I'm working on a 5-wave latent class growth model where indicators (both continuous and categorical) changed over time. I'm following your handouts on the topic and have initially estimated a simple latent growth model and established factor loading and intercept invariance without much trouble.
I now want to examine the presence of latent classes and added the type=mixture, %overall%, and classes= parts. This model results in the little discussed error: "Growth factor indicators must be all observed or all latent" and refers to one of the (latent) factors representing a time point with 5 observed indicators of which two are categorical.
After some modifications to the code I find that variables cannot be defined as categorical.
Put differently, when the two categorical variables are defined as categorical in the Variable part, the model ends in error. If this one of code is omitted yet no other changes made (i.e., same model and indicators), the model converges.
I would be very interested in what is going on here and why this happens. Thanks a lot in advance!