GLLAMM convergence question

[gllammuser]

Hello everyone,

I'm using GLLAMM for discrete random effects. but for nip>1 the convergence does not occur. I have tried almost all combinations of rescaling the variables in the model, but still facing the same problem. Could you please help?
Thanks a lot for your time and help in advance

[Joseph Coveney]

You could try fitting the model with the single integration point, saving the regression coefficient vector (e(b)) in a matrix and then using that matrix as the starting values for your nip > 1 case by using gllamm's from() option..

You might have to reconsider your model, whether it's properly specified, whether you're asking to much for the amount of data you've got, and so on. (I was under the impression that latent class models are inherently hard to deal with.)

P.S. It's best to start a new thread (topic).

P.P.S. In this list, you're urged to use your real name.