

Latent class linear mixed model? 

Message/Author 

Lei Yu posted on Tuesday, April 08, 2014  5:26 pm



Dear Drs. Muthen, I am a new user of Mplus. I have a few questions on the growth mixture models. The basic syntax, as copy and paste from the users guide, seems to imply both intercepts and slopes are contributing to the latent class. ANALYSIS: TYPE = MIXTURE; MODEL: %OVERALL% i s  y1@0 y2@1 y3@2 y4@3; i s ON x; 1. Is it possible to restrict only slope to have latent class structure? 2. based on the above example code, is latent class capturing the heterogeneity of random intercepts and slopes (i.e. after controlling for the 'effect' of x on both mean intercept and slope)? I am thinking of a random intercepts and slopes model (for longitudinal data), but would like to have random slopes follow a mixture normal (latent class), while keep random intercept a regular normal. Can growth mixture model be applied in this situation? Thanks Lei 


I think you can achieve what you want by holding the mean of i equal across classes, for example, [i] (1); in the overall part of the MODEL command. 

Lei Yu posted on Thursday, April 10, 2014  1:17 pm



Thanks, Dr Muthen That really helped. I do have a followup question, I tried to get the LoMendellRubin test and the bootstrapped likelihood test using TECH11 and TECH14. But Mplus shows the warning message that neither was available for the model with type=mixture random. Is there a statistic to determine the optimal number of classes for longitudinal data other than AIC or BIC? My sample code Data: file is 'mplusdta.dat'; Variable: names are x y1y18 t1t18; usevar =x y1y18 t1t18; TSCORES = t1t18; class = c(2); missing = ALL(1234); Analysis: type = mixture random; processor=4; STARTS = 100 8; Model: %OVERALL% i s  y1y18 at t1t18; i s on x; %c#1% [i](1); i(2); y1y18(100); [s]; s; %c#2% [i](1); i(2); y1y18(200); [s]; s; Output: tech1 tech8 tech11 tech14; 


I think BIC is best in this case. 

Back to top 

