Customer Reviews

All of Nonparametric Statistics (Springer Texts in Statistics)

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

 

 

This is a very nice and concisely written account of the field of nonparametrics. It is very bold for an author to use "all" in the title and I am sure that Wasserman does this with tongue in cheek. He has been even bolder with his other book called "All of Statistics." However, I think that it can safely be said that the essentials are covered along with many interesting topic that are not covered in a traditional nonparametrics text such as Conover's book.

The author's intentions are best expressed in this excerpt from the publisher's description. "The goal of this text is to provide the reader with a single book where they can find a brief account of many, modern topics in nonparametric inference. The book is aimed at Master's level or Ph.D. level students in statistics, computer science, and engineering. It is also suitable for researchers who want to get up to speed quickly on modern nonparametric methods.

This text covers a wide range of topics including: the bootstrap, the nonparametric delta method, nonparametric regression, density estimation, orthogonal function methods, minimax estimation, nonparametric confidence sets, and wavelets. The book has a mixture of methods and theory."

So we see that by "all" the author means a brief description of most topics in nonparametrics. Nonparametric density estimation, bootstrap, nonparametric regression, and wavelets are not covered in traditional nonparametric books and the topics are so involved that each has been treated in books solely dedicated to that topic. I and at least five other authors have written books dedicated to the bootstrap and other resampling methods. Hardle and others have written books on nonparametic regression and Silverman and others have published books on univariate nonparametric density estimation. Scott has done one on multivariate nonparametric density estimation. The topics are impressive and the coverage is good but obviously not thorough. To thoroughly treat all of these topics would take several thousands of pages!

I've read about 75% of this book, quitting near the end where the material was of less interest to me. I found this book to be largely focused the theory of nonparametrics with examples. The most important aspect of the book relative to others is that it teaches where each of the methods comes from in a rigorous fashion, i.e. it provides a wonderful base in theory of modern nonparametric methods.

On the other hand, those who want to actually apply the methods will be at a bit of a loss unless they are comfortable with scavenging CRAN for the proper functions/packages for the methods (and are successful at picking good packages). Or if you want to reinvent the wheel, that also works. I was disappointed in the practical aspect of the book in applying the methods. In the book preface, it says "data sets and code may be found at [web address]"... there is data but no code there (as of 8/17/09). At the very least, suggest package/functions in R (or some software)! Just the package/function names would be enough to get running on applying the methods and is certainly not too much to ask.

One other complaint I have with this book is that there is too much time spent on basic concepts while too little on the more complex concepts (as is often the case). The complex material is within reach if I reread it many times, however, more pages spent on these materials with corresponding less on the basics would have been preferable and resulted in a superior text.

For practitioners, I don't think this book will meet expectations. Practitioners open to other methods might instead check out The Elements of Statistical Learning.

That said, I am pleased with this book for its theory (it made me greatly appreciate jackknifing), and I would recommend this book to anyone interested in developing nonparametrics or who wants to understand some intricate nonparametric methods. It is well-written in what I anticipate was its aim: provide a strong theory base for modern nonparametric methods. For those with a milder interest, this book probably won't meet expectations.

My first thought upon receiving this slim volume went something along the lines of "Oh my, all of non-parametric statistics is much, much smaller than I expected". My other impressions are somewhat vague since (as the title of this review implies), this is not exactly a flip-through book for light reading. The content is dense ... like 100-year old ebony wood.

And after carefully examining the table of contents, I also noticed that the text has precious little to say about methods based on rank. So this is more or less a book on "modern methods". It's a very interesting selection of material - although I will probably never be able to read it.