Dear Statalisters,
I understand that in a two step IVPROBIT, the first stage is OLS estimation.
While my question is that I have binary endogenous variables in rhs
which includes 3 choices. In order to fit ivprobit estimation, may I
do the following regression?
1. In the first stage, run 3 linear probability regressions on the 3 choices;
2. Get the preditted probabilities of each of the choices;
3. Take the predictted probabilities as instruments in second stage
ivprobit estimation.
The above is my first question, my second question is that when I do
IVREG and IVPROBIT, after I get the predictted values in the first
stage, I found that people use two stata commands for second stage, if
the predictted value is "zhat", the two commands in second stage are
as followings:
a. ivreg y x (z=zhat)
b. reg y x zhat
Am I right or wrong about this? I personally believe that "a" is right.
My thrid question is that if I have multiple endogenous regressors
(both are binary varibles), I know this is very diffficult to solve,
can I do it in the following way: Generate interaction terms of the
two endogenous binary variables, thus, I have a series of dummies and
take care of them following the traditional two step method.
BTW, I am using STATA 9 SE.
Thank you very much!
Song
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/