GSEM multilevel interaction - explained variance

[Gina Groelsh]

Hello
I have calculated a multilevel interaction effect using gsem, very much the same as example 38g in stata 16 manual (p.402) titled "Random slope model, single equation formulation". It may just be a case of misunderstanding the output, but how do I find the R2 value in order to do an explained variance calculation? The example does not include this.

[Phil Bromiley]

In general, SEM is not minimizing the sum of squared errors (it is minimizing the different between the covariance matrix under the model constraints and the empirical covariance matrix) so r-squared is not generally appropriate. I guess you can calculate one by using the error variance. If you try to do so, I'd start by running a simple model in an estimator that gives r-square and then the same model in gsem. Then you can tell if your r-square calculation gives the right value.