I am attempting a multilevel model with categorical data. However, a core part of the theory involves moderation of several paths in the model. The easiest way to do this is with a multiple group model, however, the baseline multilevel model will only estimate in Mplus with montecarlo integration or Bayes.
In theory, one could use a knownclass mixture model to do the grouping. However, Mplus gives the error that paths at the between level cannot vary across classes---this is exactly what I want to test.
An alternative would be to parameterize the model with interactions, but this results in many interactions and does not estimate.