What if Parallel Analysis' factors/components do not correspond with PCA?

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

Fillis
Posts: 1
Joined: Fri Oct 28, 2016 8:17 pm

What if Parallel Analysis' factors/components do not correspond with PCA?

Postby Fillis » Fri Oct 28, 2016 8:25 pm

Hi,
I hope I am on the right board with my problem :)

I am using SPSS for my statistical analysis, and I want to apologize in advance for my lack of statistical knowledge, for it is my first time doing this.

Right now, I'm trying to decide how many components/factors to retain from the PCA. I did a PCA based on the covariance-matrix. The K1-rule (eigenvalue greater than one) and the Screetest ("elbow" in the Scree plot) do not correspond, not even close... Thats why I wanted to perform a Parallel Analysis. I used this code: https://people.ok.ubc.ca/brioconn/nfact ... ctors.html

And here comes my problem: The Parallel Analysis does its own PCA, if I get it right. I compared the eigenvalues of the PA's components with the ones my PCA put out, and they are not the same. I can't seem to find the problem or what causes the difference. Has anyone here an idea what I may be doing wrong?

I exported the SPSS output as ".htm". Sadly the terms are mainly german, but I hope some can read an understand the output it anyway.. the output includes: 1) PCA (title: "Faktorenanalyse") 2) Three different attempts to do a Parallel Analysis with rawpar.sps (see URL above) (titles each: "Matrix" and "Sequenzdiagram") 3) Attempt to du Parallel Analysis with parallel.sps (titles: "Matrix", "Korrelationen" and "Matrix" and here it is: https://www.dropbox.com/s/th7v7tao3sery ... E.htm?dl=0

Best regards

Who is online

Users browsing this forum: No registered users and 1 guest

cron