Re: st: 2sls with discrete endogenous regressor

[Austin Nichols <austinnichols@gmail.com>]

Megan Stevenson <m_stevenson@berkeley.edu>:
Did you follow the link? It describes several models, including the
-biprobit- approach.
See also:
http://www.nber.org/papers/w15539
http://faculty.smu.edu/millimet/code.html

On Wed, Sep 11, 2013 at 3:02 AM, Megan Stevenson
<m_stevenson@berkeley.edu> wrote:
> David - You are correct that there is no reason to be concerned with
> reverse causality here.  However there is a large selection bias
> problem, thus the need for an instrument for incarceration.
>
> Austin - I am looking for a non-linear model since the data is heavily
> skewed and assuming a linear relationship between the covariates and
> the outcome is creating a distortion.  A linear model will of course
> be useful as a robustness check.
>
> Does anyone know of a 2sls package with discrete endogenous variables?
>  If such a thing doesn't exist I can write the code to calculate the
> standard errors myself but it seems like it is a common enough need
> that someone would have already built the tool.
>
> Thanks!
>
> On Tue, Sep 10, 2013 at 9:54 PM, Austin Nichols <austinnichols@gmail.com> wrote:
>> Megan Stevenson <m_stevenson@berkeley.edu>:
>> Start with -ivreg2- (SSC) and a linear model, but see also
>> http://www.stata.com/meeting/chicago11/materials/chi11_nichols.pdf
>>
>> On Tue, Sep 10, 2013 at 5:57 PM, Megan Stevenson
>> <m_stevenson@berkeley.edu> wrote:
>>> Is anyone familiar with a package that will allow you to do a 2sls
>>> regression where both the endogenous dependent variable and the main
>>> dependent variable are binary?  I am trying to identify the
>>> probability of future arrest on having been incarcerated, with an
>>> instrument for incarceration.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/