Syntax using final estimates for star... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Miriam Forbes posted on Wednesday, July 31, 2013 - 1:46 am
Hi everyone,

When my factor mixture models don't terminate normally, the output provides the syntax commands to run the model using the final estimates as start values.

My question is: is there a way to request this same output in a model that does terminate normally?


Thanks very much!

Miriam
 Linda K. Muthen posted on Wednesday, July 31, 2013 - 12:42 pm
Use the SVALUES option of the OUTPUT command.
 Miriam Forbes posted on Thursday, August 01, 2013 - 1:19 am
Thanks very much, Linda.


Miriam
 Joshua Wilson posted on Friday, March 07, 2014 - 5:42 pm
I'm running a Factor Mixture Model with 10 indicators variables, 3 latent continuous factors, and specifying a 4 class solution. The model terminates normally and I've replicated the best LL value.

However, I'm trying to reorder the classes to get the largest class as %C#4% in order to run the Tech11 and Tech14 comparison test. The largest class was assigned to %C#3%. To reorder the classes, I'm using the SVALUES output as input and just changing the names of the class labels so the largest class is now last.

The trouble I'm having is that I'm positing a fully invariant model so all my start values are the same. When I run the model it doesn't converge and all cases are assigned to a single class (%C#3%), the class which was previously the largest when not inputting user-specified start values.

Do you have any suggestions to correct for this issue?
 Bengt O. Muthen posted on Friday, March 07, 2014 - 10:57 pm
I don't know what you mean by a fully invariant model. For mixtures you want as many parameters as possible to be different across the latent classes in order to better determine the latent classes.
 Joshua Wilson posted on Friday, March 07, 2014 - 11:05 pm
Thank you. I see what you're what you're saying. I haven't yet specified class specific parameters. Currently, the only parameters that differ across classes are the means of the continuous factors across classes. Is there a way to reorder the classes in this situation?
 Bengt O. Muthen posted on Monday, March 10, 2014 - 10:23 pm
You can give good starting values for the factor means. But you don't have to have the largest class last for Tech11 and Tech14.
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: