Re: st: random effects logit instrumental variables

[Roger Harbord <roger.harbord@bristol.ac.uk>]

Don't think there's any command that will do that directly in Stata. I'd 
rather doubt there is in any software package. I guess you could write a 
program to do the two stages ( i.e. regress endogenous on instruments and 
exogenous vars, predict endogenous, regress dependent var on exogeneous vars 
plus predicted endogenous) and do a clustered bootstrap to get the right 
standard errors? Would needs a bit of programming expertise, and a reasonably 
large number of clusters for validity of the bootstrap.

Any IV experts care to comment?

I'm considering doing something similar as i have a candidate for IV in which 
_both_ equations are generalised linear models (and hence beyond even -qvf- i 
think). I don't have clustered data though.

Roger.
----------------------------------------------------
Roger Harbord     mailto:roger.harbord@bristol.ac.uk
Department of Social Medicine, University of Bristol

--On 09 August 2004 13:27 +0200 esicuri@unipv.it wrote:


> I'm trying to estimate a logit model (my dependent variable is a binary
> choice) with random effects (I'm using xtlogit, re).
>
> One of my regressors (it's a proportion, a continuos variable in the 0-1
> interval) is endogenous and I have instruments for it.
>
> How can I use my instruments for the endogenous variable in my logit model
> with random effects?
>
> Thanks
>
> Elisa
*
*   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/