Simple procedure to determine optimal number of classes in a latent-class model?
Posted: 10 May 2022, 13:48
Hello Prof Hess
In latent class models, selecting the "right" number of classes is important, a procedure, as you know, often done by testing the BIC / AIC / etc of variously-sized models.
Because of the way models are specified, with the number of classes hard coded in both apollo_probabilities() and apollo_lcPars(), I suspect there may not be an easy way to do this in apollo yet. Is that correct?
Or is there a way to hard code, say, the maximum number of classes, N, one wants to test, and have apollo cycle through models with as few as 2 classes and as many as N—instead of creating N-1 hard-coded apollo_probabilities() and apollo-lcPars() functions and running them individually?
Thank you!
In latent class models, selecting the "right" number of classes is important, a procedure, as you know, often done by testing the BIC / AIC / etc of variously-sized models.
Because of the way models are specified, with the number of classes hard coded in both apollo_probabilities() and apollo_lcPars(), I suspect there may not be an easy way to do this in apollo yet. Is that correct?
Or is there a way to hard code, say, the maximum number of classes, N, one wants to test, and have apollo cycle through models with as few as 2 classes and as many as N—instead of creating N-1 hard-coded apollo_probabilities() and apollo-lcPars() functions and running them individually?
Thank you!