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]
RE: st: RE: Hausman-Taylor AR(1) estimator
From
KORAY ERCİHAN <[email protected]>
To
"[email protected]" <[email protected]>
Subject
RE: st: RE: Hausman-Taylor AR(1) estimator
Date
Fri, 7 Sep 2012 00:19:34 +0300
Dear Mark,
I first command: xi: xthtaylor ln_exports_od ln_gdp ln_sgdp dgdppc contig comlang_off ln_distcap euin i.year, endog( euin contig comlang_off)
Since xtoverid gives the following error
xtoverid error: internal reestimation of eqn differs from original
r(198)
I tried "xtoverid, noi" the output stata shows the Sargan Statistic chi-sq(3)=4.74 with p-value=0.1919
Then I started try different specifications
1- xi: xthtaylor ln_exports_od ln_gdp ln_sgdp dgdppc contig comlang_off ln_distcap euin i.year, endog( ln_gdp dgdppc comlang_off)
xtoverid
xtoverid error: internal reestimation of eqn differs from original
r(198);
xtoverid, noi
Sargan statistics = 0.121
Chi-sq(4) P-val = 0.9983
2-xi: xthtaylor ln_exports_od ln_gdp ln_sgdp dgdppc contig comlang_off ln_distcap euin i.year, endog( ln_gdp dgdppc comlang_off) vce(bootstrap)
xtoverid, noi gives the following result
xtoverid, noi
Warning - endogenous variable(s) collinear with instruments
Vars now exogenous: __00000J __00000M __00000P __00000S __00000V __00000Y
__000011 __000014 __000017 __00001A __00001D __00001G
__00001J __00001M __00001P __00001S __00001Z __000020
Warning - collinearities detected
Vars dropped: __00000O __00000R __00000U __00000X __000010 __000013
__000019 __00001C __00001F __00001I __00001L __00001O
__00001R __00000H __00000K __00000Q __00000T __00000W
__00000Z __000012 __000015 __000018 __00001B __00001E
__00001H __00001K __00001N __00001Q contig
Warning: estimated covariance matrix of moment conditions not of full rank.
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
First-stage regressions
-----------------------
First-stage regression of __00001V:
OLS estimation
--------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on pairs1
Number of clusters (pairs1) = 168 Number of obs = 2517
F( 26, 167) = 2.2e+16
Prob > F = 0.0000
Total (centered) SS = 113.0830424 Centered R2 = 0.9995
Total (uncentered) SS = 186.3030374 Uncentered R2 = 0.9997
Residual SS = .0529722933 Root MSE = .004611
------------------------------------------------------------------------------
| Robust
__00001V | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
__00000E | -1564.76 1123.943 -1.39 0.166 -3783.727 654.2076
__00000J | -.5162412 .0351062 -14.71 0.000 -.5855503 -.4469321
__00000M | -.8659744 .6635963 -1.30 0.194 -2.176093 .4441444
__00000P | 3.50e-08 4.98e-08 0.70 0.483 -6.33e-08 1.33e-07
__00000S | 8.24e-08 1.85e-07 0.44 0.657 -2.84e-07 4.48e-07
__00000V | 8.23e-08 1.88e-07 0.44 0.662 -2.88e-07 4.53e-07
__00000Y | 7.54e-08 1.62e-07 0.46 0.643 -2.45e-07 3.96e-07
__000011 | 8.80e-08 2.13e-07 0.41 0.681 -3.33e-07 5.09e-07
__000014 | 8.60e-08 2.05e-07 0.42 0.676 -3.19e-07 4.91e-07
__000017 | -58.81884 45.06964 -1.31 0.194 -147.7985 30.16085
__00001A | 6.93e-08 1.88e-07 0.37 0.714 -3.03e-07 4.41e-07
__00001D | 1.01e-07 2.67e-07 0.38 0.705 -4.25e-07 6.27e-07
__00001G | 1.58e-07 4.28e-07 0.37 0.712 -6.87e-07 1.00e-06
__00001J | 2.42e-07 5.92e-07 0.41 0.683 -9.26e-07 1.41e-06
__00001M | 2.56e-07 6.34e-07 0.40 0.687 -9.96e-07 1.51e-06
__00001P | 2.79e-07 6.87e-07 0.41 0.685 -1.08e-06 1.64e-06
__00001S | 3.29e-07 8.34e-07 0.39 0.694 -1.32e-06 1.97e-06
__00001Z | .1080825 .4199995 0.26 0.797 -.7211103 .9372752
__000020 | -222.1497 157.0535 -1.41 0.159 -532.216 87.91648
__00000I | .516241 .0351062 14.71 0.000 .4469319 .5855501
__00000L | .8659743 .6635963 1.30 0.194 -.4441445 2.176093
__000016 | 58.81886 45.06966 1.31 0.194 -30.16085 147.7986
__00001U | .9999997 9.03e-07 1.1e+06 0.000 .9999979 1.000001
__00001X | -4.87e-08 9.65e-07 -0.05 0.960 -1.95e-06 1.86e-06
__00000N | 151.295 106.8069 1.42 0.158 -59.57083 362.1609
ln_distcap | 1.402371 .9924473 1.41 0.160 -.5569892 3.36173
------------------------------------------------------------------------------
Warning: estimated covariance matrix of moment conditions not of full rank.
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
------------------------------------------------------------------------------
Included instruments: __00000E __00000J __00000M __00000P __00000S __00000V
__00000Y __000011 __000014 __000017 __00001A __00001D
__00001G __00001J __00001M __00001P __00001S __00001Z
__000020 __00000I __00000L __000016 __00001U __00001X
__00000N ln_distcap
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.9958
Test of excluded instruments:
F( 7, 167) = 2.0e+11
Prob > F = 0.0000
First-stage regression of __00001Y:
OLS estimation
--------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on pairs1
Number of clusters (pairs1) = 168 Number of obs = 2517
F( 26, 167) = 1.6e+12
Prob > F = 0.0000
Total (centered) SS = 159.7663429 Centered R2 = 0.9998
Total (uncentered) SS = 160.0142873 Uncentered R2 = 0.9998
Residual SS = .0301100832 Root MSE = .003477
------------------------------------------------------------------------------
| Robust
__00001Y | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
__00000E | 294.893 57.42321 5.14 0.000 181.524 408.262
__00000J | .0115646 .0334573 0.35 0.730 -.0544891 .0776183
__00000M | -6.176372 .482563 -12.80 0.000 -7.129082 -5.223662
__00000P | -3.91e-09 1.21e-08 -0.32 0.748 -2.79e-08 2.00e-08
__00000S | -3.39e-09 4.52e-08 -0.07 0.940 -9.26e-08 8.58e-08
__00000V | -3.43e-09 4.57e-08 -0.07 0.940 -9.36e-08 8.68e-08
__00000Y | -3.38e-09 3.94e-08 -0.09 0.932 -8.11e-08 7.43e-08
__000011 | -3.51e-09 5.17e-08 -0.07 0.946 -1.06e-07 9.86e-08
__000014 | -3.47e-09 4.97e-08 -0.07 0.944 -1.02e-07 9.46e-08
__000017 | 18.64497 2.276844 8.19 0.000 14.14987 23.14008
__00001A | -4.17e-09 4.56e-08 -0.09 0.927 -9.41e-08 8.58e-08
__00001D | -3.76e-09 6.46e-08 -0.06 0.954 -1.31e-07 1.24e-07
__00001G | -2.99e-09 1.04e-07 -0.03 0.977 -2.09e-07 2.03e-07
__00001J | -2.05e-09 1.44e-07 -0.01 0.989 -2.87e-07 2.83e-07
__00001M | -2.25e-09 1.55e-07 -0.01 0.988 -3.08e-07 3.03e-07
__00001P | -1.90e-09 1.68e-07 -0.01 0.991 -3.33e-07 3.29e-07
__00001S | -1.50e-09 2.03e-07 -0.01 0.994 -4.03e-07 4.00e-07
__00001Z | -.7893339 .1997606 -3.95 0.000 -1.183716 -.3949522
__000020 | 40.77849 7.974215 5.11 0.000 25.03523 56.52176
__00000I | -.0115646 .0334572 -0.35 0.730 -.0776182 .0544891
__00000L | 6.176372 .482563 12.80 0.000 5.223662 7.129082
__000016 | -18.64498 2.276856 -8.19 0.000 -23.14011 -14.14985
__00001U | -3.81e-09 2.21e-07 -0.02 0.986 -4.39e-07 4.32e-07
__00001X | 1 2.34e-07 4.3e+06 0.000 .9999995 1
__00000N | -27.50739 5.438 -5.06 0.000 -38.24347 -16.7713
ln_distcap | -.2601827 .0505418 -5.15 0.000 -.3599658 -.1603996
------------------------------------------------------------------------------
Warning: estimated covariance matrix of moment conditions not of full rank.
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
------------------------------------------------------------------------------
Included instruments: __00000E __00000J __00000M __00000P __00000S __00000V
__00000Y __000011 __000014 __000017 __00001A __00001D
__00001G __00001J __00001M __00001P __00001S __00001Z
__000020 __00000I __00000L __000016 __00001U __00001X
__00000N ln_distcap
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.9971
Test of excluded instruments:
F( 7, 167) = 3.3e+12
Prob > F = 0.0000
First-stage regression of __000021:
OLS estimation
--------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on pairs1
Number of clusters (pairs1) = 168 Number of obs = 2517
F( 26, 167) = .
Prob > F = .
Total (centered) SS = .001764719 Centered R2 = 0.0149
Total (uncentered) SS = .0017968437 Uncentered R2 = 0.0325
Residual SS = .0017384285 Root MSE = .000835
------------------------------------------------------------------------------
| Robust
__000021 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
__00000E | -2.690367 . . . . .
__00000J | -.0096089 . . . . .
__00000M | .1063051 . . . . .
__00000P | 2.23e-12 . . . . .
__00000S | 2.61e-11 . . . . .
__00000V | 2.59e-11 . . . . .
__00000Y | 2.28e-11 . . . . .
__000011 | 2.81e-11 . . . . .
__000014 | 2.74e-11 . . . . .
__000017 | .2186006 . . . . .
__00001A | 1.70e-11 . . . . .
__00001D | 3.31e-11 . . . . .
__00001G | 6.25e-11 . . . . .
__00001J | 1.02e-10 . . . . .
__00001M | 1.08e-10 . . . . .
__00001P | 1.20e-10 . . . . .
__00001S | 1.43e-10 . . . . .
__00001Z | .0015887 . . . . .
__000020 | -.392503 . . . . .
__00000I | .0096089 . . . . .
__00000L | -.1063051 . . . . .
__000016 | -.2186007 . . . . .
__00001U | -1.53e-10 . . . . .
__00001X | -3.55e-11 . . . . .
__00000N | .2439816 . . . . .
ln_distcap | .0025431 . . . . .
------------------------------------------------------------------------------
Warning: estimated covariance matrix of moment conditions not of full rank.
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
------------------------------------------------------------------------------
Included instruments: __00000E __00000J __00000M __00000P __00000S __00000V
__00000Y __000011 __000014 __000017 __00001A __00001D
__00001G __00001J __00001M __00001P __00001S __00001Z
__000020 __00000I __00000L __000016 __00001U __00001X
__00000N ln_distcap
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.0125
Test of excluded instruments:
F( 0, 167) = .
Prob > F = .
Summary results for first-stage regressions
-------------------------------------------
Variable | Shea Partial R2 | Partial R2 | F( 7, 167) P-value
__00001V | 0.9933 | 0.9958 | 2.0e+11 0.0000
__00001Y | 0.9776 | 0.9971 | 3.3e+12 0.0000
__000021 | 0.0122 | 0.0125 | . .
NB: first-stage F-stat cluster-robust
Underidentification tests
Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified)
Ha: matrix has rank=K1 (identified)
Kleibergen-Paap rk LM statistic Chi-sq(5)=2.77 P-val=0.7353
Kleibergen-Paap rk Wald statistic Chi-sq(5)=. P-val= .
Weak identification test
Ho: equation is weakly identified
Kleibergen-Paap Wald rk F statistic .
See main output for Cragg-Donald weak id test critical values
Weak-instrument-robust inference
Tests of joint significance of endogenous regressors B1 in main equation
Ho: B1=0 and overidentifying restrictions are valid
Anderson-Rubin Wald test F(7,167)= 16.41 P-val=0.0000
Anderson-Rubin Wald test Chi-sq(7)=116.70 P-val=0.0000
Stock-Wright LM S statistic Chi-sq(7)=149.96 P-val=0.0000
NB: Underidentification, weak identification and weak-identification-robust
test statistics cluster-robust
Number of clusters N_clust = 168
Number of observations N = 2517
Number of regressors K = 22
Number of instruments L = 26
Number of excluded instruments L1 = 7
IV (2SLS) estimation
--------------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on pairs1
Number of clusters (pairs1) = 168 Number of obs = 2517
F( 22, 167) = 206.75
Prob > F = 0.0000
Total (centered) SS = 1617.186606 Centered R2 = 0.6816
Total (uncentered) SS = 1651.722694 Uncentered R2 = 0.6883
Residual SS = 514.8476128 Root MSE = .4543
------------------------------------------------------------------------------
| Robust
__00000F | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
__00001V | 1.570001 .2646353 5.93 0.000 1.047539 2.092462
__00001Y | -1.466572 .4010023 -3.66 0.000 -2.258259 -.6748846
__000021 | -128.1563 112.4465 -1.14 0.256 -350.1562 93.84362
__00000E | -13.86337 14.29162 -0.97 0.333 -42.0789 14.35216
__00000J | -.6627835 .4548794 -1.46 0.147 -1.560839 .2352717
__00000M | .2074742 .0617934 3.36 0.001 .0854772 .3294712
__00000P | .0609254 .0493406 1.23 0.219 -.0364863 .1583371
__00000S | .0825058 .0825237 1.00 0.319 -.0804183 .2454299
__00000V | .0995982 .0854717 1.17 0.246 -.069146 .2683424
__00000Y | .2799819 .0823949 3.40 0.001 .1173121 .4426518
__000011 | .3832506 .0983113 3.90 0.000 .1891575 .5773438
__000014 | .4761469 .0998267 4.77 0.000 .2790621 .6732318
__000017 | .6613286 .1012605 6.53 0.000 .461413 .8612443
__00001A | .7126187 .1066265 6.68 0.000 .502109 .9231284
__00001D | .7121479 .1173461 6.07 0.000 .4804749 .9438209
__00001G | .6235858 .1556122 4.01 0.000 .3163652 .9308064
__00001J | .3948429 .2041333 1.93 0.055 -.0081716 .7978574
__00001M | .4205724 .2167743 1.94 0.054 -.0073988 .8485436
__00001P | .4553149 .2401146 1.90 0.060 -.0187363 .9293661
__00001S | .2717456 .2840648 0.96 0.340 -.2890753 .8325665
__00001Z | -1.177078 3.488846 -0.34 0.736 -8.065006 5.71085
__000020 | -1.021229 1.758372 -0.58 0.562 -4.492731 2.450274
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic): 2.771
Chi-sq(5) P-val = 0.7353
------------------------------------------------------------------------------
Weak identification test (Kleibergen-Paap rk Wald F statistic): .
Stock-Yogo weak ID test critical values: 5% maximal IV relative bias 13.95
10% maximal IV relative bias 8.50
20% maximal IV relative bias 5.56
30% maximal IV relative bias 4.44
Source: Stock-Yogo (2005). Reproduced by permission.
NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
------------------------------------------------------------------------------
Warning: estimated covariance matrix of moment conditions not of full rank.
overidentification statistic not reported, and
standard errors and model tests should be interpreted with caution.
Possible causes:
number of clusters insufficient to calculate robust covariance matrix
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem.
------------------------------------------------------------------------------
Instrumented: __00001V __00001Y __000021
Included instruments: __00000E __00000J __00000M __00000P __00000S __00000V
__00000Y __000011 __000014 __000017 __00001A __00001D
__00001G __00001J __00001M __00001P __00001S __00001Z
__000020
Excluded instruments: __00000I __00000L __000016 __00001U __00001X __00000N
ln_distcap
Dropped collinear: __00000O __00000R __00000U __00000X __000010 __000013
__000019 __00001C __00001F __00001I __00001L __00001O
__00001R __00000H __00000K __00000Q __00000T __00000W
__00000Z __000012 __000015 __000018 __00001B __00001E
__00001H __00001K __00001N __00001Q contig
Reclassified as exog: __00000J __00000M __00000P __00000S __00000V __00000Y
__000011 __000014 __000017 __00001A __00001D __00001G
__00001J __00001M __00001P __00001S __00001Z __000020
------------------------------------------------------------------------------
xtoverid error: internal reestimation of eqn differs from original
r(198);
3- xi: xthtaylor ln_exports_od ln_gdp ln_sgdp dgdppc contig comlang_off ln_distcap euin i.year, endog( ln_gdp ln_sgdp dgdppc comlang_off)
xtoverid, noi detected collinearities
Similar output above and its overidentication test result;
Sargan statistics: 0.113
Chi-sq(3) P-val = 0.9903
xtoverid, noi cluster(pairs1) also says "warning: estimated covariance matrix of moment conditions not of full rank." This also happened when command xthtaylor with vce(bootstrap). Both of them don't report Sargan test statistics.
In that sense, what should I do? can I believe my Sargan test after xtoverid although collinearity exists? what about collinearity since I can't use either bootstrap or xtoverid, cluster(pairs1). Can I use Sargan statistics after xtoverid, noi for my Hausman-Taylor overidentification test.
I will appreciate if you help me.
Thanks
> Subject: st: RE: Hausman-Taylor AR(1) estimator
> Date: Thu, 6 Sep 2012 20:45:27 +0100
> From: [email protected]
> To: [email protected]
>
> Koray,
>
> You'll need to show us the output with the error message about the lack
> of full rank etc. before we can offer advice about what it means and how
> to deal with it.
>
> You can use xtoverid to get cluster-robust SEs; this is a way of
> addressing the within-panel autocorrelation. This post briefly
> describes how:
>
> http://www.stata.com/statalist/archive/2011-03/msg00414.html
>
> --Mark
>
> > -----Original Message-----
> > From: [email protected] [mailto:owner-
> > [email protected]] On Behalf Of KORAY ERCIHAN
> > Sent: 06 September 2012 17:30
> > To: [email protected]
> > Subject: st: Hausman-Taylor AR(1) estimator
> >
> > Dear Statalist,
> >
> > The issues related to Hausman and Taylor estimator have been indicated
> in
> > Statalist but I couldn't find a solution for my problem.
> > I am using panel data including 168 bilateral trade relations under
> 1993-2007
> > years. I want to estimate Hausman-Taylor estimator (HT). Since I find
> > autocorrelation in my data, I want to estimate HT AR(1) too.
> >
> > I estimated two HT models for my variables. One of them (HT1) is
> having 3
> > variables in endog part and the other (HT2) has 4 variables in endog
> part. I
> > have 7 explanatory variables at hand and I'm using time dummies as
> well. The
> > overidentification test results; 0.11 with p-value: 0.9903 and 0.12
> with p-
> > value: 0.9983 for HT1 and HT2 respectively. However, I obtained these
> test
> > results after commanding "xtoverid, noi" but when I use bootstrap
> option for
> > xthtaylor, it gives an error about the lack of full rank for
> covariance matrix
> > then I can't get the overidentification test result.
> >
> > Do you think I can rely on my overidentification results since their
> p-values
> > are very high? and how can I solve the autocorrelation problem?
> >
> > I will very grateful if you share your knowledge.
> >
> > Thanks
> >
> > Koray
> >
> > *
> > * 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 the Sunday Times
> Scottish University of the Year 2011-2012
>
> 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/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/