Update Aug 2020 to my original post below: This seems like it might do the trick - An Introduction to ‘margins’
...
Hey! there are some great vignettes out there to help interpret interactive effects from "glm" or "lm" models. (see e.g., interplot, plot_model). But these dont seem to work with the output produced by the "mlogit" package, which I am using to analyze choice experiment data.
So my question: Is there a “parallel” to interplot for mlogit? Any tips out there from other choice experiment modelers?
I wish to do things such as:
-- generate relevant graphs that plot model predictions for relevant values of (interacted) independent variables,
-- use marginal effects and tests of second differences* at specific / relevant values of (interacted) independent variables to determine whether an interaction effect exists
(* i.e., “first difference” is testing statistical diff for Man vs. Women (gender); “second differences” might be testing whether “age”, together with gender, matters for a particular group.)
Thx
Scott