DIC for MGMM? PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 Qi posted on Friday, July 29, 2011 - 3:25 pm
Dear Dr. Muthén,

Is Deviance Information Criterion (DIC) available for multilevel growth mixture model? I noticed that in the latest Mplus, the "bayes" estimator is available and seems that DIC only available under bayes estimation, however, when I ran "twolevel mixture" with "estimator=bayes", I was told "Estimator BAYES is not allowed with TYPE=TWOLEVEL MIXTURE. Default estimator will be used.".

Besides, why MLR is the default estimator for "twolevel mixture", not ML?

Thanks for your enlightenment! And BIG congratulations to your recent award!!!

 Bengt O. Muthen posted on Friday, July 29, 2011 - 3:41 pm

We haven't yet gotten to introducing DIC everywhere in our Bayes procedures. And we have not yet gotten to two-level mixture Bayes. For some of the things in line to be implemented, see Version History for version 6 on our web site under the heading "Bayesian analysis using Mplus is an ongoing project. Features that are not yet implemented include:". Many more are in the pipe line.

We think that MLR protects SEs from various kinds of misspecification, and therefore is a better default than ML.
 Qi posted on Friday, July 29, 2011 - 4:41 pm
Thanks to your prompt reply! This looks like a good future research direction:-)

Mplus 6 is really exciting and I am looking fwd to the coming things.

I just want to say that this discussion board is really great (yet rare) for researchers in different fields and makes me feel much closer to top-notch experts like you, Linda, and the whole Mplus team.

Thanks again and have a great weekend!
 Yan Liu posted on Sunday, October 13, 2013 - 8:39 pm
Dear Dr. Muthén,

Is Deviance Information Criterion (DIC) available for a regular multilevel Bayesian model? Or any fit indices I can use for model selection as well as for model fit checking?

 Linda K. Muthen posted on Monday, October 14, 2013 - 9:43 am
The only way to do any testing with multilevel Bayes is the use MODEL CONSTRAINT with new parameters. No fit statistics are yet available for this.
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