Customer Reviews

Permutation Tests: A Practical Guide to Resampling Methods for Testing Hypotheses (Springer Series in Statistics)

5 star:		(2)
4 star:		(1)
3 star:		(0)
2 star:		(1)
1 star:		(0)

 

 

This is the second edition of a popular text on permutation methods which just came out in February 2000. Dr. Good has been an expert on permutation tests for over 25 years. In the 1980s he was the editor of a journal called Randomization which dealt specifically with the latest developments in permutation methods. He has also contributed to the scholarly research on this subject with a number of useful publications. In addition through his company he has done a great deal of consulting and has written many reviews on statistical software. He brings all of this valuable experience to the table in this book which emphasizes the wide variety of practical applications for this powerful tool. Permutation methods are gaining increasing popularity along with other resampling methods because of the amazing improvement in speed of digital computers over the past 15 years. This is emphasized in the book which is written at an elementary level for practitioners. It also is filled with practical advice and applications from Dr. Good's many years of experience in the pharmaceutical industry. The book has expanded from 228 to 270 pages with additional references and expansion of chapters 12 and 13 which incorporate computing advances over the 6 years since the publication of the first edition.

Good's book is a disappointment. His explanations are often opaque; he manages to make even basic things such as type I/II errors and power seem complicated. The few examples he offers are contrived; there is an almost complete lack of real-world examples. His discussion of software is hopelessly outdated and the code he provides is confined to very simple toy problems that any serious student of his book will have no problems attacking. Items cited in the book are missing in the references, and there is also a fair number of typos. Finally, (my personal pet pief), Good doesn't seem to know the distinction between "alternative" and "alternate" - he alternates between the two without realizing that one is not an alternative for the other... I highly recommend Rosenbaum's Observational Studies instead (at least for randomization inference); it is beautifully written, rigorous, and offers many real-world examples. Good does not even cite it.

David Annnis reviews this book in JASA 472 (December 2005). He begins by quoting the author's preface, where Good says he aims to "replace, rather than supplement, existing graduate level texts on testing hypotheses and decision theory."

The key points Annis makes are

* This edition is considerably expanded, including a chapter on statistical distributions and a "measure theoretic appendix."

* There is a new inclusive bibliography.

* "The chapters stand alone." Topics include one-sample tests, multiple simultaneous tests, sequential procedures, testing categorical data, multivariate procedures, and even testing space-time data.

* "The author does an admirable job presenting alternative resampling methods, most notably the bootstrap ... and classical parametric tests."

Annis characterizes the book as readable, yet also as having mathematical rigor. He concludes that it "garners high marks for its scope and clarity," recommending it to graduate students (for "rigorous treatment of diverse topics and ample exercises") and also to scientists and statistical professionals as a good reference.

3rd much enhanced edition is now available as a graduate text.