原帖见 https://bbs.pinggu.org/thread-999752-1-1.html
怀特检验的结果是
White Heteroskedasticity Test:
F-statistic 4.148468 Probability 0.011832
Obs*R-squared 11.60896 Probability 0.020509
Test Equation:
Dependent Variable: RESID^2
Method: Least Squares
Date: 12/27/10 Time: 22:08
Sample: 1 27
Included observations: 27
Variable Coefficient Std. Error t-Statistic Prob.
C -79700.74 75538.40 -1.055102 0.3028
X1 1.295425 1.026914 1.261473 0.2204
X1^2 -0.000116 0.000292 -0.398070 0.6944
X2 1490.651 1383.797 1.077218 0.2931
X2^2 -6.978429 6.337612 -1.101113 0.2827
R-squared 0.429961 Mean dependent var 1027.797
Adjusted R-squared 0.326318 S.D. dependent var 1482.430
S.E. of regression 1216.751 Akaike info criterion 17.21133
Sum squared resid 32570623 Schwarz criterion 17.45130
Log likelihood -227.3530 F-statistic 4.148468
Durbin-Watson stat 2.065441 Prob(F-statistic) 0.011832
针对是不是存在异方差有两种观点:
其一是存在,因为看
F-statistic 4.148468 Probability 0.011832
Obs*R-squared 11.60896 Probability 0.020509
来说,p值落在接受域内,因此接受原假设,承认有异方差的存在
其二观点是不存在,因为看
Variable Coefficient Std. Error t-Statistic Prob.
C -79700.74 75538.40 -1.055102 0.3028
X1 1.295425 1.026914 1.261473 0.2204
X1^2 -0.000116 0.000292 -0.398070 0.6944
X2 1490.651 1383.797 1.077218 0.2931
X2^2 -6.978429 6.337612 -1.101113 0.2827
来说,所有的交叉项和独立项的p值都不显著,再加上这里是小样本数据因此否认存在异方差。
那么上述两种观点哪个是正确的?