Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# Re: st: Re: svy:ivreg with binary outcome variables

 From Austin Nichols To statalist@hsphsun2.harvard.edu Subject Re: st: Re: svy:ivreg with binary outcome variables Date Wed, 10 Mar 2010 18:15:50 -0500

```Nisha Malhotra <nishamalhot@gmail.com>:
This question is very hard to understand due to typos and notation: do
you have variables d and Z?  Are you actually using the out-of-date
command -ivreg- rather than -ivreg2- (SSC) or -ivregress-?  Are you
discussing a coef from the first stage regression?  Please post an
example using data that is publicly available, with code starting e.g.
webuse nhanes2
svy:ivregress 2sls houssiz (highbp=age)
egen c=group(strata psu)
ivreg2 houssiz (highbp=age) [pw=finalwgt], cl(c)

On Wed, Mar 10, 2010 at 3:58 PM, Nisha Malhotra <nishamalhot@gmail.com> wrote:
> Dear Statalisters and life savers,
>
> This the regression we are carrying out
>
> C=a+bD+dZ
>
> Where D is a dichotomour variable and C is a categorial variables (0-8). D
> is endogenouse and is instrumented with (d)
>
>
> This is for a multistge survey design (Demographic Health Survey).
>
> Given that C is categorical variable - we use linear probability model in
> the first dtage despite having a binary variable D - that is we use IVREG
>
> The problem we face is as follows.
>
> If we run
> svy:Ivreg C (D=d) Z
>
> the coefficient on d (9.7) lies outside the categorical range of C. (0-8).
>
> increasing d from 0 to 1 increases C by 9.7 ???
>
> But, in case we ignore issues with the survey design and run a simple
>
> ivreg C (D=d) Z
>
> the coefficient on d lies between 0-8.
>
> Any solutions!
>
> Thanks a ton
>
> Nisha
*
*   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/
```