EFA Monte Carlo PreviousNext
Mplus Discussion > Exploratory Factor Analysis >
 Dave Flora posted on Wednesday, November 22, 2006 - 2:03 pm
I'm running a series of EFA monte carlo programs, and have noticed that the following occurs when I specify 2 factors, with half of variables loading on the first factor (say .8) with 0 loadings on the 2nd, and then the other half of the variables load on the second factor (again .8) with zero loading on the first (and factor correlation = .6):
With ML estimation, this simple structure is reproduced nicely on average across replications, but with ULS estimation, the average rotated loadings are about .4 on both factors for all variables.
Similarly, if I make the variables categorical, the average promax loadings again reproduce the population structure nicely with WLSM estimation, but not with ULS estimation.

It seems like this phenomenon is something that only happens in Monte Carlo, because when I use a real data file that fits a 2-factor model reasonably well, the ULS and WLSM Promax loadings are very similar.

Any idea what's going on?
 Bengt O. Muthen posted on Wednesday, November 22, 2006 - 2:34 pm
That sounds strange. Please send your ULS monte carlo input to support@statmodel.com.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message