Multiple group analysis with MIMIC un... PreviousNext
Mplus Discussion > Categorical Data Modeling >
 Xu, Man posted on Tuesday, March 25, 2014 - 4:24 am
In a multiple group ESEM analysis under WLSMV estimator, I have a co-variate that I am interested in checking whether it has got DIF effect with the ordinal items of interests that form latent variables.

I was wondering if this is feasible. And if there is indication of DIF, is it sufficient/reasonable to allow a regression path between the covariate and the specific item (given that under WLSMV loading and threshold are not really separate)?

Thank you.

 Linda K. Muthen posted on Tuesday, March 25, 2014 - 10:11 am
I am not sure if this is possible with ESEM. Try doing this with MODEL TEST.
 Xu, Man posted on Wednesday, March 26, 2014 - 1:46 pm
Would modification indices pick up the non-zero paths between the co-variate and the items?

For example, I regressed the co-variate to the latent factor, then I regressed the co-variate to an item (ordinal measure). The parameter is free for the latent factor, but fixed to 0 for the item. This seems to "trick" the program to produce MI indices for the non-zero path between the covariate to all the items (not just the one contrained item).

Does this more or less capture uniform DIF between the co-variate and the items?
 Linda K. Muthen posted on Thursday, March 27, 2014 - 2:03 pm
Yes, fixing one direct effect to zero will cause the modification indices to be printed. You can also use MODINDICES (ALL).
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