Carolin posted on Tuesday, June 14, 2011 - 12:57 am
I'm running a GMM with 4 timepoints. The variances (or residual variances in the model with covariates) of the growth factors are equal in all classes, the intercepts not. I read in the user guide that this is the default. Could you please explain to me why you chose this default?
Are the growth factors still random (allowing for within class variation)?
We hold the growth factor variances equal across classes in GMM to avoid possible convergence problems. You can look at plots to see if it looks like the variability differs across classes and relax the equalities accordingly.
Yes, the growth factors are still random. The growth factor variances would need to be fixed at zero for them to be fixed.
Carolin posted on Thursday, June 16, 2011 - 12:44 am