st: Marginal effects for cross-level interactions

[Dmitriy Poznyak <poznyack@yahoo.com>]

Hello everyone,


I am running a random slope model with cross-level interactions. All variables in the model are continuous. Context-level predictors are grand-mean centered. All other predictors are standardized on a scale 0-1. My context level predictors indeed happen to moderate the effect of the independent variables and the effects are in the expected direction. 

My set-up (much simplified) looks as follows:
xtmixed: y x z x*z||cluster: x
Note, that z is a context-level predictor

My goal is to compute the marginal effects for the change in x for each unit change in z in order to visualize my interactions. I am using the following magins command:

margins, at(z =(-1 0 1) x=(0 .5 1))  

This however--I believe--computes the probabilities for the dependent variable, given the change in both indep. variable and context-level moderator. While it is illustrative, it does not really tell me about the actual moderation effect of z on x. Again, I need to see the marginal effect (the mean effect will suffice) of my indep. variable on the dependent var. conditional on the change in the macro-level predictor. I assume, I should be able to plot this effect using margins command, but the exact model specification escapes me. 

Thank you for your inputs,
Dmitriy

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/