Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Generalised Probit Residual

From   "Michaud, Pierre-Carl" <>
To   <>
Subject   st: RE: Generalised Probit Residual
Date   Wed, 16 Nov 2005 11:37:27 -0800

Dear Georgios, there is a paper by Vella and Verbeek in Journal of
Econometrics (I think 1997) that constructs generalized residuals in the
case of a panel data random effect model. It is for a different
problematic than yours (theirs is for a sample selection model) but I
think you will get the insight there.
Best, Pierre-Carl.

Pierre-Carl Michaud
RAND, 1776 Main St
P.O. 2138
Santa Monica
CA 90407-2138
tel:(310) 393-0411 x7620
fax:(310) 574-3950
-----Original Message-----
[] On Behalf Of
Sent: Wednesday, November 16, 2005 11:32 AM
Subject: st: Generalised Probit Residual

Dear Statalisters,
I want to extract the generalised residuals from the simple probit model
use them as additional regressors in a second stage regression--i.e
implementing a variant of Heckman's (1981) solution to the initial
conditions problem.
The generalised residual for the Probit model as defined by Gourieroux
et al
(1987) is:
u= {phi(xb).[y-PHI(xb)]}/{PHI(xb).[1-PHI(XB)]}
The way I am thinking of extracting the generalised probit residuals is
predict xb, xb
gen norm(xb)=normxb --i.e the cumulative standard normal distr
gen normalden(xb)=normaldenxb -- i.e the standard normal desnity
gen denominator= norxb*[1-normxb]
gen numerator= normaldenxb*[y-normxb] --where y is my binary dependent
Is this right?
Then if I want to extract the error term after estimating a model using
xtprobit (random effects probit) do I follow the same procedure or is
completely wrong?
Would be most grateful if anyone could help.


*   For searches and help try:


This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index