Ty Borders posted on Thursday, March 25, 2010 - 7:41 pm
I have identified several latent growth classes for four different variables (powder cocaine use, crack cocaine use, methamphetamine use, and alcohol use). I would like to treat the classes for the first three (powder cocaine, crack cocaine, and methamphetamine) as Independent variables in a LCGA of alcohol use. Is it okay to use latent growth curves to predict other latent growth curves? Can the classes for the independent variables simply be entered as time invariant variables in an analysis of alcohol use?
If you have four categorical latent variables representing four curves, you can do a multinomial logistic regression using one as the dependent variable and the other three as independent variables, for example,
c4 ON c3 c2 c1;
This is a complex model so it may present difficulties. Do it stepwise.