steve posted on Thursday, October 04, 2007 - 7:21 am
Hi, Dr. Muthen: What is the formula used to calculate the probability scale of latent class in MPLUS ? When I used the individual group membership file output from MPLUS to calculate the probability scale of latent class, it does not match the result from MPLUS. For example: For a binary output, each individual has the most likely latent class membership. Based on the membership, I could get the frequency of different categories for this class. However the ratio is different from the output from MPLUS. So I am wondering which method the MPLUS uses for the latent class?
We use the model class probabilities based on the estimated model and multiply them by the sample size. It sounds like you are adding up the number of people with most likely class membership 1, 2, etc.