lam chen posted on Wednesday, November 14, 2012 - 9:07 am
I have a three-class LCGM with four covariates.My destination is to explore this variables' prediction on the class membership.I have tried this in two ways:
1.a LCGM with four covariates as above; 2.firstly by a unconditional LCGM,and then output the cprobabilities and conduct a multinomial logistic regression as c regresses on X1-X4(four covariates) .
The results(logistic coefficients) of these two methods were different. To my conprehension, in LCGM with covariates, c is also multinomial logistic regress on covariates.Why the results are different? Which will be the right answer to my problem?
This is most likely due to the need for direct effects between the covariates and the outcome. See Web Note 15 and the following paper which are available on the website:
Muthén, B. (2004). Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In D. Kaplan (ed.), Handbook of quantitative methodology for the social sciences (pp. 345-368). Newbury Park, CA: Sage Publications.