Michael Li posted on Friday, October 02, 2015 - 2:59 am
Dear Dr. Muthen,
I've read recommendations for conducting GMM, and most involves the following steps: identify the number of trajectories, examine covariates associated with the trajectories, and then test a final GMM that included relevant covariates.
However, I'm not quite understand that after I examine covariates significantly associated with the trajectories, why should I included relevant covariates in the final model? What's the differences between conditional and the unconditional GMM?
You don't have to take that final step. But it can be useful for example to check if there are direct influences from covariates to growth factors or outcomes - which would make decisions on the number of classes different from using the unconditinal GMM.
See also the paper on our website:
Petras, H. & Masyn, K. "General growth mixture analysis with antecedents and consequences of change." Handbook of Quantitative Criminology. Ed. Alex Piquero, Ed. David Weisburd. New York: Springer-Verlag, 2010. 69-100. download paper contact first author