I think MPLUS would give you a better set of absolute and relative model fit indices than Stata. I am not sure how easy it is in Stata to specify different mean and covariance structure within and across classes.

Hello

You can use Mplus, the free R package or a stata version 17 or more to achieve LCA/LPA.

Hi,

I use both. I run preliminary analyses with Stata, and then Mplus. The advantages are several but maybe there are already Stata packages catching up. First, there are more modification indexes in the output. Second, Mplus indicates where is the mistake in your code when there are errors. Third, Mplus is much more computationally efficient than gsem. Fourth, you can integrate multilevel analyses in Mplus (not only multigroup analysis, but actually modeling of random effects). Fifth, in Mplus you can use full-information maximum-likelihood estimation rather than a listwise deletion of cases when you have issues with incomplete cases. Sixth, maybe more importantly, I took classes with the programmer of sem in Stata and he taught using Mplus.

there are more fit statistics available in mplus. And i think you can play with the variance/cov structure easily in mplus. I think the help for stata is not good at this point in time. If you really have to, i rather you do it in R using tidylpa and can capitalise on ggplot to do the visualization