Correlation matrix is not positive definite and singular?

luinel — Fri, 26 Apr 2019 11:16:01 GMT

Hello,

I am trying to do a factor analysis (PCA) on a large data set. It has 181 columns and 163 rows, values are between -6 and 8. I have started the factor analysis and the eigenvalues and scree plot looks good.Below all the factors it says: "Warning: the Correlation matrix is not positive definite". Can someone help me understand what it means?

I try to perform the analysis using Maximum Likelihood and Common Factor Analysis with a Varimax rotations method. But I get an error message saying: "Maximum likelihood method requires a nonsingular correlation matrix. Switching to principal components method may fix the problem."

I am only a beginner in JMP but I appreciate any help on this matter.

Thank you!

/Jenny

Re: Correlation matrix is not positive definite and singular?

Mark_Bailey — Sun, 28 Apr 2019 12:24:38 GMT

You mentioned that you are a new JMP user. (Welcome!) You did not say what is your level of knowledge and experience with the PCA methods.

Select Help > Books > Multivariate Methods. Search for "not positive definite." I found this information:

topic Re: Correlation matrix is not positive definite and singular? in Discussions

Correlation matrix is not positive definite and singular?

Re: Correlation matrix is not positive definite and singular?