LCA-model fit with alternative not ne... PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 Andres Cardona posted on Wednesday, June 03, 2009 - 10:16 am

Im trying to find out which of 3 alternative (observed) classes best fit different individual level outcomes summarized in a single latent variable ("life chances"). I thought a LCA would do the job, however im struggling with 2 questions:

1. Whats the best way to include the restriction in a LCA for latent classes to be defined by observed classes? Is Example 7.21 the right model specification?

2. After estimating 3 different models for 3 different observed classes, how can I decide which model best fits the data given that the alternative classes are NOT nested? (all Ive read about this issue refers to models comparing different nested classes).

Id appreciate your expert advice on this matter. Im sure the answer is not very complicated, but i couldnt find it out by my own means. Thank you.
 Linda K. Muthen posted on Thursday, June 04, 2009 - 9:42 am
1. If you mean observed classes like gender, the KNOWNCLASS option should be used as shown in Example 7.21.

2. People use BIC to compare non-nested classes.
 Andres Cardona posted on Friday, June 05, 2009 - 5:04 am
Thanks for your answer. With observed classes I mean occupational grouppings(i.e. "manual workers" or "managers"), which are not individuals traits but "level 2" variables. Id like to restrict the LCA to fit this groups.
 Linda K. Muthen posted on Friday, June 05, 2009 - 9:23 am
Put the categorical latent variable on the between list:

CLASSES = cb (2);

[occup$1@15]; !occup = 0
[occup$1@-15]; !occup = 1
 Andres Cardona posted on Saturday, June 06, 2009 - 4:48 am
Thanks again for your reply. Ill try it that way.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message