我用eviews5.1进行了豪斯曼检验,但结果我不是很确定。我把结果贴上来,各位看是不是说拒绝原假设所以应该用固定效应?但我做出来的随机效应比较符合文章的假设,8个变量中有5个通过了t检验,而固定效应只有4个通过t检验并且关键的变量没通过检验。且我看教材上有说过,Taylor认为若T>=3,N-K>=9(其中T时间序列的数目,N横截面单元的数目,K表示回归元的个数),并且坚持随机效应模型的基本假设,则随机效应模型的估计量比固定效应模型的估计量更有效。模型是符合Taylor的条件的。各位说我应该选哪个模型好呢?非常感激!!!
Correlated Random Effects - Hausman Test
Pool: POOL01
Test cross-section random effects
Test Summary Chi-Sq. Statistic Chi-Sq. d.f. Prob.
Cross-section random 90.029116 8 0.0000
Cross-section random effects test comparisons:
Variable Fixed Random Var(Diff.) Prob.
?I 18.434235 10.750042 0.754516 0.0000
?L 11.072948 12.613884 0.497469 0.0289
?N 0.265837 0.301132 0.001244 0.3170
?X1 1.230829 1.636230 0.424397 0.5337
?X2 -0.085753 -0.094218 0.000113 0.4264
?E1 -0.001655 -0.048802 0.000531 0.0408
?E2 0.011645 0.005630 0.000054 0.4151
?E3 2.250113 4.188771 0.581123 0.0110
Cross-section random effects test equation:
Dependent Variable: ?Y
Method: Panel Least Squares
Date: 03/23/07 Time: 09:31
Sample: 1998 2005
Included observations: 8
Cross-sections included: 31
Total pool (unbalanced) observations: 244
Variable Coefficient Std. Error t-Statistic Prob.
C 2.770504 0.570523 4.856076 0.0000
?I 18.43424 1.426041 12.92687 0.0000
?L 11.07295 3.214245 3.444961 0.0007
?N 0.265837 0.150656 1.764527 0.0791
?X1 1.230829 3.117485 0.394815 0.6934
?X2 -0.085753 0.038666 -2.217797 0.0277
?E1 -0.001655 0.054974 -0.030110 0.9760
?E2 0.011645 0.038178 0.305016 0.7607
?E3 2.250113 1.474731 1.525778 0.1286
Effects Specification
Cross-section fixed (dummy variables)
R-squared 0.647070 Mean dependent var 10.50728
Adjusted R-squared 0.581649 S.D. dependent var 2.179653
S.E. of regression 1.409800 Akaike info criterion 3.670287
Sum squared resid 407.4450 Schwarz criterion 4.229260
Log likelihood -408.7750 F-statistic 9.890854
Durbin-Watson stat 1.175569 Prob(F-statistic) 0.000000
Pool: POOL01
Test cross-section random effects
Test Summary Chi-Sq. Statistic Chi-Sq. d.f. Prob.
Cross-section random 90.029116 8 0.0000
Cross-section random effects test comparisons:
Variable Fixed Random Var(Diff.) Prob.
?I 18.434235 10.750042 0.754516 0.0000
?L 11.072948 12.613884 0.497469 0.0289
?N 0.265837 0.301132 0.001244 0.3170
?X1 1.230829 1.636230 0.424397 0.5337
?X2 -0.085753 -0.094218 0.000113 0.4264
?E1 -0.001655 -0.048802 0.000531 0.0408
?E2 0.011645 0.005630 0.000054 0.4151
?E3 2.250113 4.188771 0.581123 0.0110
Cross-section random effects test equation:
Dependent Variable: ?Y
Method: Panel Least Squares
Date: 03/23/07 Time: 09:31
Sample: 1998 2005
Included observations: 8
Cross-sections included: 31
Total pool (unbalanced) observations: 244
Variable Coefficient Std. Error t-Statistic Prob.
C 2.770504 0.570523 4.856076 0.0000
?I 18.43424 1.426041 12.92687 0.0000
?L 11.07295 3.214245 3.444961 0.0007
?N 0.265837 0.150656 1.764527 0.0791
?X1 1.230829 3.117485 0.394815 0.6934
?X2 -0.085753 0.038666 -2.217797 0.0277
?E1 -0.001655 0.054974 -0.030110 0.9760
?E2 0.011645 0.038178 0.305016 0.7607
?E3 2.250113 1.474731 1.525778 0.1286
Effects Specification
Cross-section fixed (dummy variables)
R-squared 0.647070 Mean dependent var 10.50728
Adjusted R-squared 0.581649 S.D. dependent var 2.179653
S.E. of regression 1.409800 Akaike info criterion 3.670287
Sum squared resid 407.4450 Schwarz criterion 4.229260
Log likelihood -408.7750 F-statistic 9.890854
Durbin-Watson stat 1.175569 Prob(F-statistic) 0.000000