BSEM limitations PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Lois Downey posted on Thursday, January 28, 2016 - 8:30 am
In an earlier response to my post on a different thread, you indicated that it is not yet possible to use BSEM to test for scalar measurement invariance when indicators are ordered categorical variables, because the DIFF specification for thresholds of polytomous items has not yet been implemented.

Is it also the case that BSEM does not yet allow specification of informative priors for residuals for polytomous items? (I'm getting an error message that seems to imply that.)

Thanks!
 Tihomir Asparouhov posted on Monday, February 01, 2016 - 1:21 pm
With the Bayes estimator, the residual variances are fixed to 1 as in the probit regression. For the residual covariances you can specify the Inverse Wishart prior.
 Lois Downey posted on Monday, February 01, 2016 - 5:35 pm
Oh, yes! I was incorrectly specifying IW priors for the variances, in addition to the covariances. I should have fixed the variances to 1.0. Thanks very much!
 Lois Downey posted on Thursday, February 04, 2016 - 6:50 am
I now have a follow-up question. When I corrected my command file as indicated above, the run aborted without an error message. After I submitted my files to Mplus Product Support for analysis, I was told, "There is a problem with the standardized options. If you remove STDY, you will be fine." I did as instructed, and the run finished normally.

Is this is a temporary problem that will be solved in the next Mplus release, or is it a problem that is expected to continue indefinitely? (I'm unable to make a judgment about whether the residual covariances are small enough to be tolerable, whereas a similar judgment about correlations would be simple.)

Thanks!
 Linda K. Muthen posted on Thursday, February 04, 2016 - 10:46 am
We always fix every problem found in the next release. If you need something run in our developmental version, send the input and data along with your 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: