Yuchun Peng posted on Wednesday, October 14, 2009 - 2:45 am
Dear Mplus support team, I have a general issue concerned about the random effect in the GMM model. Should the random effect for slope factor (given the fact it is significant factor in the Growth model) be added during the stepwise process of deciding the optimal number of clusters. In other words, 2-Cluster with Intercept random effect->2-Cluster with Intercept and slope random effect->3-Cluster with intercept random effect->3-Cluster with intercept and slope random effect...
Or The decision on the number of cluster can be based on the GMM model with random intercept factor first. And then add the slope factor afterward to see whether it can improve the model or not I.E.: 2-Cluster with intercept random effect->3-Cluster with intercept random effect->4-Cluster with intercept random effect (saying the 4-cluster with intercept random model was optimal).. and then run 4-cluster with intercept and slope random effect
The reason why I am asking this is the computation time of adding the random effect for slope factor is very heavy (it takes more than one week for my data!!) I wonder whether it is technically appropriate to only include the slope random effect after the decision on the number of clusters was made.
If the effects are random, you should allow that when determining the number of classes.
Yuchun Peng posted on Wednesday, October 14, 2009 - 12:21 pm
Thanks for the answers!!
My question now is is there any way I can speed up the computation time? I have used the PROCESS option, however, given that the total sample size is over 20,000 and the outcome variable is a categorical variable. It has been running since a week ago. Is it normal?