Complex Samples and categorical outcomes PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
 Walt Davis posted on Wednesday, January 16, 2013 - 8:11 pm

I note there are several estimators available for complex samples with categorical outcomes. When I ran complex samples with continuous outcomes, it defaulted to MLR. (all the LVs are treated as continuous) When I run it with categorical outcomes, it defaults to WLSMV. Mainly I want to be sure that in both cases the complex sample design is being accounted for. Is there a preferred estimator in this scenario?

Also, when I specified "ESTIMATOR is MLR" for the categorical model, it refused to estimate error covariances between categorical observed. However WLSMV does this without complaint.

MLR is also causing memory problems for integration. That's not an issue if I can use WLSMV.
 Linda K. Muthen posted on Thursday, January 17, 2013 - 7:36 am
If you specify TYPE=COMPLEX and the CLUSTER option along with other complex survey options, they are taken into account unless you receive a message to the contrary.

With maximum likelihood and categorical options, the WITH option cannot be used to specify a residual covariance because each residual covariance requires on dimension of integration. You can specify them using the BY option:

f BY u1@1 u2;
f@1; [f@0];

where the residual covariance is found in the factor loading for u2.

The memory problems are because of numerical integration. WLSMV does not require numerical integration with categorical outcomes.
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