Date: Mar 27, 2011 9:45 AM
Author: Tikkuhirvi Tietavainen
Subject: Procrustes factor analysis

Hello All,

I'm trying to find a way how to use Factor analysis to improve the separation of two data sets that are measured from the same patients before and after a procedure. The sets both include 100 observations (patients) and seven variables that have two dimensions or factors. If I plot the two data sets first variable against second variable (without factor analysis), I can see a difference between the data sets. But how can I use factor analysis and the two factors (factor scores?) to compare the two data sets?

I was advised to use something called the procrustes rotation to ensure that the two data sets have the same rotation and can be compared. I have never used it, but here's what I did:

A1;%First data set, before procedure.
A2;%Second data set, after procedure.

[lambda1,psi1,T1,stats1,F1] = factoran(A1,2);%Default is varimax rotation.
[lambda2,psi2,T2,stats2,F2] =factoran(A2,2,'rotate','procrustes','target',lambda1);
figure;plot(F1(:,1),F1(:,2),'b.',F2(:,1),F2(:,2),'r+')

It would seem that the distributions are as similar as possible. What am I missing here? Am I supposed to be comparing something else than the factor scores F1 and F2?

Thank you,
Aino