DIFFTEST in Multilevel CFA PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Jonathon Little posted on Tuesday, May 05, 2009 - 6:09 pm
Dear Muthen & Muthen,

I notice that the DIFFTEST option is not available in a multilevel CFA. My analysis uses a saturated within-level model with a 4 factor measurement model fit to the between level. The estimator is WLSMV with categorical observables.

So that I may assess the improvement in fit for my between level model when compared to the same between level model with an error covariance freed, is there some way that I can do the DIFFTEST manually without being exposed to any complicated mathmatics or horrible equations?

Kind regards,

Jonahton
 Bengt O. Muthen posted on Wednesday, May 06, 2009 - 1:50 am
I am afraid it involves complicated statistics. The easiest way out is to use WLSM instead of WLSMV.
 Jonathon Little posted on Sunday, May 10, 2009 - 5:48 pm
Cheers for this Bengt.

When you say that the easiest way out is to use WLSM instead of WLSMV, am I correct in assuming that when using WLSM that a comparison of nested models is simply a matter of computing the difference between the chi-square and degrees of freedom for the two models in the same way that you would for normal theory ML?

Jonathon
 Bengt O. Muthen posted on Sunday, May 10, 2009 - 6:13 pm
Right.
 Jonathon Little posted on Thursday, July 02, 2009 - 1:53 pm
Ok this is a little confusing Bengt,

After reading page 367 of the User manual and the standard MPLUS output that is printed below:

"The chi-square value for MLM, MLMV, MLR, ULSMV, WLSM and WLSMV cannot be used for chi-square difference tests. MLM, MLR and WLSM chi-square difference testing is described in the Mplus Technical Appendices at www.statmodel.com. See chi-square difference testing in the index of the Mplus User's Guide"

Can you re-read this thread and my questions and explain this contradiction for me.

You answered "right" to my question which read "am I correct in assuming that when using WLSM that a comparison of nested models is simply a matter of computing the difference between the chi-square and degrees of freedom for the two models in the same way that you would for normal theory ML?"

Regards,

Jonathon
 Linda K. Muthen posted on Thursday, July 02, 2009 - 5:19 pm
You need to take the scaling correction factor into account when doing difference testing for WLSM. See Chi-Sqare Difference Test for MLM and MLR on the website. This applies also to WLSM.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: