

Multigroup tests of indirect effects 

Message/Author 


Dear Bengt and Linda, I'm conducting multigroup comparisons in a model where I have two multiple indicator growth curves and some other variables. In this model, some variables predict the change/slope factors through the level factors, and I would need to test whether these effects differ by gender. When I run the whole model by gender, and specify path constraints under the 'Model indirect' command, these do not influence the dfs of Chisquare value of the model (i.e., the traditional multigroup comparisons do not appear to work)  how does one conduct gender difference testing of such indirect effects in Mplus? 


Indirect effects are calculated after the model is estimated so MODEL INDIRECT has no impact on the model being estimated. If you want to test whether the indirect effects are equal across groups, you would need to use MODEL CONSTRAINT. 


Thank you Linda, this is what I aim to do. I tried the 'model constrained' but it didn't work (the error message: 'Unknown parameter label in model constraint', although I did specify the label before this). In the syntax, I first have all the directional paths in the model, then I have the 'model indirect' command under which I specify the indirect effects, and finally, I have the 'model constraint' command. There's just probably something in this syntax that I don't get..Could you tell/show me what the syntax should look like? Also, is it enough to give the parameter label and a constrain to it under the 'Model Constraint'? 


Please send your output and license number to support@statmodel.com if the information in the user's guide about MODEL CONSTRAINT does not help you solve this problem. 

Back to top 

