Customer Reviews

Finding Groups in Data: An Introduction to Cluster Analysis

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

 

 

Creating clusters for data is a very old exploratory technique. This text provides a clear up-to-date and modern approach to this multivariate exploratory data analysis technique.

I enjoyed the close comparison and some of the evaluations of the competing clustering methods. Particularly informative were the discussions of the underlying data distribution assumptions. It also was of even more use because the implementation of the algorithms has been accomplished in S-Plus (MathSoft - 12/98->v.4 ). Later chapters look at some of the model-based statistics which are thus made available to the biostatistic community. The book is particularly useful to make understandable the assumptions which can be applied to larger datasets. Such datasets are prevalent in bioinformatics, from microarray and PCR analyses -- such as P. Brown's and Somogyi's work. Tools to implement the algorithms in this book make it valuable since S-Plus is available to the academic community. Tremendously powerful understanding with this pairing of academic treatment and industrial-strength modelling software!!!

This book is well written, and it can be a useful reference for several standard clustering procedures. The software code (on paper only) is very out of date, and essentially useless. The algorithms have been implemented in Splus (and also in R), and since the documentation in Splus is really inadequate, this text should be considered the documentation.

I hope the authors will write a second edition, and include software, examples, and data on a companion disk. The quality of writing is very good, and there are a few things in this book that I cannot find in any other reference.

I am giving one star because it is technically not possible to give no stars. I was looking for theoretical background of cluster analysis. Instead I found a manual for software written when Fortran IV was "high level language" and IBM PC XT was technological marvel. Of course, program is long gone, together with Fortran IV and PC XT. About 70% of the book is verbatim copy of program output, and about 20% is detailed instruction how to prepare input data. The rest are formulas used by the program - just formulas, without any theory and analysis.

Re-publishing classic books has sense - some books will never get old. But not this one. Stay away from this book if you are interested in cluster analysis. I returned mine the same day I got it

For those readers complaining about the source code, they are currently implemented in S-Plus (and probably in the free public R project). The current S-PLUS codes are very user friendly.

If you have the S-Plus manual, you dont need this book for simple routine. Only serious statisticians, who needs in-depth information, would love this book.

Several codes and algorithms are discussed. I especially liked the author comparison of other related algorithms that provide the reader with a short survey of the techniques. Also, several methods for assessing the quality of ther clustering are proposed.

This book describes the algorithms and usage of various software programs written by the authors. Frustratingly, they aren't included and rather too much space is devoted to describing the input and output of the various programs. An analogy might be a book about painting techniques which described the techniques but failed to include any pictures. Where's the CD-ROM? The publishers need to re-think this book. The world has changed since it was first published.