logistic "interactions"

clear all
set seed 123

* recruit subjects (if only this easy...)
set obs 2000

* assign to groups
gen mbct =  mod(_n, 2)

* depression baseline score
gen d0 = rnormal(0,1)

* generate probability of relapse note - no interaction term here
gen p = 1 - min(.5*d0 + .3*mbct, 1) 
gen y = rbinomial(1,p)

* ns interaction term in the logistic regression
logistic y mbct##c.d0

* but get ES estimates in probability domainfo 1 SD above and below mean baseline 
margins, dydx(mbct) at(d0=(-1(1)1)) 


------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
1.mbct       |
         _at |
          1  |   .0162185   .0072951     2.22   0.026     .0019204    .0305166
          2  |   .1730282   .0282415     6.13   0.000     .1176759    .2283805
          3  |   .2800077   .0394602     7.10   0.000     .2026671    .3573483
------------------------------------------------------------------------------
Note: dy/dx for factor levels is the discrete change from the base level.

.