Applied Multivariate Statistical Analysis

People complain that statistics is about memorizing a bunch of formulas and when to use them. I disagree. The real problem is that the formulas almost never apply exactly to the subject at hand - they have to be adjusted to each application.That's why this book is so helpful. It gives huge numbers of results I can use immediately, but also shows me where they came from. That means that I can rephrase the formulas as needed in special computing environments, but still be sure that I'm getting a meaningful answer. The book has a secondary emphasis that I value very highly: checks that the techniques are giving meaningful answers. It's pretty silly, and perhaps dangerous, to apply a technique without knowing how good its results are. This book gives me the checks I need to measure the quality of the results from each technique.No, there's no C code to cut and paste. This is a math book. The math is clear and well-developed, though, and mostly limited to linear algebra. As I type this, I have the book open to "Discrimination and Classification," and I have my protoype C program on the screen. That's how directly applicable the book is. I admit, I haven't appplied every technique in the book. All of the book is equally well-written, though. When I need principal components or basic clustering, this is the book I'll grab first.

First: I must prefix this by saying that I am majoring in the Mathematical and Computer Sciences.This semester I decided to take a class that happened to use this text as its source. I have been extremely pleased with it: the theoretical work is excellent, the proofs are thourough, the exercises are both good and cover a broad variety of difficulties, and the tables on the CD provide excellent experience in analyzing real world data.A couple of things to keep in mind before you purchase this book, however:1) A good background in linear algebra and basic statistics is highly recommended and virtually necessary to interpret this book. Remembering the knowledge gleaned from "Sequences and Series" (often taught in Calculus II) will also prove useful. The text is good, but it is often nontrivial.2) Some kind of software that does multivariate analysis (and if nothing else, will find eigenvalues and orthonormal eigenvectors) is necessary to get the most out of this book. The software package SAS is touched on in the book, but by no means is given a comprehensive review. However, the data files on the CD-ROM should be loadable by any competant software package, so use the one you are most comfortable with.If not overly familiar with any of them, I can recommend S, SPlus, and "GNU's S" (also known as "R") for their power and flexability to work with the data presented in the book.All and all I found this to be an excellent book, definantly worthwhile if you want or need to know how to do multivariate analysis.

There have been many good theoretical texts on multivariate analysis including Anderson, Eaton and Gnandesikan. Tabachnick has written a popular applied text for the social sciences. Yet for many years this has been considered the best applied text. That is because the authors understand the theory and know how to balance it with applications. They also are excellent writers.

This is a very good, very understandable book. It could (maybe should) be used as a first course in multivariate techniques. Concepts very well explained, it needs no more than a good basic course in statistics to be fully absorbed. It also includes enough matrix algebra to be self-contained.

I followed a Multivariate Statistical Analysis Class at Business School and found this to the the best book on the subject available at the library BY FAR. One has to be at ease with Matrix Algebra though.