Anonymous posted on Sunday, March 23, 2003 - 6:41 pm
I am just beginning to read about growth mixture models and I have a question about the interpretation of the parameters in your Penn8 example (from your mplus examples). By regressing the growth parameters (f1, f2, and f3) on the covariates, are you causing the latent class variable to capture variance in heavy drinking NET the effect of the covariates?
bmuthen posted on Sunday, March 23, 2003 - 9:59 pm
I would turn it the other way around. The covariates influence the latent class membership. Latent class membership influences the growth factors f1, f2, and f3 (especially their means). The fact that the covariates also influence the growth factors directly implies that even given class membership the covariates still have a within-class influence on the growth factors.
Anonymous posted on Monday, March 24, 2003 - 12:43 pm
Why did you choose to eliminate the regression of the growth parameters on the covariates in paper #85 (app3)?
bmuthen posted on Monday, March 24, 2003 - 2:43 pm
No substantive reason for this. Typically, one would want to include them.