Cherry Chu posted on Wednesday, September 27, 2017 - 12:12 pm
I am running a two-part GMM model and I plan on investigating 2-6 latent classes for linear, quadratic, and cubic models, and for the experimental and control group separately. This means about 30 models I would have to run. I was wondering whether it is valid to run models for just 2,4,6 classes and if the BIC indicates the fit is increasingly better, then I can just keep the 6 class model without running the 3 and 5 class models?
If not, is there any other way I can investigate which model is best without running all 30? (because each model takes a long time to run)