st: Marginal effect: endogenous binary treatment with a binary outcome

[Guy Grossman <guygrossman1@gmail.com>]

Dear Stata list members,

I am fitting a model in which the outcome measured at the individual
level, is binary. The outcome is a function of an endogenous treatment
indicator (buy) assigned at a group level and a set of controls. The
treatment (buy) is a function of exogenous instrumental variable
(encourage), again assigned at the group level. I am considering using
bioprobit or cmp to fit the model. see example below. Both ioprobit or
cmp are user-written programs.

sysuse auto.dta
gen buy=round(uniform(),1)
gen encourage=round(uniform(),1)
gen outcome=round(uniform(),1)
xtile mpgcat=mpg, nq(5)

bioprobit (outcome=price mpg rep78 trunk weight buy) (buy= encourage
foreign) vce(cluster mpgcat)
cmp (outcome=price mpg rep78 trunk weight buy) (buy= encourage
foreign), ind($cmp_probit $cmp_probit) vce(cluster mpgcat)

my question is how to calculate the marginal effect of the exogenous
variable (encourage) on the probability that the outcome ==1, through
its affect on the treatment (buy)?
Thanks,
Guy
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/