Hi,

I have used the Mundlak approach (correlated random effects) to regress my dependent variable income on race (my variable of interest) and several control factors. Also, I have included an interaction term of the categorical variable: religion and the continuous variable: number of dependent children.

For my OLS and random effects regressions, I used the command: margin religion ,dydx(dependentchildren) to interpret the interaction term. However, I am unsure whether this is the correct approach for the Mundlak method as there are also meaned and demeaned variables of dependent children.

I have attached my do-file below.

Would this be the correct method and if not how can this be done?

Many thanks