Marginal effects of interaction terms after probit using inteff

[Olivia Johns]

I am investigating the effect of X (a dummy variable 1 or 0) on Y but I want to interact X with credit score groups.

Let's say I am running a probit reg

probit Y X##i.Creditgroup, vce(cluster group)

I want to eventually see how X affects Y differently across the credit groups. I know that computing marginal effects of interaction terms using margins is incorrect. So following Norton et al, I am trying to use "inteff".

So I converted my variables so that I don't have factor variables since inteff can't handle factor variables.

so assume there are 4 credit groups and X is a binary, so i have now,

gen x_credit4 = (X== 1 & creditgroup==4)
gen x_credit3 = (X== 1 & creditgroup==3)
gen x_credit2 = (X== 1 & creditgroup==2)
gen x_credit1 = (X== 1 & creditgroup==1)

gen credit4 = (creditgroup==4)
gen credit3 = (creditgroup==3)
gen credit2 = (creditgroup==2)
gen credit1 = (creditgroup==1)

probit Y X x_credit4 x_credit3 x_credit2 x_credit1, vce(cluster zipcode_year)
inteff Y X x_credit4 x_credit3 x_credit2 x_credit1

and I get an "r(9);" error!

Any help will be greatly appreciated. (any other methods to potentially compute the marginal effects)

[Jeff Pitblado (StataCorp)]

You can use margins to compute marginal effects of one variable within groups of another variable.
Here is how you would do that using your example.

probit Y X##i.Creditgroup, vce(cluster group)
margins r.X, within(Creditgroup)