Using margins command after medeff

[Hana Ashida]

Hi all,

I used medeff comand to see mediation effects. Because my outcome variable and mediator variable are binary, I used probit model in the medeff as below.

medeff (regress M T X1 X2 X3)(probit Y M T X1 X2 X3), mediate(M) treat(T) vce(cluster X1)
margins, dydx (M T)

*X1, X2 and X3 are control varables.

Then, I got the following result. (edyrtotal=T, and wealthq=M)

How should I interpret this result?
Is my understanding correct that the dy/dx for T (edyrtotal) is direct effect and M (wealthq) is the indirect effect?

Any help on the topic is appreciated!

Thank you in advance.
Regards,

[Felix Bittmann]

Are you sure you should use margins after medeff? See https://journals.sagepub.com/doi/pdf...867X1201100407
The command itself provides the relevant statistics for you, like ACME and so on.

[Hana Ashida]

Thank you, I was misunderstanding.