Subscribe Bookmark RSS Feed

Kaiser-Meyer-Olkin (KMO)

dysvick

Community Trekker

Joined:

Jun 23, 2011

Prior to performing PCA or Factor Analysis it is a good idea to perform two tests to determine whether components or factors will result from the analysis or whether it will be a waste of time. The Kaiser-Meyer-Olkin index (KMO) of sampling adequacy and Bartlett's test for sphericity are such tests. Where is the KMO test in JMP?
2 REPLIES
mpb

Super User

Joined:

Jun 23, 2011

I don't believe JMP has a direct calculation for KMO. I'm sure someone will correct me if I'm wrong. You can do it yourself with some effort. Under Multivariate you can produce Correlations and Partial Correlations and then turn them into data tables using Make Into Data Table. Then you can manipulate them manually or via scripting to calculate KMO from its definition.

Alternatively you can assess collinearity by looking at the VIF values. They can be gotten from the Multivariate platform by asking for Inverse Correlations. The diagonals of this matrix are the VIF values for the associated variables.
dysvick

Community Trekker

Joined:

Jun 23, 2011

Thank you for your reply. I am aware of the methodology for computing KMO. I was hoping for a more direct solution. SAS does it as a normal step in Factor Analysis and PCA. This is so basic to research, but labor intensive manually, that you would think the JMP folks would have recognized the importance long ago. That being said, I have sent numerous corrections for the textbooks written under the SAS Press logo to correct errors in the documented PCA methodology for the JMP application. Apparently, there is no peer review of published (and expensive) books purporting to provide step-by-step guidance for conducting multivariate analyses.