Estimated means for multilevel mixtur...
Message/Author
 Patrick Malone posted on Wednesday, October 20, 2004 - 4:02 pm
I'm running a multilevel mixture model -- students nested within schools, with two classes defined by a single continuous student-level indicator, and a number of student- and school-level predictors of both class membership and the value of the predictor (within one class).

I'm specifying the mean structure for the class indicator at the between level, but when I run the TECH4 output, I get estimated means for the indicator at both the between and within levels.

What is the interpretation of the within-level means of the indicator? There are no intercepts apparent in the within-level portion of the results?

Thanks
 Linda K. Muthen posted on Thursday, October 21, 2004 - 11:02 am
In TECH4, within only variables have their means on the within level. Between only variables and variables that are both between and within have their means on the between level. If a dependent variable is regressed on a within only variable or set of variables, the model implied mean of the dependent variable is a function of the within variable means of the independent variables and is reported on the within level. I added TECH4 to Example 10.1 to figure this out. You can do this also.
 Patrick Malone posted on Thursday, October 21, 2004 - 12:52 pm
Thanks. My confusion is in how to interpret the two different means. Is the between-level mean the mean of the between-unit aggregates? If so, what is the within-level mean?

Thanks
 Linda K. Muthen posted on Thursday, October 21, 2004 - 1:46 pm
The one on the between level is the between part of the mean of y implied by the model. The one on the within level is the within part of the mean implied by the model. The model estimated mean is the sum of these.
 Patrick Malone posted on Sunday, October 24, 2004 - 5:23 pm
Great, thanks.
 Nordin abd Razak posted on Tuesday, January 23, 2007 - 7:27 pm
Dear Linda,

I run the multilevel mixture model, with cross-level interaction effects. However, I didn't get the R-square for latent variables in the structural path model. Is it MPLUS does not produce R-square for this kind of model??
 Linda K. Muthen posted on Wednesday, January 24, 2007 - 7:59 am
R-square is not available in this situation.
 Nordin abd Razak posted on Wednesday, January 24, 2007 - 3:49 pm
Thank you Linda