Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Hansen J with orthog option


From   "Schaffer, Mark E" <[email protected]>
To   <[email protected]>
Subject   st: RE: Hansen J with orthog option
Date   Sat, 4 Mar 2006 22:34:14 -0000

Priya,

> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of 
> Priydarshini Mahabir
> Sent: 03 March 2006 11:54
> To: [email protected]
> Subject: st: Hansen J with orthog option
> 
> Thank you very much for your suggestion on first stage F and estout.
> 
> The following query might be very similar to one posted by a 
> statalister recently. But I am still very confused and would 
> like to ask following: 
> 
> I am running a regression (below), with the gmm robust 
> cluster option to cater for any arbitrary heteroskedasticity 
> and group correlation. Based on intuition, variable lchx is 
> correlated with the error term. I want to formally test this. 
> Since ivendog cannot be used with robust or cluster , I use 
> the "orthog" option.
> 
> Regression:
> 
> ivreg2 limp lgdpm lcapm lgdpx lcapx  ldist lareap landl 
> island border comlang comcol colony lchx (= lchgdp iecor), 
> ffirst gmm robust cluster(pid) orthog(lchx)
> 
> ----test reult----
> --------------------------------------------------------------
> ----------------
> Hansen J statistic (Lagrange multiplier test of excluded 
> instruments):  90.466
>                                                               
>                Chi-sq(2) P-val =    0.0000
> -orthog- option:
> Hansen J statistic (eqn. excluding suspect orthog. 
> conditions):          0.022
>                                                               
>                 Chi-sq(1) P-val =    0.8829
> 
> C statistic (exogeneity/orthogonality of suspect 
> instruments):          90.445
>                                                               
>                  Chi-sq(1) P-val =    0.0000
> Instruments tested:   lchx
> --------------------------------
> Included instruments: lgdpm lcapm lgdpx lcapx ldist lareap 
> landl island border
>                       comlang comcol colony lchx Excluded 
> instruments: lchgdp iecor
> --------------------------------------------------------------
> ------------------------------
> Questions:
> 
> 1. The C stat confirms that lchx is in fact endogenous and IV 
> technique is appropriate. Is this correct?

Yes.

> 2. But which Hansen J stat would indicate whether instruments 
> are valid or not?

Where the output says:

Hansen J statistic (eqn. excluding suspect orthog. conditions):   0.022
                                             Chi-sq(1) P-val =    0.8829

This is the J statistic for the equation that "excludes the suspect
orthogonality condition", i.e., that treats lchx as endogenous.  (The
"suspect orthogonality condition" is the claim that lchx is exogenous.)

> What is the difference between the two

The J statistic that treats lchx as exogenous = 90.466, i.e., it's big.
This suggests that one or more variables treated as exogenous (either
regressors or excluded instruments) actually isn't, or the equation is
misspecified in some other way.

The J statistic that treats lchx as endogenous = 0.022, i.e., it's
small.  This suggests that the variables you treat as exogenous actually
are.

> and why is there such a large disparity in magnitude?

Precisely because treating lchx as exogenous - adding it to the
orthogonality conditions - is dubious.  You have evidence that it's
endogenous, precisely because if you treat it as exogenous, the J
statistic goes through the roof.

Note, by the way, that if you now go ahead and estimate the equation but
treat lchx as endogenous,

ivreg2 limp lgdpm lcapm lgdpx lcapx  ldist lareap landl 
island border comlang comcol colony (lchx = lchgdp iecor), 
ffirst gmm robust cluster(pid)

you will get a J stat that is slightly different from 0.022.  The reason
is that ivreg2 uses a modified J stat when constructing the C test
statistic so that a positive test statistic is guaranteed.

Our 2003 Stata Journal article (Baum-Schaffer-Stillman) has a more
detailed discussion.

HTH.

--Mark

> Many thanks
> Priya
> 
> 
> Research Postgraduate
> Leverhulme Centre for Research on Globalisation and Economic 
> Policy C42, School of Economics The University of Nottingham 
> University Park Nottingham,
> NG7 2RD
> UK
> Tel: + 44  (0)115 951 5870
> 
> >>> [email protected] 03/02/06 4:58 pm >>>
> Priya,
> 
> If you have one endogenous variable, and you want a robust 
> first-stage F-statistic, then you can extract it from the 
> saved -e(first)- matrix.
> Just do an -ivreg2- estimation, type -mat list e(first)-, and 
> you'll see where the F-stat is saved.
> 
> Cheers,
> Mark
> 
> > -----Original Message-----
> > From: [email protected]
> > [mailto:[email protected]] On Behalf Of 
> > Priydarshini Mahabir
> > Sent: 02 March 2006 16:23
> > To: [email protected]
> > Subject: st: First stage F and estout
> > 
> > Hi,
> > I am using the following command to estimate a gravity model (panel 
> > data)
> > 
> > ivreg2 limp lgdpm lcapm lgdpx lcapx  ldist lareap landl 
> island border 
> > comlang comcol colony iicor iecor (lhkchx= lhkchgdp ichcor ihkcor), 
> > first gmm robust cluster(pid)
> > 
> > I am tabulating the results using the estout command. 
> >  estout *, cells(b(star fmt(%9.3f)) se(par)) style (fixed) 
> stats( r2 j 
> > jp idstat idp )
> > 
> > My question is how to get the first stage F statistic (test for 
> > excluded instruments)  in the table as well? When I include 
> "F" within 
> > the stats( ), i only get the 2nd stage F stat.
> > 
> > Is there any other command for tabulating regression results? 
> > Many thanks
> > 
> > Priya
> > 
> > Research Postgraduate
> > Leverhulme Centre for Research on Globalisation and Economic Policy 
> > C42, School of Economics The University of Nottingham 
> University Park 
> > Nottingham,
> > NG7 2RD
> > UK
> > Tel: + 44  (0)115 951 5870
> > 
> > 
> > This message has been checked for viruses but the contents of an 
> > attachment may still contain software viruses, which could 
> damage your 
> > computer system:
> > you are advised to perform your own checks. Email 
> communications with 
> > the University of Nottingham may be monitored as permitted by UK 
> > legislation.
> > 
> > 
> > *
> > *   For searches and help try:
> > *   http://www.stata.com/support/faqs/res/findit.html 
> > *   http://www.stata.com/support/statalist/faq
> > *   http://www.ats.ucla.edu/stat/stata/ 
> > 
> > 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html 
> *   http://www.stata.com/support/statalist/faq 
> *   http://www.ats.ucla.edu/stat/stata/
> 
> This message has been checked for viruses but the contents of 
> an attachment may still contain software viruses, which could 
> damage your computer system:
> you are advised to perform your own checks. Email 
> communications with the University of Nottingham may be 
> monitored as permitted by UK legislation.
> 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 
> 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index