Please forgive me if this answer is already available somewhere, but I've searched all day and been unable to find a solution - I'm an Mplus syntax novice. I am running a series of mixture models using DCONTINUOUS and DCATEGORICAL for distal outcomes predicted by the results of a latent class analysis (I had originally tried DU3STEP, but got the 'more than 20% classification error' message).
First, is there any way to add control variables, such that as C predicts a Y, controlling for X1's effect on Y (with X1 being excluded from the LCA)?
Second, is it possible to add a moderator to the DCONTINUOUS distal outcome, such that the interaction of the classes and an X2 predicts the outcome?
No, those features are not available with the DCAT/DCON method. See also the paper on our website:
Asparouhov & Muthén (2013). Auxiliary Variables in Mixture Modeling: 3-Step Approaches Using Mplus. Accepted for publication in Structural Equation Modeling. An earlier version of this paper is posted as web note 15. Appendices with Mplus scripts are available here.
Thank you very much for the quick reply. Would the most appropriate option then be to use the class probabilities as predictors (alongside controls and moderators) in regression, or would you recommend a different approach?
If your entropy is high (say > 0.8), you may want to use dummy variables to represent most likely class membership and use that in the regression. Otherwise, I am not sure which method is the least disadvantageous.