Customer Reviews

Practical Nonparametric Statistics (Wiley Series in Probability and Statistics)

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

 

 

Since I now have a copy of the third edition I can now say that many of my comments on the second edition still hold. The book is authoritative, clearly written and very much applications oriented. Conover has done a good job of updating it with recent developments. He provides a nice introductory treatment of bootstrap among other things.

Standard statistics make assumptions about how the data are distributed, then give results based on the assumed distribution. Two big problems are that the distribution buried in the analysis may not be the right one, and that the assumption might not even be visible in the analysis. "Nonparametric statistics" (NPS) make no assumptions about the distribution. They work no matter how the data are distributed. Even better, they sometimes work to determine whether the standard techniques have any hope of giving answers.

For the practitioner, this book is the broadest catalog I know of how-to for NPS: when each analysis applies and how to apply it. Even more, it gives insight into how some of the tests work. That gives the reader a better chance to understand each technique's strengths, weaknesses, and applicability. For the student, including self-taught, it's a clear and well-organized textbook. The exercises are varied and generally meaningful, and half have answers (though little discussion of how the answers were derived).

I wish the book gave more background, including how some of the distributions are derived. Most times, seeing more of the derivation gives me more confidence in using an analysis. Face it, almost every real-life situation needs to be bashed a bit to fit the format expected by a test. Knowing more of the background gives me more assurance that my machinations don't break any important assumptions. Still, it's the author's choice to emphasize practice over theory and I have to respect that.

More seriously, I would like to see the bootstrapping section enlarged. Many modern applications, particularly in biology, deal with data so complex that they define analysis or even real understanding. Bootstrapping is just one of many randomization and resampling techniques used for such data. More discussion on the design and analysis of resampling techniques would have been very useful.

The book meets its goals, though, and does so admirably. I'm not a stat specialist, but this is the book I'll recommend for heavy users who want a little more than rote recitation of analytic techniques.

This is a very impressive book. All concepts are introduced in an elementary fashion, with derivation following only after an example of the technique. The explanations are lucid and the extensive lists of references very helpful. I would heartily recommend this book to anyone interested in robust estimators and nonparametric methods.

At first glance this textbook appears to be a well-written and thorough introduction to nonparametric statistics. The range of research studies presented and the list of references is far more comprehensive than what is found in other familiar texts, such as Siegel and Castellan. Upon further inspection, however, one discovers large gaps in Conover's treatment. A vast number of research studies that have contributed to current knowledge of nonparametric methods are missing. For example, I have been doing Monte Carlo studies the area for about twenty years, and some fifteen or twenty of the studies that I consult frequently in connection with my own work, and which I am sure many other investigators find necessary to be familiar with, are nowhere cited. This major oversight cries out for explanation.
A feature of the text that is attractive upon first reading is the discussion of the so-called rank transformation and its relation to familiar nonparametric tests. The fact that certain nonparametric methods, such as the Wilcoxon-Mann-Whitney test, produce the same results as parametric methods like the Student t test performed on ranks replacing scores, is valuable information for researchers as well as theorists. Conover deserves credit for introducing these interesting and important findings at a time when they were either overlooked or swept under the rug by other authors of widely-used introductory textbooks.
Unfortunately, however, Conover does not delve into the implications of the rank transformation for the classification of scales of measurement into nominal, ordinal, interval, and ratio scales and its significance for practical use of statistical methods. The above mentioned equivalence raises serious problems for this venerable classification. Curiously, Conover adheres to the old scheme. Further and somewhat more venturesome discussion could have added substantially to the worth of the text. It remains for other investigators to explore more fully the inevitable implications of the rank transformation for statistics education as well as statistical theory and practice.

This book is a great reference for the scientist needing to know how best to interpret their results. It has clear explanations for when to use particular tests as well as how to use the tests when multiple observers are rating the same event. Analogous parametric tests are mentioned along with comparisons (A.R.E.) of their relative efficiency.

This was much better info than I got from the statisticians at my workplace - of course, the statisticians want to keep my business, not tell me how to do it!

I am involved with environmental statistics software development. When I need to look up something related to nonparametric statistics I find Conover is the first place I turn to. It is written well, does not require that the user have advanced mathematical skills, and contains an extensive list of references. My use is limited to the 2nd edition, but I would guess the 3rd edition is only better.

I purchased this book for work since many of the performance metrics/data we study at an electric utility do not follow a normal distribution.

When reading the term "Practical" in the book title I had hoped for the following:
1. What each technique is and is used for;
2. Some real world examples for application;
3. Assumptions and conditions;
4. How to apply the technique;
5. Pitfalls to avoid in its application.

Instead the book is organized in a very academic manner where theory is presented without practical or reasonably accessible application examples.

In summary, the book is okay, but the term "Practical" in the book title is highly misleading. A more "Practical" source for some simple and useful (although limited intro to) nonparametrics is contained in Ramsey's "Intermediate Statistics for Dummies". I was hoping Conover's book would help me pick up additional practical nonparametric techniques, but I've only found tedious algorithms. I don't mind the math, but a good book should not overlook intuition.

I ended up only browsing through this book and ended up needing to return it and the seller was absolutely horrible about returns.
I think in the end I spent $15 returning it to them, and I never saw a cent of my money from the return.