Friday, February 03, 2006

st: RE: Re: Binary endogenous variable in a panel data set


> -----Original Message----- > From: > [] On Behalf Of > jyotsna puri > Sent: 03 February 2006 13:01 > To: > Subject: st: Re: Binary endogenous variable in a panel data set > > Hi all, > > I have an unbalanced panel dataset. > My dependent variable of interest is continuous but there is > also an endogenous binary variable - which I must instrument for. > I have used xtivreg to estimate this model. However xtivreg > does not account for the binary nature of the endogenous > variable (my understanding is that it assumes the endogenous > variable to be continuous).

This isn't the case. It's the same as generic IV - you don't need to make this assumption for the standard properties of IV to carry through. If you are looking for increased efficiency, though, you could consider a variation on a procedure recommended by Wooldridge in his 2002 book, which is similar to what you propose below except that you use the generated variable as an _instrument_ for the endogenous regressor. That way, you don't have to bootstrap.



> > Can some one advise me on any utility/route to deal with this? > > One possibility is to do a 2 step routine: Estimate the > equation for the endogenous binary variable using xtprobit. > Take the predicted values, and use it in the second stage > equation and bootstrap the standard errors for the second stage. > > Has any one done this and can you advise me on the possible pitfalls? > > Best, > -Jo > > > > Jyotsna (Jo) Puri > > I am supporting the education of underprivileged kids. You can too: > > hp?runnerid=2005TA486 > > > * > * For searches and help try: > * > * > * > >

* * For searches and help try: * * *


<< Home

This page is powered by Blogger. Isn't yours?