Issues running Finite Mixture Models (FMMs) in STATA 15

[Jake Sully]

The main issue here is that I get the error message "initial values not feasible" after running more than three FMMs. I'm running the models on sub datasets that are the result of random sampling of 50% of the original dataset which stays the same in each iteration. Whether I'm running the models on a loop/program or not, doesn't seem to matter, the result is the same. A few more details, the FMM is that of a negative binomial regression, the data has panel variables, and the total number of observations being modelling exceeds 70k.

If you have any further questions, please let me know, thanks!

[Weiwen Ng]

Read SEM intro 12, particularly the section on getting better starting values with gsem and categorical latent values - I'm pretty sure you can skip the section about continuous latent variables. Also, it may be worth thinking about simplifying the model - for example, a Poisson model may be adequate.

Also, did you mean that you get the error message after running more than 3 FMMs, or after running an FMM with more than 3 classes? Not quite clear what you mean. It is likely to help if you post the exact commands you typed, and the exact error message, in code delimiters.