st: inteff: calulations by hand

[Florian Seliger <florian.seliger@gmx.de>]

Dear Statalist,

I have a question concerning the calculation of the marginal effect as done by "inteff".

I tried to calculate the marginal effect by hand and got other results compared to the results obtained with inteff.

let's assume I want to estimate the following model:

probit mneu im agree imXagree $xlist

where imXagree is the interaction effect.

The specification with inteff is:

inteff mneu im agree imXagree $xlist

Now, I tried to do the following:

******
predict xb, xb
gen me=normalden(xb)*(-xb)*(_b[im]+_b[imXagree]*agree]*(_b[agree]+_b[imXagree]*im])+normalden(xb)*_b[imXagree]

sum me
*****

Shouldn't I get the same mean as with inteff (I did not!), have I understood something wrong or made a mistake?


Thank you!
Best,
Florian



*
*   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/