Starting values in LGMM PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Anthony Mancini posted on Tuesday, September 02, 2008 - 9:18 pm
In a 4-class LGMM using a continuous outcome, I am trying to designate a specific class as the referent, so that covariates predict that class in relation to each of the others.

However, when I use starting values (mean values for the intercept and slope derived from a previous run), the appropriate class is never the 4th class (the referent). I have tried using starting values for all of the classes and for the 4th class alone.

In my searches of the discussion board, I have not been able to determine what I am doing wrong. How are starting values determined in this context? Any help is appreciated.
 Bengt O. Muthen posted on Tuesday, September 02, 2008 - 9:34 pm
The latest version of Mplus should give the covariate effects under all possible referent classes.

Otherwise, you can also give starting values for the class probabilites in logit scale - which might be needed if classes are of unequal size.
 Anthony Mancini posted on Wednesday, September 03, 2008 - 8:23 pm
Okay...I will try using starting values in logit scale. (Or I'll get computing to install the latest version.) Thanks as usual!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: