I've read quite a lot in the discussion board on the issue of predicting continuous outcomes with categorical variables, in this case latent classes. As far as I understand, the only possible way is by obtaining the means of the outcomes for each level of the latent class.
Here is my problem:
In my study I have modelled two parallel latent classes, let's call them Alpha(A) and Beta(B), each one presenting with 3 levels/trajectories), which are also allowed to correlate with one another. As the outcput for the outcome means gives results for the probability of being in one of the levels in the class, I end up with 9 different means, as my 9 probabilities cover being in class A1B1; A1B2; A1B3; A2B1 and so on...
What I actually was looking for, was the mean of my outcomes for the 3 levels/trajectories in class A, and for the 3 classes/trajectories in class B.
That is exactly what I have done, but as I explained instead of getting 6 means of y (3 for c11 c12 c13 and 3 for c21 c22 c23) I get 9 y means, one for each Latent Class Pattern (i.e. c11c21; c11c22; c11c23; c12c21 and so on).
Is there any way of obtaining only the means for c11 c12 c13 c21 c22 c23?