 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: RE: ivreg2 2sls, gmm2s and autocorrelation test
Marie-Helene,
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> Marie-Hélène Felt
> Sent: Wednesday, October 22, 2008 2:35 PM
> To: [email protected]
> Subject: RE: st: RE: ivreg2 2sls, gmm2s and autocorrelation test
>
> Hello,
> so, to finish with this question, and because nobody seems to
> have any idea
> about why I have such results with abar after 2SLS robust or
> GMM2S robust, I
> have a few very simple questions:
>
> 1). to estimate a regression with one endogenous regressor
> AND PROBABLY H , what
> is best: 2SLS H robust or GMM2S H robust? what would you do?
> 2). to test for AC after that, what would you use: ivactest or abar?
> 3). to estimate a regression with one endogenous regressor
> AND PROBABLY H AND
> AC, what is best: 2SLS HAC robust or GMM2S HAC robust? what
> would you do?
Can I suggest that you compare the results of estimating with ivreg2, ivreg, and ivregress, and then abar?
I experimented a bit with ivreg2 and ivreg, and the results suggest that your question is a question about abar and not about which estimator you use to do your estimation. Most likely abar is picking up information about the estimation - e.g., whether you used -robust- or not - and then giving you a test statistic that is appropriate.
With respect to questions 1 and 3, the difference is that GMM2S is efficient as well as consistent, whereas IV is just consistent. On the other hand, Hayashi (in his 2000 textbook) suggests that GMM2S makes greater demands on the data and in practice may not perform as well as IV in some circumstances.
HTH.
Yours,
Mark
>
> thanks a lot,
>
> Marie Helene
>
> Selon "Schaffer, Mark E" <[email protected]>, 21.10.2008:
>
> > Marie-Helen,
> >
> > > -----Original Message-----
> > > From: [email protected]
> > > [mailto:[email protected]] On Behalf Of
> > > Marie-Hélène Felt
> > > Sent: Tuesday, October 21, 2008 6:34 PM
> > > To: [email protected]
> > > Subject: Re: st: RE: ivreg2 2sls, gmm2s and autocorrelation test
> > >
> > > I'm sorry I didn't mention it:
> > > I have:
> > > c:\ado\plus\i\ivreg2.ado
> > > *! ivreg2 2.2.08 15oct2007
> > > *! authors cfb & mes
> > > *! see end of file for version comments
> > >
> > > c:\ado\plus\a\abar.ado
> > > *! abar 1.1.0 9 Nov 2007
> > > *! David Roodman, Center for Global Development, Washington,
> > > DC, www.cgdev.org
> > >
> > > would a question of version explain these results?
> >
> > Possibly. Updating is a good idea anyway, so you should
> probably update and
> > see if the problem goes away.
> >
> > --Mark
> >
> > > Selon "Schaffer, Mark E" <[email protected]>, 21.10.2008:
> > >
> > > > Marie-Helene,
> > > >
> > > > > -----Original Message-----
> > > > > From: [email protected]
> > > > > [mailto:[email protected]] On Behalf Of
> > > > > Marie-Hélène Felt
> > > > > Sent: Tuesday, October 21, 2008 5:14 PM
> > > > > To: [email protected]
> > > > > Subject: st: ivreg2 2sls, gmm2s and autocorrelation test
> > > > >
> > > > > hello,
> > > > >
> > > > > I'm using IVREG2 to estimate a regression with one endogenous
> > > > > regressor.
> > > > > I noticed that the results of -abar- (test for AC) are really
> > > > > different after a
> > > > > 2SLS H robust estimation and a GMM2S H robust estimation.
> > > > > After 2SLS it seems
> > > > > that I have AC, but not after GMM2S...but it's the same
> > > equation I'm
> > > > > estimating!!
> > > >
> > > > The first step in these things is always to check that you
> > > have the latest
> > > > versions installed (and also to tell us which version of
> > > Stata you're using).
> > > >
> > > > I have
> > > >
> > > > . which ivreg2
> > > > c:\ado10\plus\i\ivreg2.ado
> > > > *! ivreg2 2.2.09 17jul2008
> > > > *! authors cfb & mes
> > > > *! see end of file for version comments
> > > >
> > > > . which abar
> > > > c:\ado10\plus\a\abar.ado
> > > > *! abar 1.1.0 9 Nov 2007
> > > > *! David Roodman, Center for Global Development, Washington, DC,
> > > > www.cgdev.org
> > > >
> > > > What about you?
> > > >
> > > > --Mark
> > > >
> > > > > I'm working with time series, and not with cross sectional
> > > > > time series, so I'm
> > > > > wondering if I'm allowed to use -abar- after both ivreg2
> > > > > estimations (2sls and
> > > > > gmm2s).
> > > > > If indeed I'm allowed to use it, how should I understand
> > > > > these results?
> > > > > Would you suggest to use -ivactest- rather than -abar-??
> > > > >
> > > > > I report hereafter my results.
> > > > >
> > > > > Thank you for your help,
> > > > >
> > > > > Marie Helene
> > > > >
> > > > > . ivreg2 lnpda lntxus lnpvus lntxca lnpvca lnipja
> > > > > (lnqda=lntxda lnpvda), robust
> > > > >
> > > > > IV (2SLS) estimation
> > > > > --------------------
> > > > >
> > > > > Estimates efficient for homoskedasticity only
> > > > > Statistics robust to heteroskedasticity
> > > > >
> > > > > Number
> > > > > of obs = 148
> > > > > F( 6,
> > > > > 141) = 4.96
> > > > > Prob >
> > > > > F = 0.0001
> > > > > Total (centered) SS = 1.439041186
> > > > > Centered R2 = -0.0540
> > > > > Total (uncentered) SS = 26084.88825
> > > > > Uncentered R2 = 0.9999
> > > > > Residual SS = 1.516815821 Root
> > > > > MSE = .1012
> > > > >
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > | Robust
> > > > > lnpda | Coef. Std. Err. z P>|z|
> > > > > [95% Conf. Interval]
> > > > > -------------+------------------------------------------------
> > > > > ----------------
> > > > > lnqda | -.0943358 .0205959 -4.58 0.000
> > > > > -.1347031 -.0539685
> > > > > lntxus | .3162592 .3225253 0.98 0.327
> > > > > -.3158787 .9483971
> > > > > lnpvus | .1865913 .2975095 0.63 0.531
> > > > > -.3965166 .7696993
> > > > > lntxca | -.2562668 .3309884 -0.77 0.439
> > > > > -.9049921 .3924585
> > > > > lnpvca | -.1924842 .3003962 -0.64 0.522
> > > > > -.7812501 .3962816
> > > > > lnipja | .2326394 .2220086 1.05 0.295
> > > > > -.2024894 .6677681
> > > > > _cons | 12.7047 1.322835 9.60 0.000
> > > > > 10.11199 15.29741
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Underidentification test (Kleibergen-Paap rk LM statistic):
> > > > > 13.650
> > > > > Chi-sq(2)
> > > > > P-val = 0.0011
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Weak identification test (Kleibergen-Paap rk Wald F
> > > > > statistic): 8.977
> > > > > Stock-Yogo weak ID test critical values: 10% maximal IV size
> > > > > 19.93
> > > > > 15% maximal IV size
> > > > > 11.59
> > > > > 20% maximal IV size
> > > > > 8.75
> > > > > 25% maximal IV size
> > > > > 7.25
> > > > > Source: Stock-Yogo (2005). Reproduced by permission.
> > > > > NB: Critical values are for Cragg-Donald F statistic and
> > > > > i.i.d. errors.
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Hansen J statistic (overidentification test of all
> > > > > instruments): 0.533
> > > > > Chi-sq(1)
> > > > > P-val = 0.4653
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Instrumented: lnqda
> > > > > Included instruments: lntxus lnpvus lntxca lnpvca lnipja
> > > > > Excluded instruments: lntxda lnpvda
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > >
> > > > > . abar, lags(6)
> > > > > Warning: The Arellano-Bond test is only valid for time series
> > > > > only if they are
> > > > > ergodic.
> > > > > Arellano-Bond test for AR(1): z = 5.77 Pr > z = 0.0000
> > > > > Arellano-Bond test for AR(2): z = 4.34 Pr > z = 0.0000
> > > > > Arellano-Bond test for AR(3): z = 3.48 Pr > z = 0.0005
> > > > > Arellano-Bond test for AR(4): z = 2.02 Pr > z = 0.0437
> > > > > Arellano-Bond test for AR(5): z = 0.47 Pr > z = 0.6380
> > > > > Arellano-Bond test for AR(6): z = 0.96 Pr > z = 0.3350
> > > > >
> > > > > . ivreg2 lnpda lntxus lnpvus lntxca lnpvca lnipja
> > > > > (lnqda=lntxda lnpvda), gmm2s
> > > > > robust
> > > > >
> > > > > 2-Step GMM estimation
> > > > > ---------------------
> > > > >
> > > > > Estimates efficient for arbitrary heteroskedasticity
> > > > > Statistics robust to heteroskedasticity
> > > > >
> > > > > Number
> > > > > of obs = 148
> > > > > F( 6,
> > > > > 141) = 4.89
> > > > > Prob >
> > > > > F = 0.0001
> > > > > Total (centered) SS = 1.439041186
> > > > > Centered R2 = -0.0525
> > > > > Total (uncentered) SS = 26084.88825
> > > > > Uncentered R2 = 0.9999
> > > > > Residual SS = 1.514566079 Root
> > > > > MSE = .1012
> > > > >
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > | Robust
> > > > > lnpda | Coef. Std. Err. z P>|z|
> > > > > [95% Conf. Interval]
> > > > > -------------+------------------------------------------------
> > > > > ----------------
> > > > > lnqda | -.0941792 .0205948 -4.57 0.000
> > > > > -.1345443 -.0538141
> > > > > lntxus | .3132924 .3224997 0.97 0.331
> > > > > -.3187954 .9453801
> > > > > lnpvus | .1590178 .2951029 0.54 0.590
> > > > > -.4193732 .7374088
> > > > > lntxca | -.2397403 .3302135 -0.73 0.468
> > > > > -.8869469 .4074662
> > > > > lnpvca | -.163539 .2977688 -0.55 0.583
> > > > > -.7471551 .4200772
> > > > > lnipja | .2394675 .2218115 1.08 0.280
> > > > > -.1952751 .6742101
> > > > > _cons | 12.61117 1.316618 9.58 0.000
> > > > > 10.03065 15.19169
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Underidentification test (Kleibergen-Paap rk LM statistic):
> > > > > 13.650
> > > > > Chi-sq(2)
> > > > > P-val = 0.0011
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Weak identification test (Kleibergen-Paap rk Wald F
> > > > > statistic): 8.977
> > > > > Stock-Yogo weak ID test critical values: 10% maximal IV size
> > > > > 19.93
> > > > > 15% maximal IV size
> > > > > 11.59
> > > > > 20% maximal IV size
> > > > > 8.75
> > > > > 25% maximal IV size
> > > > > 7.25
> > > > > Source: Stock-Yogo (2005). Reproduced by permission.
> > > > > NB: Critical values are for Cragg-Donald F statistic and
> > > > > i.i.d. errors.
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Hansen J statistic (overidentification test of all
> > > > > instruments): 0.533
> > > > > Chi-sq(1)
> > > > > P-val = 0.4653
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > > Instrumented: lnqda
> > > > > Included instruments: lntxus lnpvus lntxca lnpvca lnipja
> > > > > Excluded instruments: lntxda lnpvda
> > > > > --------------------------------------------------------------
> > > > > ----------------
> > > > >
> > > > > . abar, lags(6)
> > > > > Warning: The Arellano-Bond test is only valid for time series
> > > > > only if they are
> > > > > ergodic.
> > > > > Arellano-Bond test for AR(1): z = 0.99 Pr > z = 0.3202
> > > > > Arellano-Bond test for AR(2): z = 0.99 Pr > z = 0.3203
> > > > > Arellano-Bond test for AR(3): z = 0.99 Pr > z = 0.3232
> > > > > Arellano-Bond test for AR(4): z = 0.99 Pr > z = 0.3233
> > > > > Arellano-Bond test for AR(5): z = 0.83 Pr > z = 0.4050
> > > > > Arellano-Bond test for AR(6): z = 0.85 Pr > z = 0.3944
> > > > >
> > > > > *
> > > > > * 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/
> >
> >
>
>
> *
> * 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/