BookmarkSubscribe
Choose Language Hide Translation Bar
RonB
Community Trekker

PCA and T-Square T^2

Multivariate Control Chart (from the control chart menu) shows results for Principal Components on Covariances. Doing Principal Components (from the multivariate methodes menu), also on Correlations gives the same eigenvalues and cumulative percentage. Both are scaled.

 

However, the eigenvectors and pc formulas (those that are save to the data table) are not the same.

 

Am I missing something?

 

Thanks

0 Kudos
1 REPLY 1

Re: PCA and T-Square T^2

Hello Ron,

 

There is a note at the bottom of the multivariate control chart report that states that the Eigenvectors have been divided by the square root of the Eigenvalues.  This does not happen in the Multivariate Method platform and as best as I can tell this is the difference in results that you are seeing between the two platforms.

 

HTH

Bill