Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: RE: Fw: reverse causality and sample size

 From "Millimet, Daniel" <[email protected]> To "[email protected]" <[email protected]> Subject st: RE: Fw: reverse causality and sample size Date Wed, 26 Jan 2011 14:50:06 +0000

```It sounds like an irrelevant question because in the first-stage model for x3 (your model 2), you should include the other exogenous variables from the structural model (model 1).  So, x1 and x2 belong in model 2.  Thus, this becomes observation-specific rather than just an industry-level regression.

****************************************************
Daniel L. Millimet, Professor
Department of Economics
Box 0496
SMU
Dallas, TX 75275-0496
phone: 214.768.3269
fax: 214.768.1821
web: http://faculty.smu.edu/millimet
****************************************************

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Maarten buis
Sent: Wednesday, January 26, 2011 6:33 AM
To: stata list
Subject: st: Fw: reverse causality and sample size

--- On Wed, 26/1/11, Victoria Alvarez wrote:
> I was not able to send my message through the statalist.
> Can I ask you to send  it for me?

> Dear Statalist,
>
> I'm estimating a model specification where I suspect a
> variable to be endogenous  because of reverse-causality.
>
> The model - that I call model 1 - looks as follows
> Y = b0 + b1X1 + b2X2 + b3X3 + u
> Where X3 is the potential endogenous variable. Note that
> this model refers to a dataset of 1000 observations
> referring to 80 industries.
> Then I run a two stage  regression.
>
> At the first stage, I regress the following model - that
> I call Model 2-:
>
> X3 = a0 + a1Z1 + a2Z2 + a3Z3 + a4Z4 + v.
>
> Model 2 is applied using a dataset with 80 observations
> referring to the 80 industries.
>
> I use regress twice and compute the standard errors
> accounting for the inclusion of a predicted regressor in
> the second step. In the second step, I regress the
> Model 1 over the 1000 observations and substituting X3 by
> the predicted X3 in the first stage.
>
>
> Is it a problem if the two datasets that I use in the two
> regressions have a different size?
>
> I would appreciate any help in this matter.
> Best,
> Victoria

*
*   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/

*
*   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/
```