Comparing MLR models with type complex PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Delphine Gross posted on Wednesday, February 08, 2006 - 11:09 am
I estimated two models with type = complex and estimator = MLR. I want to compare their chi-square. My problem is that I have to estimate the same model with estimator = ML to compare the chi-square (as indicated in one of your appendix) but I cannot do that because it is not possible to estimate a model ML and complex. What can I do to still obtain a chi-square difference test that is correct?
 Linda K. Muthen posted on Wednesday, February 08, 2006 - 3:11 pm
To obtain the ML chi-square, multiply the MLR chi-square by the scaling correction factor.
 davide morselli posted on Wednesday, April 08, 2009 - 2:54 pm
I need to compare a two-level MIMIC model to the same model with random slope. Using MLR (with Algorithm=integration) it gives me the message THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE
TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE
FIRST-ORDER DERIVATIVE (.....) PROBLEM INVOLVING PARAMETER 23.

but then

THE MODEL ESTIMATION TERMINATED NORMALLY

I heard that it is not a relevant error, but then Mplus can't calculate Chi-square, CFI, RMSEA, only IC test (BIC, AIC). THe WLS estimation doesn't give this error, but I can't use it with TYPE = TWOLEVEL RANDOM, for the computetion of the random slopes.

I could compare the IC but It is a pretty weak test.

I think the problem is taht the baseline model does not converge with the MLR. How can I overcome it?

Thnx in advance for your reply
 Linda K. Muthen posted on Wednesday, April 08, 2009 - 4:40 pm
Please send your input, data, output, and license number to support@statmodel.com.
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: