LCA Stata15 - discontinuous region encountered

[Marie Baron]

Hello,

I am a beginner with LCA models. I am trying to run latent class models on Stata 15 but I have difficulties.
My model is
gsem (mhi5_76 dlang_d2 <-, logit) ( srh_c dpal ddrktype socsupp_c <-, mlogit), lclass(B 3) startvalues(randompr, draws(20))

I tried to run the model from two to five class to be able to chose the best model.
However, the program runs for the two class but doesn't work for the four and five latent class model.
it sometimes run for the three latent class model, but not systematically.

The error message is the following : discontinuous region encountered

Looking at the help menu for this error and tried the "difficult" and it didn't work either.

I am not sure what are the next steps after and I have a hard time understanding what might be the sources of the problem.

Thank you for any help you could give me.

Marie

[Red Owl]

Marie,

You might find some help in the responses from Jeff Pitblado of StataCorp in #4 and #5 in a previous thread at https://www.statalist.org/forums/forum/general-stata-discussion/general/1397167-latent-class-analysis-with-stata-15-gsem-problem

I would recommend you give special attention to the -nontolerance- option added in #4 and then to the followup in #5.

Red Owl
Stata/IC version 15.1, Windows 10 (64-bit)

[Marie Baron]

Thank you very much Red Owl.

It seems to be working. I have to investigate more about these options, but it seems to be the solution.

Thanks

[Donghui Wang]

Dear Red Owl,
I encountered the same issues but the links you provided above are no longer valid.

[Weiwen Ng]

Strange. When I click the link, I also get an message that it's invalid. When I searched for the post manually, I found it, and the URL appears the same.

https://www.statalist.org/forums/for...5-gsem-problem

This should work.