Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Marginal effects after heckman

From   Maarten buis <>
Subject   Re: st: Marginal effects after heckman
Date   Mon, 27 Sep 2010 12:11:22 +0000 (GMT)

--- On Mon, 27/9/10, Fred Dzanku wrote:
> How can I get marginal effects of the (probit) selection
> equation after running a heckman selection model by maximum
> likelihood? I estimated a model in the form
> margins, dydx(*) atmeans predict(equation(select)) ///
> vce(unconditional)
> but what I get as marginal effects are exactly the same as
> the coefficients of the selection equation.

You asked for the marginal effects on the linear predictor (xb)
of the selection equation. This is ofcourse exactly the same as
the coefficients you already got. What you want are the marginal
effects on -normal(xb)-, i.e. the probability of being selected.
I don't see a prediction option that will directly give you that
(I looked in -help heckman postestimation-), but you can get it
by specifying the -expression()- option in -margins-, like so:

*---------------- begin example ------------------
webuse womenwk, clear

// the coefl option is only there so I know
// how the selection equation is called
// (in this case Stata chose a very sensible name)
heckman wage educ age, ///
        select(married children educ age) coefl

margins, dydx(*) expression(normal(xb(select)))
*----------------- end example --------------------
(For more on examples I sent to the Statalist see: )

Hope this helps,
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index