Regularized Common Factor Analysis?

[genemroz]

I've been using Proc Factor to undertake common factor analysis of a large dataset.  But the results have been plagued with Heywood cases.  I've done some reading that suggests a method called Regularized Common Factor Analysis (RCFA) can help in these cases. But it's not clear to me exactly what this is nor how or even whether this can be accomplished with Proc Factor or some other SAS Proc.  I'm looking for some guidance on how to proceed or where to find further information.

Thanks in advance,

Gene

[sbxkoenk]

I guess you use PROC FACTOR with METHOD=PRINCIPAL.
I guess you also use a PRIORS statement or a PRIORS= value other than PRIORS=ONE.

Correct?

And indeed ... the handling of Heywood cases by the HEYWOOD option only has an effect for the iterative methods (PRINIT, ALPHA, ULS, ML). For the noniterative methods (PRINCIPAL, IMAGE, HARRIS) there is no effect by the option.

The FACTOR Procedure
Heywood Cases and Other Anomalies about Communality Estimates

https://go.documentation.sas.com/doc/en/statug/latest/statug_factor_details10.htm

BR, Koen

[genemroz]

Thanks for the link to the documentation.  I'll study that more closely...but I think I'm looking for something different.  Namely, I'm trying to learn about a specific and specialized form of factor analysis that goes by the moniker of "Regularized Common Factor Analysis (RCFA)" and how to implement it within SAS.

[sbxkoenk]

I know you are looking for something different.

Namely this:

New Trends in Psychometrics

Regularized Common Factor Analysis

Sunho Jung and Yoshio Takane

Department of Psychology, McGill University, 1205 Dr. Penfield Avenue, Montreal, QC, H3A 1B1, Canada

In SAS you find regularization methods for regression (penalized regression methods), but I do not see any regularization methods built into PROC FACTOR. ☹️

• L1 regularization is known as LASSO.
• L2 regularization is known as ridging. 

BR, Koen