Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: RE: RE: xtoverid error: internal reestimation of eqn differs from original
From
"Schaffer, Mark E" <[email protected]>
To
"[email protected]" <[email protected]>
Subject
st: RE: RE: xtoverid error: internal reestimation of eqn differs from original
Date
Tue, 28 Jan 2014 17:26:47 +0000
Sezer,
This is odd - all the coefficients are slightly but noticeably different.
The first-stage F stat is also suspiciously high - almost 2000. This corresponds to the t stat of 44 (!) for hlthlm in the first-stage results.
Can you check a few other things?
- Are you using the latest version of -ivreg2-? From within Stata,
which ivreg2, all
will tell you.
- When you estimate using -xtivreg- with the -ec2sls- option, do the results (main equation and first stage) differ from what -xtoverid,noi- reports?
- When you use the -first- option with -xtivreg-, do the first-stage results differ from what -xtoverid,noi- reports?
BTW, since your equation is exactly identified when you use the default G2SLS estimator, this last point implies that you should be able to get the first-stage F from the first-stage results for official -xtivreg- (just square the t-stat). In other words, you can get a compatible first-stage F stat without using -xtoverid-.
--Mark
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of SEZER ALCAN
> Sent: 28 January 2014 16:11
> To: [email protected]
> Subject: st: RE: xtoverid error: internal reestimation of eqn differs from
> original
>
> I copied original estimation and xtoverid, noi results. As you can see same
> number of observation is used and coefficients are not substantially
> different. You may see it below.
>
> I very much appreciate your comments
> Regards,
> Sezer Alcan
>
>
>
> xtivreg y $xvars (sahgd = hlthlm), re
>
> G2SLS random-effects IV regression Number of obs = 10051
> Group variable: FKIMLIK Number of groups = 3784
>
> R-sq: within = 0.0479 Obs per group: min = 2
> between = 0.5865 avg = 2.7
> overall = 0.5116 max = 4
>
> Wald chi2(26) = 5525.18
> corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000
>
> ------------------------------------------------------------------------------
> y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> sahgd | .0681581 .0329966 2.07 0.039 .003486 .1328303
> age | .0841604 .0058326 14.43 0.000 .0727286 .0955922
> age2 | -.0008967 .0000755 -11.88 0.000 -.0010446 -.0007488
> wed | .122449 .0210754 5.81 0.000 .081142 .163756
> illit | -.1637081 .0654399 -2.50 0.012 -.291968 -.0354482
> secprim | .0974501 .0232707 4.19 0.000 .0518404 .1430599
> highsch | .2458178 .0269148 9.13 0.000 .1930658 .2985698
> techigh | .2928797 .0262241 11.17 0.000 .2414815 .344278
> univ | .5371905 .0325742 16.49 0.000 .4733463 .6010348
> nutriengh | .1259238 .0120452 10.45 0.000 .1023156 .1495319
> socsecprb | .2143147 .0183127 11.70 0.000 .1784225 .2502069
> prof | .4363263 .0348374 12.52 0.000 .3680462 .5046064
> manag | .4134413 .0367269 11.26 0.000 .3414578 .4854248
> craft | -.0143766 .0231956 -0.62 0.535 -.0598391 .0310859
> machop | -.0575744 .02341 -2.46 0.014 -.1034571 -.0116917
> unskll | -.1693637 .0235825 -7.18 0.000 -.2155845 -.1231429
> nutriunskll | -.0274815 .0302797 -0.91 0.364 -.0868286 .0318655
> sector3 | .0279124 .019475 1.43 0.152 -.0102579 .0660826
> sector5 | -.0255151 .0249776 -1.02 0.307 -.0744704 .0234402
> sector6 | -.1065776 .0218064 -4.89 0.000 -.1493174 -.0638379
> firmsize10 | -.0530375 .0196092 -2.70 0.007 -.0914708 -.0146041
> firmsize2049 | .0864132 .023699 3.65 0.000 .039964 .1328624
> firmsize50p | .1883012 .0214952 8.76 0.000 .1461715 .2304309
> perm | .1811101 .020985 8.63 0.000 .1399803 .2222399
> yr2008 | -.1080605 .0104045 -10.39 0.000 -.128453 -.0876681
> yr2009 | -.0215809 .0089087 -2.42 0.015 -.0390417 -.0041201
> _cons | -1.27803 .1101712 -11.60 0.000 -1.493962 -1.062099
> -------------+----------------------------------------------------------------
> sigma_u | .40067359
> sigma_e | .36180592
> rho | .55084324 (fraction of variance due to u_i)
> ------------------------------------------------------------------------------
> Instrumented: sahgd
> Instruments: age age2 wed illit secprim highsch techigh univ nutriengh
> socsecprb prof manag craft machop unskll nutriunskll sector3
> sector5 sector6 firmsize10 firmsize2049 firmsize50p perm
> yr2008 yr2009 hlthlm
> ------------------------------------------------------------------------------
>
> . xtoverid
> xtoverid error: internal reestimation of eqn differs from original
> r(198);
>
> . xtoverid, noi
>
> First-stage regressions
> -----------------------
>
> First-stage regression of __00000I:
>
> OLS estimation
> --------------
>
> Estimates efficient for homoskedasticity only
> Statistics consistent for homoskedasticity only
>
> Number of obs = 10051
> F( 27, 10024) = 774.81
> Prob > F = 0.0000
> Total (centered) SS = 948.7104765 Centered R2 = 0.2241
> Total (uncentered) SS = 2272.334237 Uncentered R2 = 0.6761
> Residual SS = 736.1052268 Root MSE = .271
>
> ------------------------------------------------------------------------------
> __00000I | Coef. Std. Err. t P>|t| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> __00000O | -.0012507 .003727 -0.34 0.737 -.0085564 .0060551
> __00000R | -.0000673 .0000483 -1.39 0.164 -.000162 .0000275
> __00000U | -.0085974 .0141249 -0.61 0.543 -.036285 .0190902
> __00000X | -.210671 .042279 -4.98 0.000 -.2935463 -.1277956
> __000010 | .0274273 .0149386 1.84 0.066 -.0018553 .0567099
> __000013 | .0184523 .0177786 1.04 0.299 -.0163974 .0533019
> __000016 | .0513269 .01727 2.97 0.003 .0174743 .0851794
> __000019 | .0530563 .0218709 2.43 0.015 .010185 .0959275
> __00001C | .0344548 .0085254 4.04 0.000 .0177433 .0511663
> __00001F | .0304538 .0124402 2.45 0.014 .0060685 .0548392
> __00001I | -.0071152 .0235747 -0.30 0.763 -.0533263 .0390959
> __00001L | .0168858 .0254501 0.66 0.507 -.0330014 .0667731
> __00001O | -.0067492 .015419 -0.44 0.662 -.0369735 .0234751
> __00001R | -.0004078 .01575 -0.03 0.979 -.031281 .0304653
> __00001U | -.0480332 .0157943 -3.04 0.002 -.0789933 -.0170732
> __00001X | .0131181 .0212184 0.62 0.536 -.0284742 .0547104
> __000020 | -.0202486 .0132233 -1.53 0.126 -.0461689 .0056717
> __000023 | .0242722 .016993 1.43 0.153 -.0090374 .0575819
> __000026 | -.0206897 .0145767 -1.42 0.156 -.049263 .0078835
> __000029 | .0081375 .0134997 0.60 0.547 -.0183246 .0345997
> __00002C | .0016234 .0162664 0.10 0.921 -.030262 .0335087
> __00002F | -.0160256 .0147908 -1.08 0.279 -.0450186 .0129673
> __00002I | .0114303 .0143138 0.80 0.425 -.0166277 .0394883
> __00002L | .0069627 .0073903 0.94 0.346 -.0075237 .0214491
> __00002O | -.0058741 .0063412 -0.93 0.354 -.0183042 .006556
> __00000E | .9547274 .0680486 14.03 0.000 .8213384 1.088116
> __00000L | -.4283008 .0097704 -43.84 0.000 -.4474527 -.4091489
> ------------------------------------------------------------------------------
> Included instruments: __00000O __00000R __00000U __00000X __000010
> __000013
> __000016 __000019 __00001C __00001F __00001I __00001L
> __00001O __00001R __00001U __00001X __000020 __000023
> __000026 __000029 __00002C __00002F __00002I __00002L
> __00002O __00000E __00000L
> ------------------------------------------------------------------------------
> F test of excluded instruments:
> F( 1, 10024) = 1921.65
> Prob > F = 0.0000
> Angrist-Pischke multivariate F test of excluded instruments:
> F( 1, 10024) = 1921.65
> Prob > F = 0.0000
>
>
>
> Summary results for first-stage regressions
> -------------------------------------------
>
> (Underid) (Weak id)
> Variable | F( 1, 10024) P-val | AP Chi-sq( 1) P-val | AP F( 1, 10024)
> __00000I | 1921.65 0.0000 | 1926.83 0.0000 | 1921.65
>
> Stock-Yogo weak ID test critical values for single endogenous regressor:
> 10% maximal IV size 16.38
> 15% maximal IV size 8.96
> 20% maximal IV size 6.66
> 25% maximal IV size 5.53
> Source: Stock-Yogo (2005). Reproduced by permission.
>
> Underidentification test
> Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified)
> Ha: matrix has rank=K1 (identified)
> Anderson canon. corr. LM statistic Chi-sq(1)=1616.87 P-val=0.0000
>
> Weak identification test
> Ho: equation is weakly identified
> Cragg-Donald Wald F statistic 1921.65
>
> Stock-Yogo weak ID test critical values for K1=1 and L1=1:
> 10% maximal IV size 16.38
> 15% maximal IV size 8.96
> 20% maximal IV size 6.66
> 25% maximal IV size 5.53
> Source: Stock-Yogo (2005). Reproduced by permission.
>
> Weak-instrument-robust inference
> Tests of joint significance of endogenous regressors B1 in main equation
> Ho: B1=0 and orthogonality conditions are valid
> Anderson-Rubin Wald test F(1,10024)= 7.17 P-val=0.0074
> Anderson-Rubin Wald test Chi-sq(1)= 7.19 P-val=0.0073
> Stock-Wright LM S statistic Chi-sq(1)= 7.19 P-val=0.0073
>
> Number of observations N = 10051
> Number of regressors K = 27
> Number of endogenous regressors K1 = 1
> Number of instruments L = 27
> Number of excluded instruments L1 = 1
>
> IV (2SLS) estimation
> --------------------
>
> Estimates efficient for homoskedasticity only
> Statistics consistent for homoskedasticity only
>
> Number of obs = 10051
> F( 27, 10024) = 927.71
> Prob > F = 0.0000
> Total (centered) SS = 2432.217133 Centered R2 = 0.3602
> Total (uncentered) SS = 5444.007778 Uncentered R2 = 0.7142
> Residual SS = 1556.070469 Root MSE = .3935
>
> ------------------------------------------------------------------------------
> __00000G | Coef. Std. Err. z P>|z| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> __00000I | .0888147 .0331225 2.68 0.007 .0238959 .1537336
> __00000O | .0757184 .0054114 13.99 0.000 .0651122 .0863246
> __00000R | -.0008138 .0000703 -11.58 0.000 -.0009515 -.000676
> __00000U | .1304926 .0205185 6.36 0.000 .090277 .1707081
> __00000X | -.1801946 .0620044 -2.91 0.004 -.3017211 -.0586681
> __000010 | .1117898 .0217153 5.15 0.000 .0692286 .154351
> __000013 | .2030786 .0258334 7.86 0.000 .152446 .2537111
> __000016 | .2834297 .0251911 11.25 0.000 .234056 .3328033
> __000019 | .5128589 .0318645 16.09 0.000 .4504056 .5753122
> __00001C | .1253982 .0124417 10.08 0.000 .1010129 .1497835
> __00001F | .2136176 .0180921 11.81 0.000 .1781577 .2490774
> __00001I | .4131903 .0342317 12.07 0.000 .3460974 .4802832
> __00001L | .436081 .0369653 11.80 0.000 .3636303 .5085317
> __00001O | -.019046 .0223885 -0.85 0.395 -.0629268 .0248347
> __00001R | -.0736599 .0228685 -3.22 0.001 -.1184813 -.0288385
> __00001U | -.1782438 .0229984 -7.75 0.000 -.2233199 -.1331676
> __00001X | -.0256015 .0308117 -0.83 0.406 -.0859914 .0347884
> __000020 | .0228741 .0192233 1.19 0.234 -.0148028 .060551
> __000023 | -.0349952 .0246815 -1.42 0.156 -.0833701 .0133797
> __000026 | -.1084179 .0211765 -5.12 0.000 -.149923 -.0669128
> __000029 | -.0628365 .0196054 -3.21 0.001 -.1012624 -.0244106
> __00002C | .0772072 .0236188 3.27 0.001 .0309153 .1234991
> __00002F | .1776297 .0214788 8.27 0.000 .1355321 .2197273
> __00002I | .1943209 .0207961 9.34 0.000 .1535613 .2350805
> __00002L | -.1077913 .0107296 -10.05 0.000 -.128821 -.0867616
> __00002O | -.0214963 .0092137 -2.33 0.020 -.0395548 -.0034378
> __00000E | -1.095742 .1031156 -10.63 0.000 -1.297845 -.8936388
> ------------------------------------------------------------------------------
> Underidentification test (Anderson canon. corr. LM statistic): 1616.868
> Chi-sq(1) P-val = 0.0000
> ------------------------------------------------------------------------------
> Weak identification test (Cragg-Donald Wald F statistic): 1921.654
> Stock-Yogo weak ID test critical values: 10% maximal IV size 16.38
> 15% maximal IV size 8.96
> 20% maximal IV size 6.66
> 25% maximal IV size 5.53
> Source: Stock-Yogo (2005). Reproduced by permission.
> ------------------------------------------------------------------------------
> Sargan statistic (overidentification test of all instruments): 0.000
> (equation exactly identified)
> ------------------------------------------------------------------------------
> Instrumented: __00000I
> Included instruments: __00000O __00000R __00000U __00000X __000010
> __000013
> __000016 __000019 __00001C __00001F __00001I __00001L
> __00001O __00001R __00001U __00001X __000020 __000023
> __000026 __000029 __00002C __00002F __00002I __00002L
> __00002O __00000E
> Excluded instruments: __00000L
> ------------------------------------------------------------------------------
> xtoverid error: internal reestimation of eqn differs from original
> r(198);
>
> ________________________________________
> From: [email protected] [owner-
> [email protected]] on behalf of Schaffer, Mark E
> [[email protected]]
> Sent: Tuesday, January 28, 2014 5:51 PM
> To: [email protected]
> Subject: st: RE: RE: xtoverid error: internal reestimation of eqn differs from
> original
>
> Sezer,
>
> > -----Original Message-----
> > From: [email protected] [mailto:owner-
> > [email protected]] On Behalf Of SEZER ALCAN
> > Sent: 28 January 2014 15:42
> > To: [email protected]
> > Subject: st: RE: xtoverid error: internal reestimation of eqn differs from
> > original
> >
> > Dr Schaffer,
> > Although xtoverid gives error, xtoverid, noi still reports under
> identification
> > test, weak identification test and sargan statistics.
> > What if internal estimation results by ivreg2 after xtoverid,noi command
> > does show any collinearity (no regressors are dropped) is it possible to rely
> > on these test results?
>
> They should be OK, but because they don't match the original estimation you
> can't use them with the original estimation results - they don't apply.
>
> In what respects to the original results differ from the xtoverid results? Do
> they use the same number of observations? Same number of coefficients
> reported? Are the coefficients very slightly different or substantially
> different? Are the same variables being treated as endogenous? (One
> possibility is that collinearity with the excluded instruments is causing an
> endogenous regressor to be treated as if it is exogenous.)
>
> If the results aren't confidential, you can post them here and we can try to
> work out what is going on.
>
> --Mark
>
> > Regards,
> > Sezer Alcan
> > ________________________________________
> > From: [email protected] [owner-
> > [email protected]] on behalf of Schaffer, Mark E
> > [[email protected]]
> > Sent: Wednesday, January 22, 2014 10:16 PM
> > To: [email protected]
> > Subject: st: RE: xtoverid error: internal reestimation of eqn differs from
> > original
> >
> > Sezer,
> >
> > > -----Original Message-----
> > > From: [email protected] [mailto:owner-
> > > [email protected]] On Behalf Of SEZER ALCAN
> > > Sent: 22 January 2014 18:11
> > > To: [email protected]
> > > Subject: st: xtoverid error: internal reestimation of eqn differs from
> > > original
> > >
> > > When I use xtoverid command in stata after estimation by xthtaylor or
> > xtivreg
> > > re /xtivreg fe, I get "xtoverid error: internal reestimation of eqn differs
> > from
> > > original". I read all replies under this subject; Dr Schaffer says
> > > that it may be because of collinearity. However when I execute "
> > > xtoverid, noisily " I do not detect any collinearity.
> >
> > Strictly speaking, my suggestion isn't that collinearity is causing the internal
> > reestimation to vary from the original estimation. Rather, it's how Stata's
> > official xthtaylor or xtivreg handles the collinearity vs. how the internal
> > reestimation by xtoverid (using ivreg2) handles it.
> >
> > If there is perfect collinearity, Stata estimators will typically drop enough
> > regressors or instruments to make the collinearity go away. But which of
> the
> > collinear regressors get dropped can vary. And this can make the
> estimated
> > coefficients vary even though, in some sense, it's the same model.
> >
> > If you look at the internal estimation results by ivreg2 that xtoverid,noi
> > reports, you should see at the bottom of the main output whether ivreg2 is
> > dropping any collinear regressors or instruments. You won't be able to
> easily
> > map the ones that are being dropped to your dataset, because xtoverid
> uses
> > temporary variables, but it should at least be a good clue to what is going
> on.
> >
> > HTH,
> > Mark
> >
> > >
> > > I was wondering if this problem was due to unbalanced nature of the
> > > panel I was using. If so what other test can be executed to test over
> > > identification restrictions in an unbalanced panel.
> > > If collinearity isn't the problem of the error, can xtoverid, noi
> > > test results be trusted ?
> > >
> > >
> > > I work wiht Stata 12 and latest version of xtoverid is installed.
> > >
> > > Thanks all in advance,
> > > Sezer Alcan
> > > *
> > > * For searches and help try:
> > > * http://www.stata.com/help.cgi?search
> > > * http://www.stata.com/support/faqs/resources/statalist-faq/
> > > * http://www.ats.ucla.edu/stat/stata/
> >
> >
> > -----
> > Sunday Times Scottish University of the Year 2011-2013 Top in the UK for
> > student experience Fourth university in the UK and top in Scotland
> (National
> > Student Survey 2012)
> >
> > We invite research leaders and ambitious early career researchers to join
> us
> > in leading and driving research in key inter-disciplinary themes.
> > Please see www.hw.ac.uk/researchleaders for further information and
> how
> > to apply.
> >
> > 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/faqs/resources/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/faqs/resources/statalist-faq/
> > * http://www.ats.ucla.edu/stat/stata/
>
>
> -----
> Sunday Times Scottish University of the Year 2011-2013
> Top in the UK for student experience
> Fourth university in the UK and top in Scotland (National Student Survey
> 2012)
>
>
> We invite research leaders and ambitious early career researchers to
> join us in leading and driving research in key inter-disciplinary themes.
> Please see www.hw.ac.uk/researchleaders for further information and how
> to apply.
>
> 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/faqs/resources/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/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
-----
Sunday Times Scottish University of the Year 2011-2013
Top in the UK for student experience
Fourth university in the UK and top in Scotland (National Student Survey 2012)
We invite research leaders and ambitious early career researchers to
join us in leading and driving research in key inter-disciplinary themes.
Please see www.hw.ac.uk/researchleaders for further information and how
to apply.
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/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/