Computing marginal effect for interaction terms after xtprobit

[Anny Yu]

I'm implementing a dynamic correlated random effects probit model with the following command:
xtprobit employment i.lagged##(c.edu i.origin i.female c.age i.hhtype i.initialstatus c.hhtype_bar), i(id)
How can I compute marginal effect of a fully interacted model on Stata, as I know margins would not give the marginal effect (dydx) for interaction terms?

Many, many thanks indeed!

[Clyde Schechter]

The reason why -margins, dydx()- does not calculate marginal effects for interaction terms is because they don't exist. So you shouldn't try to compute them.

[Phil Bromiley]

Let me expand a little on Clyde's comment. With interactions, the influence of any of the interacting variables depends on the values of the other variables. It makes sense to talk about the marginal effects for the original variables at differing values of the interacting variables and this is what Stata will let you do. You need to interpret both main and interaction together to have anything that is sensible (except that the main is the influence of the main variable if the interacting variable equals 0). The margin of an interaction means nothing since the interaction can't change without the original variables changing.

[Saad Baloch]

Hello. I have a quick question? I estimated a static random effect probit model using xtprobit in > > > stata 14. > >after the computation, I have the coefficients of all included variables. after that I tried to estimate > > > Marginal effects using margins, dydx(*) predict(pu0). but unfortunately I am not getting any value. you can see below: ------------------------------------------------------------------------------ | Delta-method | dy/dx Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- degofpc | 0 (omitted) csr | 0 (omitted) strflex | 0 (omitted) degpccsr | 0 (omitted) ------------------------------------------------------------------------------ Can anyone guide me what could be the possible problem?