I have a question that to my knowledge isn't addressed yet on this forum.

I am running a probit regression with an interaction term containing a continuous variable. I have standardized this variable in such a way that the unit is the standard deviation increase.

Say y is my standardized continuous variable, then when calculating the marginal effects, I am using code of the form:

Code:
margins, dydx(x) at(y =(-1 0 1))
My question is how to go about figuring out which values of y to take when calculating the marginal effects, when there are so many possible values (e.g. I could look at the change from -.5 to .5 or -1 to 0 or 0 to 1, etc). Is this just a matter of subject matter knowledge, i.e. looking at what makes most sense given the research aims and the meaning of the variable? Or is there a more standard way to approach this?

To my knowledge, articles often report a single value for 'a standard deviation change' when this could mean many things.

Hopefully my question makes sense. Thanks!