Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Testing weak instruments agter reg3


From   Ekrem Kalkan <kalkan.ekrem@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Testing weak instruments agter reg3
Date   Fri, 4 Dec 2009 09:33:19 +0200

Thank you very much for your comment.

Ekrem.

2009/12/3 Austin Nichols <austinnichols@gmail.com>:
> Ekrem Kalkan <kalkan.ekrem@gmail.com> :
> The advice on a first-stage F greater than 10 is about 7 years out of
> date now (Stock and Yogo overturned in 2002 that advice from Staiger
> and Stock from 1997, more or less), and applies only to a single
> endogenous regressor, in any case. -ivreg2- (from SSC) will give you
> relevant diagnostics (read the help file) if you estimate one equation
> at a time; if you have a recursive system of equations, these will be
> very good measures of the strength of your excluded instruments.  If
> not, they will be imperfect, but better than nothing.
>
> On Thu, Dec 3, 2009 at 3:40 AM, Ekrem Kalkan <kalkan.ekrem@gmail.com> wrote:
>> Dear Stata users,
>>
>> I am estimating a 3SLS regression (for an Almost Ideal demand system)
>> with reg3 command. I do not know how to test whether my "excluded
>> instruments" are weak or not (correlation with endogenous regressors).
>> I know that with "ivreg2" there are some diagnostics for weak
>> instruments. For example, the partial F-statistic of the reduced form
>> regression is required to be higher than 10.  Do you think that  these
>> diagnostics are valid for system estimation like reg3?
>
> *
> *   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index