How to compare the output of PCA analysis?

Hello to everyone.

I have this problem: I run a PCA analysis using the princomp function on two 8x3 matrices. Then, in order to find out whether they behave similarly or not I would like to perform a comparison of such vectors.

I found two ways: the first was proposed by Krzanowski (1979) (see ikaros.teipir.gr/era/era2/fullpap/B26.doc on page 4), but the problem is that applying the equation 3 on page 4 I always, i.e. regardless of the input data, that the similarity factor (called Sim is always maximum, in my case 3: how can I use it then?) and the second by Yang and Shahabi (2004) that I still have to try.

What is your suggestion?

Thank you.

Answer 1

[James Cai]

http://freesourcecode.net/matlabprojects/57085/pca-similarity-factor-in-matlab#.WsZ4QYjwaUk

[Walter Roberson]

Jeremy Milne comments:

This response is no help

[Walter Roberson]

I disagree Jeremy Milne: it is a link to source code that implements the kind of similarity measure that the poster appears to have been asking for.