dear all~先呈上我的步骤:
reg etr noofemployees capint gearing1 audittotalsales forsubsales big41not0 dumoil dummining if roce>=0.03
Source | SS df MS Number of obs = 342
-------------+------------------------------ F( 8, 333) = 11.64
Model | 1.07580661 8 .134475826 Prob > F = 0.0000
Residual | 3.84772688 333 .011554735 R-squared = 0.2185
-------------+------------------------------ Adj R-squared = 0.1997
Total | 4.92353348 341 .014438515 Root MSE = .10749
------------------------------------------------------------------------------
etr | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
noofemploy~s | 1.12e-08 7.65e-08 0.15 0.884 -1.39e-07 1.62e-07
capint | -.0454856 .0305143 -1.49 0.137 -.1055108 .0145396
gearing1 | -.0287844 .0425808 -0.68 0.500 -.1125457 .0549769
audittotal~s | .0531247 .0649089 0.82 0.414 -.0745585 .180808
forsubsales | -5.498491 31.58945 -0.17 0.862 -67.63852 56.64154
big41not0 | .079356 .0335961 2.36 0.019 .0132687 .1454434
dumoil | .2455811 .027769 8.84 0.000 .1909563 .300206
dummining | .0712409 .024491 2.91 0.004 .0230644 .1194174
_cons | .1658436 .0330159 5.02 0.000 .1008976 .2307896
------------------------------------------------------------------------------
. estat imtest, white
White's test for Ho: homoskedasticity
against Ha: unrestricted heteroskedasticity
chi2(39) = 81.76
Prob > chi2 = 0.0001
Cameron & Trivedi's decomposition of IM-test
---------------------------------------------------
Source | chi2 df p
---------------------+-----------------------------
Heteroskedasticity | 81.76 39 0.0001
Skewness | 3.93 8 0.8631
Kurtosis | 5.23 1 0.0222
---------------------+-----------------------------
Total | 90.92 48 0.0002
---------------------------------------------------
. reg etr noofemployees capint gearing1 audittotalsales forsubsales big41not0 dumoil dummining if roce>=0.03, robust
Linear regression Number of obs = 342
F( 8, 333) = 8.71
Prob > F = 0.0000
R-squared = 0.2185
Root MSE = .10749
------------------------------------------------------------------------------
| Robust
etr | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
noofemploy~s | 1.12e-08 3.37e-08 0.33 0.740 -5.51e-08 7.76e-08
capint | -.0454856 .0282476 -1.61 0.108 -.1010519 .0100807
gearing1 | -.0287844 .0388389 -0.74 0.459 -.1051848 .0476161
audittotal~s | .0531247 .0125581 4.23 0.000 .0284216 .0778279
forsubsales | -5.498491 17.34388 -0.32 0.751 -39.61587 28.61889
big41not0 | .079356 .0359637 2.21 0.028 .0086114 .1501006
dumoil | .2455811 .0453295 5.42 0.000 .1564129 .3347494
dummining | .0712409 .0237919 2.99 0.003 .0244395 .1180424
_cons | .1658436 .0352816 4.70 0.000 .0964407 .2352464
------------------------------------------------------------------------------
. estat imtest, white
White's test for Ho: homoskedasticity
against Ha: unrestricted heteroskedasticity
chi2(39) = 81.76
Prob > chi2 = 0.0001
Cameron & Trivedi's decomposition of IM-test
---------------------------------------------------
Source | chi2 df p
---------------------+-----------------------------
Heteroskedasticity | 81.76 39 0.0001
Skewness | 3.93 8 0.8631
Kurtosis | 5.23 1 0.0222
---------------------+-----------------------------
Total | 90.92 48 0.0002
---------------------------------------------------
我的问题是,我已经在第二次regression的时候加了robust了,为什么第二次white test的结果和第一次还是完全一样呢?In this case, how can I tell if the robust option reduces heteroscedasticity?
PS:本人stata 0 基础,被告知毕业论文要用,一筹莫展ing,请不吝赐教~谢谢:)