<>
webuse auto,clear
probit foreign price mpg, nolog
ivprobit foreign price (mpg = weight displacement gear_ratio), nolog
As you can see, ivprobit provides a test of endogeneity (the null is
exogeneity) of the regressors that are instrumented. But if you have
some suspect regressors and no instruments, how can you contemplate
estimating an ivprobit model? You can't test for IV vs non-IV
solutions without a specified IV model.
Kit Baum | Boston College Economics & DIW Berlin | http://ideas.repec.org/e/pba1.html
An Introduction to Stata Programming
| http://www.stata-press.com/books/isp.html
An Introduction to Modern Econometrics Using Stata | http://www.stata-press.com/books/imeus.html
On May 9, 2009, at 02:33 , Sergio wrote:
I have several doubts about Hausman and Smith-Blundell test for
exogeneity
(probexog). I?m estimating a Probit model and I suspect that one
of the
explanatory variables is not exogenous. Then, don?t know how can I
do test for
endogeneity:
1. Is Hausman Test an appropriate test for endogeneity in probit
regressions,
how can I perform It to compare two probit specifications? I?ve seen
some
examples of Hausman test but always comparing probit vs others
estimation
methods (i.e. IVprobit)
2. I try to run Smith-Blundell test, but reading some threads
(http://www.stata.com/statalist/archive/2006-04/msg00574.html) and
stata
commands probexog , I can see that some instruments are required
for it, and
I don?t have any variable for this purpose.
3. I need at least to conclude that suspect variable is endogenous,
and there is
not data at hand to estimate IVprobit, but can I do this if I can´t
perform
Smith-Blundell test?
*
* 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/
