Dear all,
I am interested in estimating a multinomial logit model with a
continuous endogenous variable. As far as I know, there is no command in
Stata which can deal with this.
The dependent variable of the multinomial logit is z (three classes) and
the endogenous variable (y) is continuous.
One way to do this is employ a two-stage multinomial logit where the
first stage estimates y through OLS and the second stage uses the
predicted values of y as instruments.
Since the second-stage estimation should be corrected for the inclusion
of a predicted value as a regressor, I thought to implement this
correction via boostrapping. My question is: do I need to compute the
two-step corrected variance-covariance matrix by hand or the boostrap
command can automatically deal with this?
Below is a sketch of the program I've thought about, but I am not sure
whether I need to make some further corrections for the standard errors.
Thanks in advance for your help,
Chiara
program myprog
// Stage 1
regress y x1 x2 x3 z1 z2
predict double yhat, xb
// Stage 2
mlogit z x1 x2 x3 yhat
end
bootstrap "myprog" _b _se, reps(1000) dots
--
[email protected]
--
http://www.fastmail.fm - Access your email from home and the web
*
* 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/
