Ekrem,
The problem is a fundamental econometric problem, and it doesn't matter whether you use xtivreg2,fe or ivreg2 with dummies - it's the same problem.
If you check out their paper, you'll see that the bias in the usual Eicker-Huber-White-sandwich-robust VCE is proportional to 1/(T-1). If T is large enough for you to conclude the bias is small, you should be OK. But if T is small, then any tests that use this VCE won't be consistent.
Cheers,
Mark
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> Ekrem Kalkan
> Sent: 10 January 2010 22:56
> To: [email protected]
> Subject: Re: st: RE: RE: ivhettest
>
> Thank you Mark , it now works.
>
> Regarding the problem with fixed-effects, do you think that I
> can use ivhettest after the command ivreg2 by adding panelid
> dummies as regressors instead of a regression with xtivreg2.
>
> Ekrem.
>
> 2010/1/11 Schaffer, Mark E <[email protected]>:
> > Ekrem,
> >
> > This is a bug I accidentally introduced into a recent
> update of ivhettest, and which I subsequently corrected
> shortly thereafter. You must be one of the few who
> downloaded the buggy version.
> >
> > If you reinstall ivhettest, the new version should work.
> >
> > That said, the Stock-Watson point still stands - unless
> your data are large-T-large-N (i.e., you have a reasonable
> number of observations in the time series dimension as well
> as the cross-section dimension), the test provided by
> ivhettest won't be consistent with the fixed effects estimator.
> >
> > Best wishes,
> > Mark
> >
> >> -----Original Message-----
> >> From: [email protected]
> >> [mailto:[email protected]] On Behalf Of Ekrem
> >> Kalkan
> >> Sent: 10 January 2010 22:21
> >> To: [email protected]
> >> Subject: Re: st: RE: RE: ivhettest
> >>
> >> First I set my panel units: "xtset productid marketid"
> >> "marketid" is created by egen marketid=group(city month).
> >>
> >> Then, I run the following:
> >>
> >> ivreg2 y (x1 x2 = z1 z2 z3 ) productdummies citydummies
> monthdummies
> >> othervariables
> >>
> >> Here is the trace of "ivhettest, bpg" after this regression:
> >>
> >> - version 7.0
> >> - local version 1.1.7
> >> - syntax [varlist(default=none)] [if] [in] [, ivlev ivsq
> ivcp fitlev
> >> fitsq ph phnorm phsym nr2 bpg all ]
> >> - if "`e(cmd)'" != "ivreg" & "`e(cmd)'" != "ivreg2" &
> "`e(cmd)'" !=
> >> "ivgmm0" & "`e(cmd)'" != "regress" { = if "ivreg2" != "ivreg" &
> >> "ivreg2" != "ivreg2" & "ivreg2" != "ivgmm0"
> >> & "ivreg2" != "regress" {
> >> error 301
> >> }
> >> - if "`e(fwlcons)'`e(partialcons)'" != "" { = if "0" != "" {
> >> - di in r "ivhettest not allowed after ivreg2 with partial
> >> (previously
> >> fwl) option"
> >> ivhettest not allowed after ivreg2 with partial (previously
> >> fwl) option
> >> - error 499
> >> }
> >>
> >> I hope this helps.
> >> Thanks.
> >>
> >> Ekrem.
> >>
> >>
> >> 2010/1/10 Martin Weiss <[email protected]>:
> >> >
> >> > <>
> >> >
> >> > " There is not the "partial" command."
> >> >
> >> >
> >> > -partial- would be an option to -ivreg2-, not a command. It is
> >> > probably best if you just show the sequence of commands
> as typed in
> >> > the command line, and as the FAQ request. In the absence of this
> >> > information, the problem is difficult to diagnose. Also
> >> -set trace on-
> >> > and report the lines around the error.
> >> >
> >> >
> >> > HTH
> >> > Martin
> >> >
> >> > -----Ursprüngliche Nachricht-----
> >> > Von: [email protected]
> >> > [mailto:[email protected]] Im Auftrag
> von Ekrem
> >> > Kalkan
> >> > Gesendet: Sonntag, 10. Januar 2010 22:53
> >> > An: [email protected]
> >> > Betreff: Re: st: RE: RE: ivhettest
> >> >
> >> > Hello,
> >> >
> >> > My observations are in 3-dimensional space: Product x
> City x Time.
> >> >
> >> > I have 93 products as cross-sectional units. Instead of using
> >> > commands like xtivreg or xtivreg2, I use dummy variables
> for each
> >> > cross-sectional units and run the model with command
> >> "ivreg2". There
> >> > is not the "partial" command.
> >> >
> >> > My equation is below:
> >> >
> >> > ivreg2 y (x1 x2 = z1 z2 z3 ) productdummies citydummies
> >> monthdummies
> >> > othervariables
> >> >
> >> > Then I implement "ivhettest" and get the error I had wrote
> >> previously.
> >> >
> >> > Thanks.
> >> >
> >> > Ekrem.
> >> >
> >> >
> >> > 2010/1/10 Schaffer, Mark E <[email protected]>:
> >> >> Ekrenm,
> >> >>
> >> >> In addition to what Eric and Martin have said, I should
> >> note that the
> >> >> test employed by -ivhettest- (as well as Stata's -estat
> >> hettest- and
> >> >> others, I suspect) isn't valid after panel data estimation
> >> with fixed
> >> >> effects.
> >> >>
> >> >> The reason is that the test based on the contrast between the
> >> >> classical non-robust VCV and the heteroskedastic-robust
> >> VCV, as per
> >> >> White (Econometrica 1980). However, as Stock and Watson
> >> >> (Econometrica 2008) have shown, the standard
> >> >> Eicker-Huber-White-robust-sandwich VCV isn't consistent for the
> >> >> standard fixed effects estimator. Thus the test isn't
> >> valid, because
> >> >> in the presence of heteroskedasticity it's contrasting two
> >> different inconsistent VCVs.
> >> >>
> >> >> --Mark
> >> >>
> >> >>> -----Original Message-----
> >> >>> From: [email protected]
> >> >>> [mailto:[email protected]] On Behalf
> >> Of DE SOUZA
> >> >>> Eric
> >> >>> Sent: 10 January 2010 20:56
> >> >>> To: '[email protected]'
> >> >>> Subject: st: RE: ivhettest
> >> >>>
> >> >>> It says what is says. You used the "partial" option
> with ivreg2.
> >> >>> This partials out the effects of certain variables.
> >> >>> When you use this option, you cannot use ivhettest after it.
> >> >>>
> >> >>> If you did not use the "partial" option with the command
> >> -ivreg2-,
> >> >>> please indicate exactly what you did.
> >> >>>
> >> >>>
> >> >>> Eric de Souza
> >> >>> College of Europe
> >> >>> BE-8000 Brugge (Bruges)
> >> >>> Belgium
> >> >>>
> >> >>> -----Original Message-----
> >> >>> From: [email protected]
> >> >>> [mailto:[email protected]] On
> Behalf Of Ekrem
> >> >>> Kalkan
> >> >>> Sent: 10 January 2010 21:50
> >> >>> To: [email protected]
> >> >>> Subject: st: ivhettest
> >> >>>
> >> >>> Dear Stata users,
> >> >>>
> >> >>> After running a regression with "ivreg2" command, I run
> >> "ivhettest"
> >> >>> command for testing for heteroscedasticity. But, I
> >> receive the error
> >> >>> below:
> >> >>>
> >> >>> . ivhettest, bpg
> >> >>> ivhettest not allowed after ivreg2 with partial (previously
> >> >>> fwl) option
> >> >>>
> >> >>> Could you tell me what does this mean and how can I test
> >> >>> heteroscedasticiy after an IV regression with panel data.
> >> >>>
> >> >>> Thank you.
> >> >>>
> >> >>> Ekrem Kalkan
> >> >>> Turkish Competition Authority.
> >> >>> *
> >> >>> * 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/
> >> >>>
> >> >>
> >> >>
> >> >> --
> >> >> 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/
> >> >
> >>
> >> *
> >> * 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/
>
--
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/
