Customer Reviews

Principal Components Analysis (Quantitative Applications in the Social Sciences)

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

 

 

I am a big fan of this little "green book" statistical series. Thanks to it, I already taught myself Logit Regression, Cluster Analysis, Discriminant Analysis, Factor Analysis, and Correspondence Analysis. Most of these were excellent; "Principal Component Analysis" (PCA) was good.

The reasons I don't consider it excellent like some of the others are: First, the terminology is kind of dated and confusing. The author talks about of Latent Roots and Latent vectors when the more common names nowadays are Eigenvalues and Eigenvectors. Also, the author mentions in the introduction, he will explain most concepts without relying on Matrix Algebra. Yet, he does to a great extent. If you are not familiar with Matrix Algebra, you will be forced to learn it to better understand this book. Finally, the author gives you many formulas that are sometimes difficult to understand, especially when he rarely fleshes out the related calculations in a concrete example.

Despite the negative comments mentioned above, the book has an equal or greater number of strong points too. Let's face it PCA is complicated. There is no way to make it appear really simple and easy to understand. This book is the kind you have to read, underline, work through examples, and review again. I suspect any book on PCA would be similar.

In view of the above, the author takes interesting examples out of the social science. He develops a strong foundation in PCA. He also does a good job of showing how PCA is at the foundation of many other multivariate analysis methods.

This green book series has allowed me to hang in there and keep up within an intense quantitative group of a major financial institution on the West Coast. Without them, I would have been left behind. For the record, I am an MBA type and not a quant type. So, if I can understand these books, so can you. I recommend this one book if you need to understand PCA. Just accept upfront, it is not going to be easy reading. But, it does the job of explaining PCA.

I have used PCoA on a few occasions and was hoping this book would explain how the method actually works. This was not the case. The book assumes more knowledge on matrix algebra than I have. I read through it but learned very little that I didn't already know, and what I wanted to know wasn't there. The examples are marinally useful.

I found it interesting to be given the option to "upgrade" this item after I purchased it. I promptly did so, to test out this feature, and was greeted with the "search inside this book" functionality, but with full access to the book. This let me get a jump start reading though the book.

This book is very clear for an academic paper and provided a good jumping off point to review my rusty linear algebra. The technique it describes is great for distilling data with high dimensionality and low correlation (a tag cloud for instance) into a smaller set of highly correlated variables (such as could be mapped to a plane for a visual representation).

Like most reference books, initially I skimmed through this and now have it close at hand to aid in the project(s) that inspired it's purchase.