Implications of doing a confirmatory factor analysis with a correlation matrix as input instead of a variance-covariance matrix?

Asked Jun 04 '14 at 03:31

Active Jun 04 '14 at 03:31

Viewed 196 times

Is this possible, and if so what are the implications of doing things one way rather than another.

Is one approach generally preferable? So far I have only been taught to use a variance-covariance matrix.

asked Jun 04 '14 at 03:31

Absolutely the same way (corr matrix is cov matrix of standardized variables). Actually, in FA corr is used more often. However, the results, of course, will be different, if the variances of the variables were not all the same. – ttnphns Jun 04 '14 at 06:12
There once was a discussion about implications: http://stats.stackexchange.com/q/62677/3277. It is about PCA, but the same is valid about FA. – ttnphns Jun 04 '14 at 06:14

0 Answers0