Ben Porter posted on Monday, December 30, 2013 - 12:06 pm
Dear Drs. Muthen, I am running a model with one known class with 2 levels and one latent class with 4 levels. I have gotten the model to estimate parameters for each of the eight distinct classes (e.g. g#1.c#1). I wanted to test likelihood of class membership between known groups (e.g. liklihood of c1 given g1 = likelihood of c1 given g2). Unfortunately, the class orders within groups are not parallel when using random starts so c on g does not measure the specific comparisons I am interested in.
Beyond using a c on g statement, I was wondering if it is possible to manually specify thresholds for latent classes within known groups to allow comparisons of class probabilities when classes are not ordered similarly within groups. If this is not possible, is there a way to reorder the classes to be parallel while still using random starts. Thank you