Multilevel mixture modeling PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
 PAUL R. SWANK posted on Tuesday, March 20, 2007 - 11:56 am
What we have are a number of variables having to do with certain schools and students. The student measures are outcomes, the school measures are more facillities and process. We wanted to indentify latent classes of schools based on their facilities and process data as well as the student outcome data. However, we are told that we can't define latent classes in terms of the between subjects variables. We had hoped not to have to aggregate the student level data. Do you see a way around this?
 Bengt O. Muthen posted on Tuesday, March 20, 2007 - 4:46 pm
Interesting question - I am glad you asked. With the introduction of Mplus version 4.2, it is possible to create latent classes on level 2 as well as level 1. Such level 2 classes can have as indicators observed level 2 variables (school measures) as well as the level-2 variation in individual's (students') variables (the latent continuous variables of random intercepts and slopes). The version 4.2 User's Guide addendum gives several examples of such modeling.
 IYH Boon posted on Thursday, May 05, 2011 - 9:38 am
I am in a somewhat similar situation. I have repeated measures for a level 2 variable (school characteristics observed over a sequence of 9 years). Ideally, I'd like to model these data using LCGA or GMM, and then use the resulting latent classes to predict student-level outcomes (test scores observed after the last time point). Is this possible? Are there examples that I can look at? Thanks in advance.
 Bengt O. Muthen posted on Thursday, May 05, 2011 - 6:57 pm
Yes, you can specify level-2 latent classes by say

between = cb(2);

where the level-2 growth factor means change over cv. Then you let the between-level part of the student-level outcomes vary as a function of cb.

You should have a look at

Henry, K. & Muthén, B. (2010). Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling, 17, 193-215.

which is on our web site with Mplus scripts under Papers, Multilevel Mixture Modeling. Although in a cross-sectional as opposed to longitudinal setting, this paper gives you a better view of what can be done.
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