I am trying to run a 4 class GGMM with a sample size of 1,000,000. I can get the model to run using subsamples of 1,000 and even 10,000 -- but when I increase the sample size, the model will not converge. I tried to increase the convergence criteria to .005, but it did not help. I still get the following error message:
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO A CHANGE IN THE CLASS COUNTS DURING THE LAST E STEP.
AN INSUFFICENT NUMBER OF E STEP ITERATIONS MAY HAVE BEEN USED. INCREASE THE NUMBER OF M ITERATIONS. ESTIMATES CANNOT BE TRUSTED. THE CLASS COUNTS CHANGED IN THE LAST EM ITERATION FOR CLASS 1.
I think this is because with even the most minor of parameter adjustment, at least one person would change classes, which would generate the above error message. Is there a way to adjust the convergence requirements for the class counts for this with such a large sample size? Or is there some other problem?