Margins for a logistic regression model with contemporaneous and lagged regressors.

[MET]

How should the margin instruction be written to account for the fact that the contemporaneous and lagged regressors are not independent?

[Clyde Schechter]

Perhaps I am missing something, but I don't see any need to do anything special here. In general, the various regressors in a model are typically not independent of each other, even when they are not related by a lag operator. The regression command itself has presumably dealt with this already by the time you get to running -margins-.

By the way, it is the norm in this community that we use our real first and last names as our usernames. It helps maintain a professional atmosphere. Please click on "CONTACT US" and send the forum operator a request to change the username on your account accordingly. Thank you in advance.

[MET]

Thank you Clyde, will do. I meant to have a situation such as X{t} = a + b*X{t-1} + u{t}. Is it correct to evaluate the marginal effect of X{t} at the average of X{t-1}?.

[Clyde Schechter]

Well, this is a somewhat different story because you have a regressor that is a lag of your dependent variable. So the problem is the potential for autoregressive error structure. If your regression model handles this appropriately, you can run -margins- without worry. -margins- itself doesn't deal with these issues: it just uses the information in the estimates from the regression itself.
Autoregressive error structure is not something that comes up much in my field, and I don't have the expertise to advise you. But I can assure you that once you get the regression right, you can just apply -margins- in the usual way.