moderated mediation model

[Rachel Chen]

Dear Professor,
I'm doing mediated effects now, and in the model, my mediated variables are two categorical variables. Since M+-1SD only makes sense for continuous variables, what should I do if my M is 0.
Thank you all.

[Erik Ruzek]

Is your moderator continuous? If so, you could plot a separate slope for each of the groups comprising the mediators. Here is a much simpler example:

sysuse nlsw88.dta, clear
reg wage i.married##c.age 
margins married, at(age=(35(1)44))
marginsplot

[Rachel Chen]

Thank you for your answer. But my moderating variable is also a categorical variable. Then, what have I done？Thank you very much

[Erik Ruzek]

You can still use margins. Play with the marginsplot command to get different types of graphs. See this Stata youtube video on margins with categorical by categorical interactions. See also here.