

Class membership in LGC mixture model... 

Message/Author 

John posted on Sunday, January 01, 2006  8:47 am



Hello  even though this is a more general question, I would be happy if I could get some advice on it: Problem: I have dichotomous indicators repeatedly measured at four points in time (each one year apart, panel study). An initial latent class analysis showed that a fourclass model fits best to the data. Further, a plot of the LCclasess shows that two of the four classes 'converge' at the fourth measurement point  they seem to have the same probalities for positively answering the single item at time 4. I now want to model the data as a latent growth curve mixture model. Question: Using latent growth curve mixture modeling (or any other approach), how could I estimate the probability that members of one latent class do change or do not change their classmembership (=trajectories) for the period under study? Is there any applied literature? Best regards & happy new year, John 

John posted on Sunday, January 01, 2006  8:49 am



sorry for the following error: I estimate the whole model using several dichotomous indicators, not just one single indicator 


With one latent class variable, individuals cannot change class over time. You might consider latent transition analysis. 

Back to top 

