Customer Reviews

Applying Contemporary Statistical Techniques

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

 

 

I am a professor of Linguistics with an interest in applying contemporary statistics to my field. People in my field have rarely heard of many of the modern methods mentioned here, such as bootstrapping, Winsorizing, 20% means trimming, etc. Wilcox does a great job of making this information accessible and convincing the reader of the need for more modern methods of statistics. What I especially like is that these methods are much more practical for applied research than traditional theoretical statistics based on assumptions of normality and homogeneity of variance (which we rarely find in real data!).The best feature of this book, however, is that Wilcox has written scripts for S-PLUS and R so that you can actually APPLY what he recommends in an easy-to-implement way. The text tells you where to find the commands, and then uses them to perform the necessary tests. When I had a problem getting one of the commands to work correctly, I emailed Wilcox and he promptly gave me a revised version of it that worked. I heartily recommend this book to anyone trying to get up to speed on more modern methods of statistics. I think it would be an excellent book for a research methods course in any social science field where 'classic' statistics are regularly done.

And I do note that I have read this actual book! Wilcox does have an entire chapter on nonparametric tests in this book.

This book accomplishes something that has been needed for a long time: A critical examination of the shortcomings of many of the statistical methods that have been widely used in psychology, education, and the social sciences, combined with a presentation of more modern and lesser known methods that overcome these shortcomings. It addresses the assumptions made in traditional statistical tests based on normal-curve theory and points out in a clear and convincing way how failure of these assumptions can lead to trouble. The book conveys a marvelous sense of ongoing discovery--that there is progress in statistics, that methods of the past are not always the best, and that newer and better methods are continually evolving. This is a far cry from the message in the overabundance of elementary textbooks used in introductory courses that all too often present statistical methods as if everything is known and finalized.
The introductory chapters of the book describe elementary concepts in statistics "with a difference," by focusing on properties of various methods that are typically ignored or glossed over in textbooks. This prepares the way for understanding of the difficulties to be examined later. The description of the historical origins of familiar methods is admirable.
The main flaw in the book is lack of treatment of nonparametric methods and their relation to the tests based on parametric assumptions. In the past, one of the reactions of applied statisticians and researchers to failure of assumptions has been a switch to nonparametric methods. However, these tests have their own assumptions and problems, and a thoroughgoing treatment of these problems and their relation to the ones mentioned above also is badly needed. Unfortunately, nonparametric tests are mentioned only briefly toward the end of the book. But despite this gap in the argument, I still give the book five stars because its other features are excellent. If run-of-the-mill statistics texts deserve two or three stars, then a five-star scale is not enough: This book deserves nine or ten.
(Edited note: Something went wrong here! I intended this to be a review of "Fundamentals of Modern Statistical Methods: Substantially Improving Power and Accuracy," not "Applying Contemporary Statistical Techniques," which I have not read!)