Bookmark and Share

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.


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

RE: RE: st: Valid instrument test for exactly identified regression


From   DE SOUZA Eric <eric.de_souza@coleurope.eu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   RE: RE: st: Valid instrument test for exactly identified regression
Date   Wed, 15 Jun 2011 18:02:40 +0200

Instruments are valid if they are (i) exogenous and (ii) relevant
The discussion here should make clear that you cannot test for exogeneity when you have exactly as many instruments as you have endogenous regressors.
Relevance means that (i) they are correlated with the endogenous regressors, (ii) the part of the instruments that are not correlated with the exogenous regressors is strongly correlated with the endogenous regressor (the instruments are strong). Relevance can be tested.  If you are using the command -ivreg2- written by Mark Schaffer and others, the tests are produced automatically and further details can be obtained by using the options -first- or ffirst-.  

If your instruments are weak then you have a problem. The latest (third) edition of Stock and Watson on Econometrics has a good elementary discussion of the problem and ways round it.

In any case you should read the help file for ivreg2


Eric de Souza
College of Europe
Brugge (Bruges), Belgium
http://www.coleurope.eu



-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of etanebay@yahoo.com
Sent: 15 June 2011 17:42
To: statalist@hsphsun2.harvard.edu
Subject: Re: RE: st: Valid instrument test for exactly identified regression

Hi, thanks for the discussion. 

I understand Hansen J cannot be used, since I have 1 IV for 1 endog. 

However, even though more IVs are good, for now I need to show that my single IV is valid.  (I checked numerous published papers with exact id, but no one seems to bother showing their one IV is valid). 
So I need to show that the IV doesn't corr with the u_it from 2nd stage. 

I have done what Justina said, which to me made intuitive sense in checking if IV corr with u_it. The coefficient is not sig at 90%. 

I then did the direct thing, by predicting the u_it (using steps from this board) and then pwcorr u_it and IV. The 2 are not correlated. I'm not an econometrician (obviously!), so was wondering if you all think these steps are problematic, now that there isn't a direct test I can do. 

Thanks - Eric





Sent from my BlackBerry Wireless Handheld Powered by Gee! from StarHub

-----Original Message-----
From: "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
Sender: owner-statalist@hsphsun2.harvard.edu
Date: Wed, 15 Jun 2011 14:32:52
To: <statalist@hsphsun2.harvard.edu>
Reply-To: statalist@hsphsun2.harvard.eduSubject: RE: RE: st: Valid instrument test for exactly identified regression

Justina,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Justina 
> Fischer
> Sent: 15 June 2011 14:19
> To: statalist@hsphsun2.harvard.edu; statalist@hsphsun2.harvard.edu
> Subject: Re: RE: st: Valid instrument test for exactly identified 
> regression
> 
> well, validity implies that the exclusion restriction is satisfied. 
> At least I learned this in grad school...

True.  But it's not testable if the equation is just-identified.  You only have degrees of freedom available for testing if it's overidentified (hence the name of the test - Eric's point).

> Nevertheless, he needs more and better instruments.

Indeed!

--Mark

> JF
> -------- Original-Nachricht --------
> > Datum: Wed, 15 Jun 2011 14:10:22 +0100
> > Von: "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
> > An: statalist@hsphsun2.harvard.edu
> > Betreff: RE: st: Valid instrument test for exactly identified 
> > regression
> 
> > Justina,
> > 
> > I don't think the test you proposed makes sense, to be
> honest.  (Maybe
> > you had in mind a test of the exogeneity of the endogenous
> regressor?)
> > But your conclusion - find a second instrument - is sensible.
> > 
> > --Mark
> > 
> > > -----Original Message-----
> > > From: owner-statalist@hsphsun2.harvard.edu
> > > [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf
> Of Justina
> > > Fischer
> > > Sent: 15 June 2011 13:56
> > > To: statalist@hsphsun2.harvard.edu
> > > Subject: Re: st: Valid instrument test for exactly identified 
> > > regression
> > > 
> > > so it  was significant at the 10% level ? That's no good.
> > > Try to find a second instrument for decent testing (e.g.a
> quadratic
> > > term of your first instrument) .
> > > JF
> > > 
> > > -------- Original-Nachricht --------
> > > > Datum: Wed, 15 Jun 2011 12:45:23 +0000
> > > > Von: etanebay@yahoo.com
> > > > An: statalist@hsphsun2.harvard.edu
> > > > Betreff: Re: st: Valid instrument test for exactly identified 
> > > > regression
> > > 
> > > > Hi thanks, I've already tried that informally and the IV wasn't 
> > > > significant at 95%. But is there a Hansen J-like test I can
> > > do that is more formal?
> > > > Thanks. 
> > > > 
> > > > 
> > > > Sent from my BlackBerry Wireless Handheld Powered by Gee! 
> > > from StarHub
> > > > 
> > > > -----Original Message-----
> > > > From: "Justina Fischer" <JAVFischer@gmx.de>
> > > > Sender: owner-statalist@hsphsun2.harvard.edu
> > > > Date: Wed, 15 Jun 2011 13:34:03
> > > > To: <statalist@hsphsun2.harvard.edu>
> > > > Reply-To: statalist@hsphsun2.harvard.eduSubject: Re: st: Valid 
> > > > instrument test for exactly identified regression
> > > > 
> > > > well, going back to your econometric textbook you could
> > > test whether
> > > > the instrument is significant when added to the main
> > > regression (exclusion
> > > > restriction)   - it should not be....
> > > > 
> > > > Justina Fischer
> > > > -------- Original-Nachricht --------
> > > > > Datum: Wed, 15 Jun 2011 10:10:32 +0000
> > > > > Von: etanebay@yahoo.com
> > > > > An: statalist@hsphsun2.harvard.edu
> > > > > Betreff: st: Valid instrument test for exactly identified
> > > regression
> > > > 
> > > > > Hi all,
> > > > > I have a model that is exactly identified, so the
> > > xtivreg2 command
> > > > > gives me a zero for the Hansen J statistic.
> > > > > Can you please advise: how do I test the validity of the
> > > IV, that it
> > > > > doesn't correlate with the errors in the structural equation?
> > > > > I know the IV is relevant from the first stage (1st stage 
> > > > > F-test, weak-instrument robust inference tests -- all
> reject null at 99%).
> > > > > Thanks!
> > > > > E
> > > > > Sent from my BlackBerry Wireless Handheld Powered by
> Gee! from
> > > > > StarHub
> > > > > 
> > > > > *
> > > > > *   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/
> > > > 
> > > > --
> > > > Justina AV Fischer, PhD
> > > > Senior Researcher
> > > > Faculty of Economics
> > > > University of Mannheim
> > > > 
> > > > homepage: http://www.justinaavfischer.de/
> > > > e-mail: javfischer@gmx.de
> > > > papers: http://ideas.repec.org/e/pfi55.html
> > > > 
> > > > 
> > > > *
> > > > *   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/
> > > 
> > > --
> > > Justina AV Fischer, PhD
> > > Senior Researcher
> > > Faculty of Economics
> > > University of Mannheim
> > > 
> > > homepage: http://www.justinaavfischer.de/
> > > e-mail: javfischer@gmx.de
> > > papers: http://ideas.repec.org/e/pfi55.html
> > > 
> > > 
> > > *
> > > *   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/
> > > 
> > 
> > 
> > --
> > Heriot-Watt University is a Scottish charity registered
> under charity
> > number SC000278.
> > 
> > 
> > *
> > *   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/
> 
> --
> Justina AV Fischer, PhD
> Senior Researcher
> Faculty of Economics
> University of Mannheim
> 
> homepage: http://www.justinaavfischer.de/
> e-mail: javfischer@gmx.de
> papers: http://ideas.repec.org/e/pfi55.html
> 
> 
> *
> *   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/
> 


--
Heriot-Watt University is a Scottish charity registered under charity number SC000278.


*
*   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/

*
*   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   |   Site index