conditional indirect effects for interaction terms

[Hale Isaac]

I am running the following regression and I am trying to compute the conditional indirect effects of X on L for three levels of my moderator variable (the interaction terms of x*L in the regression, where L is the moderator):

gsem (y <- L i.x i.x#c.L i.ethnic, mlogit)

1. Mean – 1 standard deviation
2. Mean
3. Mean + 1 standard deviation

Is there anyway to do this in stata?

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(y x L)
. .  .
0 0 13
. .  .
0 0  3
. .  .
3 0 16
. .  .
0 0 11
. .  .
0 1 15
. .  .
0 0  2
. .  .
2 0 16
. .  .
0 0 11
. .  .
0 0 16
. .  .
0 0 16
. .  .
0 0 13
. .  .
0 0 16
. .  .
0 0 11
. .  .
0 0 12
. .  .
1 0 11
. .  .
0 0 14
. .  .
3 1 15
. .  .
0 0 11
. .  .
1 0  3
. .  .
0 0 11
. .  .
0 1  5
. .  .
0 1 15
. .  .
0 0 11
. .  .
0 0  5
. .  .
1 0  3
. .  .
2 1 16
. .  .
1 0 14
. .  .
0 1  5
. .  .
1 0  8
. .  .
0 0  8
. .  .
1 0 15
. .  .
0 0 12
. .  .
0 0 11
. .  .
3 0 16
. .  .
1 0 11
. .  .
0 0 12
. .  .
0 0 11
. .  .
0 0 11
. .  .
0 0 11
. .  .
0 0 11
. .  .
0 0  0
. .  .
0 0 16
. .  .
0 0 14
. .  .
0 0 11
. .  .
0 0 11
. .  .
0 1 13
. .  .
3 1 13
. .  .
3 0 16
. .  .
0 0 12
. .  .
0 0 11
end

This paper explains how do this in SPSS but not sure how to replicate it in Stata.

Borau, S., El Akremi, A., Elgaaied-Gambier, L., Hamdi-Kidar, L., & Ranchoux, C. (2015). Analysing moderated mediation effects: Marketing applications. Recherche et Applications en Marketing (English Edition), 30(4), 88-128.

[Phil Bromiley]

First, I don't see that you have mediation – to have mediation you need more than one equation. If you are simply using this to do a standard multinomial logit, then I would go to the mlogit command. I suspect the post estimation discussion includes some illustration of how to do predictive values. The predicted values or effects are done using the margins command.

[Hale Isaac]

The full moderated mediation model that I am using is:

gsem (m <- i.x L i.x#c.L i.ethnic) (y <- L i.x i.x#c.L i.ethnic, mlogit)

I have graphed the interactions using the margins command but I am not sure if I can calculate the conditional indirect effects at high and low levels of my moderator using margins. I am simply trying to replicate what is done here: https://stats.idre.ucla.edu/stata/fa...tion-in-stata/ using nlcom