GLM with log link and Poisson regression PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 James Algina posted on Thursday, March 13, 2014 - 12:35 am
I have bivariate multilevel data and would like to use the approach presented by Tihomir in GLM.pdf.

My model constraint code so far is
model constraint:
new W2int W2trt W2sL1 W2sL2 W4int W4trt W4sL1 W4sL2;
mean2=exp(W2int + W2trt*group + W2sl1*RL1_COV + W2sl2*RL2_COV);
mean4=exp(W4int + W4trt*group + W4sl1*RL1_COV + W4sl2*RL2_COV);

Is there a way to add a bivariate random intercept to the model?

Also the printout indicates that the df are negative and an identification warning. But the results are similar to what I get with univariate models in separate runs. Is the identification warning incorrect?

Thanks,

Jamie
 Tihomir Asparouhov posted on Friday, March 14, 2014 - 4:37 am
>Is there a way to add a bivariate random intercept to the model?

Formally speaking no, since the constraint feature is not available in the current version of Mplus for multilevel models. If your clusters are of limited size however you can do it by rewriting the model as a multivariate model (long-to-wide transformation).


> Also the printout indicates that the df are negative and an identification warning. But the results are similar to what I get with univariate models in separate runs. Is the identification warning incorrect?

Are you using the latest Mplus 7.11? I think this was already fixed.
Please send the example to support@statmodel.com. There should be no messages about DF, there should be no chi-square. It is something we overlooked. It is an error in the presentation but the model estimates should be fine.
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: