i have two questions about average marginal effects. any
thoughts/suggestions would be very much appreciated.
1) does anybody here know how to compute average marginal effects & their
standard errors for ivprobit estimates? mfx appears to only present
marginal effects for one observation at a time (i.e., not the average of
_N different marginal effects). and the user-created margeff does not
support ivprobit.
2) is there an easy way to compute average marginal effects and their
standard errors for nonlinear functions of regressors? for instance, i'm
running an ivprobit of y on x (using some instrument z) and i'd like to
estimate the average of dy/d(x^2) for my _N observations & get the
appropriate standard error.
