# Re: st: Q on reg3 output for model fit

 From John Antonakis To statalist@hsphsun2.harvard.edu Subject Re: st: Q on reg3 output for model fit Date Sun, 10 Feb 2013 18:55:31 +0100

```You might not have the right version of overid.

Typing -which overid- give me:

. which overid
*! overid V2.0.7 05feb2010
*! Authors C F Baum, Vince Wiggins, Steve Stillman, Mark Schaffer

Typing -help overid- gives the following in the relevant section:

-------------------------------------------------------------------------------------------------
help for overid
-------------------------------------------------------------------------------------------------

```
Calculate tests of overidentifying restrictions after ivreg, ivreg2, ivprobit, ivtobit, reg3
```
[snip]

```
----+ Three-stage least squares +----------------------------------------------------------
```
```
overid will report an overidentification statistic after system estimation with reg3. As Davidson and MacKinnon (2004, p.532) indicate, a Hansen-Sargan test of the overidentifying restrictions is based on the 3SLS criterion function evaluated at the 3SLS point and interval parameter estimates. Under the null hypothesis, the statistic is distributed Chi-squared wih (G*L - K) degrees of freedom, where G is the number of simultaneous equations. The procedure will take proper account of linear constraints on the parameter vector imposed during estimation.
```
```
So, either you don't have the right version or you have done something wrong in setting up reg3.
```
HTH,
J.

__________________________________________

John Antonakis
Professor of Organizational Behavior
Director, Ph.D. Program in Management

University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
__________________________________________

On 10.02.2013 18:47, Jason R Franken wrote:
> Thanks John,
>
```
> By the way, when I tried -overid- after 3SLS, STATA returned the following error message
```> "overid works only after ivreg, ivreg2; use overidxt after xtivreg" ...
>
> Jason
>
>
> ----- Original Message -----
> From: "John Antonakis" <John.Antonakis@unil.ch>
> To: statalist@hsphsun2.harvard.edu
> Sent: Sunday, February 10, 2013 11:22:41 AM
> Subject: Re: st: Q on reg3 output for model fit
>
> OK...that chi-square is just a Wald test that the predictors
> simultaneously = 0. It is not an overidentification statistic. You
> should run that statistic in any case if you are overidentified by using
> the userwritten command -overid- (available from SSC) to ensure that the
> model constraints are tenable.  After running reg3 just type -overid-.
> Although a negative r-square is not necessarily a bad thing, it could
```
> indicate misspecification, and the overidentification stat will catch that.
```>
> Best,
> J.
>
> P.S. it is always good and saves everyone time if you post your output
> so we are all on the same page.
>
> __________________________________________
>
> John Antonakis
> Professor of Organizational Behavior
> Director, Ph.D. Program in Management
>
> Faculty of Business and Economics
> University of Lausanne
> Internef #618
> CH-1015 Lausanne-Dorigny
> Switzerland
> Tel ++41 (0)21 692-3438
> Fax ++41 (0)21 692-3305
> http://www.hec.unil.ch/people/jantonakis
>
> Associate Editor
> __________________________________________
>
> On 10.02.2013 17:58, Jason R Franken wrote:
```
>> The chi-square statistic I'm talking about is posted to the right of R-square in 3SLS output from reg3 (or an F-statistic in the 2SLS version). I'm in doubt that it is the identification statistic described in the prior response to this post, since it is significant in text book examples with positive R-square at: http://www.ats.ucla.edu/stat/stata/examples/greene/greene16.htm.
```>>
```
>> I thought this chi-square statistic is for the test of whether the model is no better better predictor of the dependent variable than its mean, but the negative R-squares I observe can occur when "ybar is a better predictor of y (in the sum-of-squares sense) than Xb" according to http://www.stata.com/support/faqs/statistics/two-stage-least-squares/. Hence, I shouldn't see a negative r-square when the chi-square test statistic rejects the null that ybar is as good as my model.
```>>
>> I'm not sure what the chi-square statistic is testing ...
>>
>> Thanks again.
>>
>> Jason
>>
>> ----- Original Message -----
>> From: "John Antonakis" <John.Antonakis@unil.ch>
>> To: statalist@hsphsun2.harvard.edu
>> Sent: Sunday, February 10, 2013 2:26:58 AM
>> Subject: Re: st: Q on reg3 output for model fit
>>
>> I would worry more about the chi-square statistic of fit (i.e.,
>> overidentification statistic, which is what I assume you are talking
>> about). If that is significant, the model is misspecified and an
>> unlikely candidate to have generated the data. So, I would rethink the
>> model.
>>
>> Best,
>> J.
>>
>> __________________________________________
>>
>> John Antonakis
>> Professor of Organizational Behavior
>> Director, Ph.D. Program in Management
>>
>> Faculty of Business and Economics
>> University of Lausanne
>> Internef #618
>> CH-1015 Lausanne-Dorigny
>> Switzerland
>> Tel ++41 (0)21 692-3438
>> Fax ++41 (0)21 692-3305
>> http://www.hec.unil.ch/people/jantonakis
>>
>> Associate Editor
>> __________________________________________
>>
>> On 10.02.2013 07:17, Jason R Franken wrote:
```
>>> I'm running 3SLS using reg3 and get negative R-square (consistent with the explanation at http://www.stata.com/support/faqs/statistics/two-stage-least-squares/). As I understand, this can happen when the mean of the dependent variable is a better predictor than the model. However, the p-values of the chi-square statistic are significant. What if anything can I determine about model fit from the R-square and chi-square statistics in the output of reg3?
```>>>
>>> Thank you in advance for any insight you can provide.
>>>
>>> Jason Franken
>>>
>>>
>>> Jason Franken
>>> Assistant Professor
>>> School of Agriculture
>>> Western Illinois University
>>>
>>> *
>>> *   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/
>> *
>> *   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/
>> *
>> *   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/
>
> *
> *   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/
> *
> *   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/

*
*   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/
```