I would appreciate instruction how to generate the marginal effects in a gsem mediation model such as:

gsem (social -> behavior,) (social -> health, family(binomial) link(probit)) (behavior -> health, family(binomial) link(probit)

In particular, I would like to generate the marginals for the indirect effect of social, mediated by behavior, on health.

I have thus far generated the overall or average indirect effect, standard error, and confidence intervals with

nlcom _b[health:behavior]*_b[behavior:social], post

I wish to generate the marginal indirect effects for several points over the range of social for subsequent plotting of these in relation to health.