 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: ivreg2: difference in first regression results
I noticed that my ivreg2.ado for Stata8 which I use at home produces different
results (in the first regression: F test of excluded instruments etc.) than my
ivreg2.ado which I use in my office with Stata9. The results are shown below.
The second equation is exactly the same, but in the first t values and the F
statistic are very different. The whole thing is run with cluster and robust
controls which I thought were identical in both cases. But they obviously are
not. So am I doing something wrong?
(The more recent version of ivreg2 seems to use degrees of freedom based on
clusters, not on the number of observations which seems to be the reason why F
critical values in the first regression differ between the two versions.)
Any ideas why this is the case and which one of the version is "correct"?
Thanks a lot. Branko
USING IVREG2 FOR STATA 8
---------------------------------------------------------
. ivreg2 rlaw lngdppc90 war_cum fuel96 comyears polity2 (alt = cpdemocracy
urban_90) if year>1995, gmm first robu
> st cluster(contcod)
First-stage regressions
-----------------------
First-stage regression of alt:
Source | SS df MS Number of obs = 191
-------------+------------------------------ F( 7, 183) = 52.38
Model | 182.967522 7 26.1382175 Prob > F = 0.0000
Residual | 91.3256713 183 .499047384 R-squared = 0.6671
-------------+------------------------------ Adj R-squared = 0.6543
Total | 274.293194 190 1.44364839 Root MSE = .70643
------------------------------------------------------------------------------
alt | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lngdppc90 | -.4497775 .1950786 -2.31 0.022 -.8346698 -.0648852
war_cum | -.1374908 .0387878 -3.54 0.000 -.2140196 -.0609621
fuel96 | .2283111 .6113502 0.37 0.709 -.9778901 1.434512
comyears | -.0585286 .0103033 -5.68 0.000 -.0788571 -.0382001
polity2 | .0468558 .0153691 3.05 0.003 .0165323 .0771792
cpdemocracy | .6989806 .1595599 4.38 0.000 .384167 1.013794
urban_90 | .0111668 .0062011 1.80 0.073 -.0010681 .0234017
_cons | 7.950307 1.705855 4.66 0.000 4.584635 11.31598
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.1002
Test of excluded instruments:
F( 2, 183) = 10.19
Prob > F = 0.0001
Summary results for first-stage regressions:
Shea
Variable Partial R2 Partial R2 F( 2, 183) P-value
alt 0.1002 0.1002 10.19 0.0001
GMM estimation
--------------
Number of clusters (contcod) = 24 Number of obs = 191
Total (centered) SS = 87.75379385 Centered R2 = 0.7018
Total (uncentered) SS = 105.2330001 Uncentered R2 = 0.7513
Residual SS = 26.16757786 Root MSE = .37
------------------------------------------------------------------------------
| Robust
rlaw | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
alt | .3499127 .2774387 1.26 0.207 -.1938572 .8936826
lngdppc90 | .9954777 .1508931 6.60 0.000 .6997326 1.291223
war_cum | -.0844177 .0573268 -1.47 0.141 -.1967762 .0279407
fuel96 | -.2100487 .3116242 -0.67 0.500 -.8208209 .4007235
comyears | -.0163427 .0232756 -0.70 0.483 -.0619621 .0292767
polity2 | -.0243998 .0210711 -1.16 0.247 -.0656984 .0168989
_cons | -8.358753 2.63221 -3.18 0.001 -13.51779 -3.199716
------------------------------------------------------------------------------
Hansen J statistic (overidentification test of all instruments): 1.459
Chi-sq(1) P-val = 0.22717
------------------------------------------------------------------------------
Instrumented: alt
Instruments: lngdppc90 war_cum fuel96 comyears polity2 cpdemocracy urban_90
------------------------------------------------------------------------------
-------------------------------------------------------------------------------------------------------------------------------
USING IVREG2 FOR STATA9
----------------------------------------------------------
. ivreg2 rlaw lngdppc90 war_cum fuel96 comyears polity2 (alt = cpdemocracy
urban_90) if
> year>1995, gmm2s robust cluster(contcod) first
First-stage regressions
-----------------------
First-stage regression of alt:
OLS estimation
--------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on contcod
Number of clusters (contcod) = 24 Number of obs = 191
F( 7, 23) = 13.08
Prob > F = 0.0000
Total (centered) SS = 274.2931937 Centered R2 = 0.6671
Total (uncentered) SS = 841 Uncentered R2 = 0.8914
Residual SS = 91.32567134 Root MSE = .7064
------------------------------------------------------------------------------
| Robust
alt | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lngdppc90 | -.4497775 .3415128 -1.32 0.201 -1.156251 .2566955
war_cum | -.1374908 .079412 -1.73 0.097 -.301767 .0267854
fuel96 | .2283111 .9041596 0.25 0.803 -1.642086 2.098708
comyears | -.0585286 .0190362 -3.07 0.005 -.0979081 -.0191492
polity2 | .0468558 .0253491 1.85 0.077 -.0055828 .0992943
cpdemocracy | .6989806 .3894204 1.79 0.086 -.1065968 1.504558
urban_90 | .0111668 .0114127 0.98 0.338 -.0124422 .0347759
_cons | 7.950307 3.012348 2.64 0.015 1.71879 14.18182
------------------------------------------------------------------------------
Included instruments: lngdppc90 war_cum fuel96 comyears polity2 cpdemocracy
urban_90
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.1002
Test of excluded instruments:
F( 2, 23) = 2.17
Prob > F = 0.1370
Summary results for first-stage regressions
-------------------------------------------
Variable | Shea Partial R2 | Partial R2 | F( 2, 23) P-value
alt | 0.1002 | 0.1002 | 2.17 0.1370
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(2)=. P-val= .
Kleibergen-Paap rk Wald statistic Chi-sq(2)=. 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(2,23)= 1.39 P-val=0.2694
Anderson-Rubin Wald test Chi-sq(2)=3.01 P-val=0.2220
Stock-Wright LM S statistic Chi-sq(2)=2.52 P-val=0.2830
NB: Underidentification, weak identification and weak-identification-robust
test statistics cluster-robust
Number of clusters N_clust = 24
Number of observations N = 191
Number of regressors K = 7
Number of instruments L = 8
Number of excluded instruments L1 = 2
2-Step GMM estimation
---------------------
Estimates efficient for arbitrary heteroskedasticity and clustering on contcod
Statistics robust to heteroskedasticity and clustering on contcod
Number of clusters (contcod) = 24 Number of obs = 191
F( 6, 23) = 59.99
Prob > F = 0.0000
Total (centered) SS = 87.75379385 Centered R2 = 0.7018
Total (uncentered) SS = 105.2330001 Uncentered R2 = 0.7513
Residual SS = 26.16757785 Root MSE = .3701
------------------------------------------------------------------------------
| Robust
rlaw | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
alt | .3499127 .2774387 1.26 0.207 -.1938572 .8936826
lngdppc90 | .9954777 .1508931 6.60 0.000 .6997326 1.291223
war_cum | -.0844177 .0573268 -1.47 0.141 -.1967762 .0279407
fuel96 | -.2100487 .3116242 -0.67 0.500 -.8208209 .4007235
comyears | -.0163427 .0232756 -0.70 0.483 -.0619621 .0292767
polity2 | -.0243998 .0210711 -1.16 0.247 -.0656984 .0168989
_cons | -8.358753 2.63221 -3.18 0.001 -13.51779 -3.199716
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic): .
Chi-sq(2) P-val = .
------------------------------------------------------------------------------
Weak identification test (Kleibergen-Paap rk Wald F statistic): .
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): 1.459
Chi-sq(1) P-val = 0.2272
------------------------------------------------------------------------------
Instrumented: alt
Included instruments: lngdppc90 war_cum fuel96 comyears polity2
Excluded instruments: cpdemocracy urban_90
------------------------------------------------------------------------------
Development Research, World Bank
Email: [email protected] or branko_mi@yahoo.
tel: 202-473-6968
World Bank, Room MC 3-559
1818 H Street NW
Washington D.C. 20433
For "Worlds Apart" see
http://www.pupress.princeton.edu/titles/7946.html
Website:
http://econ.worldbank.org/projects/inequality
For papers see also:
http://econpapers.hhs.se/
http://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=149002
*
* 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/