I am conducting GMM to examine trajectories of smoking behavior across 3 timepoints.
DATA: file is smokers.dat; VARIABLE: NAMES ARE vid new t1 t2 t3; MISSING = all(-1234); USEV ARE vid t1 t2 t3; CATEGORICAL IS vid; CLASSES = c(2); ANALYSIS: TYPE = MIXTURE MISSING; STARTS = 1000 20; STITERATIONS = 20; ALGORITHM = INTEGRATION; MODEL: %OVERALL% i s | t1@0t2@1t3@2; i s ON vid; c#1 ON vid; OUTPUT: standardized tech8 tech11 tech14;
I have read all of the GMM articles on the website and have two questions:
1. Based on theory and BIC/Entropy data I would select the 3 class model. Based on BLRT, the 2 class model would be preferred. Would you recommend using the BLRT?
2. How do I assess the influence of the covariate (vid)?
1. I would choose the one that makes the most sense substantively?
2. Do you mean how do you tell if the relationship is statistically significant? Column 3 of the output is the ratio of the parameter estimate to the standard error of the parameter estimate. This is a z-value.