If you are unsure about such issues, just check in Stata
******************
sysuse auto, clear
qui probit foreign weight mpg
mfx
qui reg foreign weight mpg
mfx
******************
and see that the marginal effects are not equal, and neither are the
p-values... (mpg is significant at alpha=5% in the -probit- vs. -regress-).
Note also that -mfx- is redundant for -regress- as the linear model has
constant marginal effects whereas -probit- requires you to specify at which
value of the covariates it should evaluate the marginal effects. The default
for -mfx- is the mean of the regressors (see -h-mfx-).
There is always a reason why Stata does not offer something, such as -resid-
for -predict- after -probit-. That should make you think twice before
constructing things like
******************
sysuse auto, clear
qui probit foreign weight mpg
predict pred
g resid= pred -for
******************
HTH
Martin
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Nirina F
Sent: Sunday, August 24, 2008 10:23 PM
To: [email protected]
Subject: st: probit vs. reg postestimation
Hello,
My dependent variable is a binary variable. Onn the following example
y=b1x1+b2x2+b3x3+e,
if I use reg to estimate b's , I should get more or less the same
results as the marginal effects after probit.
Now after reg, I can use predict e1, resid but after probit how can I
get the same thing as e1?
and then if I want to do 2sls, ivreg2 won't be as ivprobit right?
Thank you very much in advance for your help.
Nir
*
* 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/
*
* 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/
