Customer Reviews

Wavelet Methods in Statistics with R (Use R!)

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

 

 

This book is an excellent introduction to wavelet techniques in statistics. It takes a very heavily applied approach and sometimes reads like a set of different tutorials on accomplishing certain tasks in R with the author's own wavethresh package. In truth, one can almost think of it as a very detailed manual. This may sound like a criticism but it is in fact one of the book's greatest strengths. There are already plenty of treatments out there on the theory of wavelets and their applications in a variety of fields ranging from the very mathematical to the heavily applied.

However, apart from this book, I have been unable to find any other book that actually deals with the nitty-gritty of implementation using a particular software package. The mathematical details of wavelet analysis are certainly interesting and should be looked into, but it's also nice to have a hugely accessible and detailed cookbook on how to implement techniques quickly and painlessly. Recommended to anyone who is looking to apply wavelets in their field.

This is an interesting book, but is very condensed and suffers (for the non-expert) from a lack of intuitive reasoning on why we are doing what and when. Then there is the R-package itself. The limitations of dyadic sets of data are apparent when one comes, for example, to time-series analysis. Yet, the R-software in "wavethresh" for non-decimated wavelets seems to restrict the user to dyadic data and to full decomposition rather than partial decomposition, and cannot handle "reflection" rather than "periodic" boundaries. (According to Percival and Walden (2000) these relaxations or adjustments are some of the main advantages of non-decimated wavelets). One of the potentially most interesting parts of the book is the section on time-series dealing with locally-stationary processes for which the author is an authority. Yet, the section leaves one a bit bewildered with not much sense as to what has been achieved at the end. Clearly, the book is designed for graduate and post-graduate students with much more than my limited mathematical and statistical skills, but the fact that the software is available on non-proprietary R is to be commended. Nevertheless, it is to be hoped that the book and the software may become more user friendly over time (like for example PC-Give (although proprietary) which made advances in econometric analysis much more readily available to all). I should stress that in writing this review I am handicapped by my limited mathematical abilities and by my restricted understanding of the use of wavelets, but I think that time-series applications of wavelets can be an important tool in an economist's tool-box.