table of Critical Values of Fisher's F Distribution; Appendix P: Tables for Complete Ordering Problems; Appendix Q: Tables for Subset Selection Problems; Appendix R: Tables for Gamma Distribution Problems; Appendix S: Tables for Multivariate Selection Problems; Appendix T: Excerpt of Table of Random Numbers; Appendix U: Table of Squares and Square Roots; Bibliography; References for Applications; Index for Data and Examples; Name Index; Subject Index.

June, 1999 / xxvi + 569 pages / Softcover / ISBN 0-89871-439-7

This book is a comprehensive review of important concepts supporting the design of experiments methodology. It covers elements ranging from basic data collection techniques to interpretation of ANOVA tables. In addition, the book provides explanations of the integral role DOE plays in the six sigma and robustness concepts, which are so prevalent within R&D and manufacturing organizations around the world.

Dr. Stamatis not only focuses on the statistical concepts supporting the DOE methodology, but he also emphasizes their rationales, applications and interpretations. Basic statistical ideas are clearly presented, so that the non-statistician can understand, as well as apply the concepts. Pertinent illustrations are provided for every relevant concept, most with numerical examples, and many with cases examples demonstrated through output from various commercially available software.

This book is a must for every Engineer and Scientist who ever plans to use design of experiments in his or her work. The book is well suited for both the self-learner and the experienced researcher. The materials are presented in an interesting and down to earth, readable text. The power of the methodology is obvious to the reader by the end of every chapter. In brief, this book is an excellent presentation of a potentially difficult and dry subject. I highly recommend this book!

Dr. Stamatis covers from Traditional Experimental Design to Taguchi Design in this one source for understanding all the hype around advanced statistics for industrial applications. He even clearly explains the roll of Robust Design, which may be the only true way to achieve six sigma in the future.

I highly recommend this book for anyone planning to use statistics in a business setting.

SPC is one of the most powerful tools available to any organization or workgroup that wants to implement continuous improvement. Unfortunately, it is not widely used outside of manufacturing or companies that are committed to quality. One of the reasons is that is perceived to be difficult to learn. This wonderful book changes that by introducing statistical process controls in a clear, gentle manner.

The book is divided into modules, each of which builds upon the preceding one, and can be used as a training text or as a self-study guide. The first module covers the basics: causes of variation, tools (historgrams, control charts, variable and attribute charts).

Modules 3 and 4 go deeper into the tools, explaining why you would use them, how to use them and how to interpret them. This is the heart of the book.

Machine and process capability, the subject of module 5, can be applied beyond the shop floor. For example, I work as an information technology consultant and was able to apply the knowledge from this module to project estimation and controls, service level measurement and quality assurance processes. This information is also applicable to other areas and will be useful to anyone who works at a company registered as ISO-9000.

Module 5 covers all of the common quality problem-solving tools ranging from brainstorming to scatter diagrams. IT consultants and practitioners will find the sections on cause and effect diagrams and Pareto analysis useful for process improvement for defect identification and removal, and other related objectives.

Elements of a TQM system covered in module 8 may have been better placed in module 1, but it is thorough and a good starting point for anyone who is new to quality.

This book finishes with a module that provides the answers and solutions to practice problems from the preceding modules, which underscores its value as a class test or self-study guide.

I recommend this book to associates who either have never heard of SPC (and there are a lot of them) or think it is beyond their ability to grasp. It is impossible to have a viable, effective program of continuous improvement without SPC. The authors have done a remarkable job of writing a book that lives up to its title by simplifying SPC. As such they have made an important contribution to quality by making this effective tool available to anyone who will take the time to read the book and apply what they learn.

SPC Simplified came along at just the right time. I needed help developing statistical analysis that I could interpret to upper management. I used this text to assist me in constructing my first run charts, variance analysis, error analysis, root cause analysis, brainstorming session and my first Process Cause and Effect Analysis. Great job. A good buy for anyone in Performance Improvement.

This book is now being used at the university I attend for a Data Analysis Course. The examples provided were used as projects for out class. I learned a lot about real world problems that can be solved and modeled using statistics.

I recommend the software MiniTAB to accompany this book.

I especially liked the chapter on HIV and mortality (very interesting results)!

I took my introductory mathematical statistics course from Brad Efron and Charles Stein at Stanford. It gave me a strong foundation and understanding of decision theory, hypothesis testing and the power of exponential families of distributions. Although I understood the Neyman - Pearson theory of hypothesis testing and confidence intervals, I was not completely comfortable applying what I had learn when confronted with real problems. Students even at the best universities in the US still need consulting experience before they are comfortable with the tools they learn.

That is why this book so intrigued me. It approaches the theory and methods from applications first. Each chapter poses an interesting real problem and then progresses to a solution, introducing only the necessary tools. Many important statistical tools and topics are covered this way. It is particularly good for students interested in biostatistics as many of the applications fall in that area. In fact the subject of maternal smoking and infant health is treated in Chapter 1 and revisited in Chapter 10 with additional data to consider.

It is designed for an advanced undergraduate course for statistics majors and has been successfully implemented by the authors at Berkeley. I think this can really work and make mathematical statistics interesting. However, the student should not think that he or she can come out of this class and jump right into consulting. The biggest problem in consulting is working with the client to pose a well formulated problems that address their questions. In this text the authors have already done that long and difficult task for us and we are left in a position to learn the subject matter and the statistical tools needed for the solution.

The second edition came out in 2001 and although my previous review is listed here it refers to the first edition. Professor Reiss sent me a copy of the second edition and I browsed through it to see what has changed. Except for the words "Second Edition" in the upper left corner of the cover page the cover is the same as the original. A closer look through the text shows that there are substantial changes. The original text was 316 pages long with 35 references and a CD Rom on the back page. The new edition is 443 pages with 51 references and a CD Rom on the back page (this is version 3.0 of Xtremes).

The preface to the second edition tells you precisely what is added. There are 8 new contributing authors who are Stuart Coles , Jurg Husler, Daniel Dietrich, Dietnar Pfeifer, Humberto Vaquera, Jose Villasenor, Pieter van Gelder and Dan Lungu and apparently the two main authors will encourage more contributors for a third edition. The authors are very much interested in demonstrating applications of extreme value theory using their Xtremes software and generously invite others to join in.

The structure and theme of the book has not changed. Section I on modeling and analysis has replaced the section on robust statistics with a section called heavy and fat-tailed distributions. The sections are slightly longer in the second edition. Chapter 2 has an additional section called the auto-tail-dependence function.

Part II on inference for parametric models includes a whole new chapter on Poisson Processes (Chapter 7). In Part III on multivariate methods, Chapter 9 on multivariate maxima includes a new section on the Gumbel-McFadden Model and Chapter 10 a new section on bivariate peaks over a threshold.

Part IV on topics in Hydrology, Insurance and Finance is totally revised and consists of Chapters 11-14 in place of the original Chapters 9-11. The old Chapters 9 and 10 are now Chapters 12 and 13 respectively.

In Part V there are again five case studies but they are totally new ones with the new authors that are acknowledged in the preface.

In the appendix they have replaced the description of the XPL programming language with the StatPascal language.

The software XTREMES was introduced by Falk, Husler and Reiss in their 1994 book "Laws of Small Numbers: Extremes and Rare Events". That book was mainly theoretical and the software was in an MS-DOS version for PCs. This text was published in 1997. For this text they supply a Windows (3.1 , 95, NT) version on a CD ROM.

In this book the emphasis is on applications in insurance, finance, hydrology and other fields. The first 10 chapters develop the theory and teach the use of XTREMES presenting dialog boxes and descriptions. The text is divided into 5 parts. Part I deals with modeling and data analysis, part II covers statistical infrence for parametric models, in part III elements of multivariate analysis are introduced, part IV emphasizes the application areas and part V is a collection of case studies using XTREMES. There are five case studies. One presented by Reiss but other presented by notable researchers including Tai Hsing, Jurg Husler, Ana Ferreira, Edgar Kaufmann and Cornelia Hillgartner. The appendices provide additional details on XTREMES. This is a very unique text that is valuable to anyone interested in doing research or applications of extreme value theory. Includes coverage of the parametric bootstrap.

Mason, Gunst and Hess have done a great deal of consulting in industry which has provided them with an understanding of the design and analysis of linear and nonlinear models. In fact, I have had some direct experience with Bob Mason from Southwest Research Institute. When I was working at a medical device company owned by St. Jude Medical Inc., Southwest Research Institute had been contracted to analyze data from one of our animal experiments for defibrillation thresholds. As the new statistician in the company I needed to get acquainted with their work. The book illustrates many designs using real world problems. The first 5 chapters (88 pages) cover elementary statistical concepts and methods. The next 6 chapters (145 pages) cover designs. Most standard designs are covered including full factorials, balanced incomplete blocks, latin squares, fractional factorials and nested and crossover designs. There is also a chapter on response surface designs including rotatable designs, central composite designs and Box-Behnken designs. The rest of the book, some 368 pages, deals with statistical analysis of data from designed experiments. It makes a very good reference source.

The only disadvantage of it is that there have been many advances in the design of experiments since 1989 when the book was published. The topic of robust parameter design is not covered because much of the development occurred after 1989. Hints of the topic and mention of the Taguchi approach appear only on pages 108-110. To learn much more about the recent developments in the design of experiments see Hamada and Wu (2000) "Experiments". My review of Hamada and Wu can be found on Amazon.

Anova,multi stage charterlisti

During my Ph.D.,I have found this book very useful. It covers a lot of new results in econometrics. I think it is not only a reference book for graduate students but also a guide for academic researchers.

I think this book is a good reference point for graduate students and academicians alike. It provides a very nice review of testing issues as well as empirical bayesian methods.

Snedecor first wrote this excellent elementary applied text while at Iowa State (late 1940s or early 1950s). When Bill Cochran arrived in Iowa he helped out with the revision. I was very popular and was revised by Cochran many times even after Snedecor died. Well written and often used in elementary courses this book is also a good reference source for statistical methods.

This book does not require calculus but is rigorous. The examples are real life agricultural data and yet the book is fascinating to read. If I had to have one book on practical statistics this would be it.

This is the best book in statistical quality control. A very good start for undergraduates. However, problems are not suitable for mathematicians because they are very very simple. The logic behind quality control is best presented in this book. I wish the author can increase the number of problems in this book by adding some difficult excercises.

Well written book within 500 pages. Dr. Ryan has done a great job in presenting the material without much mathematical details. A good reference for any one from industry and can be good text for University.