Eviews8.0多元回归后怀特检验有异方差，如何修正？

16565

收藏 2016-05-07

EXPO GDP DIS已经是lnExpo等命名为EXPO等
Dependent Variable: EXPO
Method: Least Squares
Date: 05/07/16 Time: 13:08
Sample: 1 60
Included observations: 60

Variable Coefficient Std. Error t-Statistic Prob.

C -1.281757 2.807124 -0.456609 0.6497
DIS -1.410979 0.278825 -5.060452 0.0000
GDP 1.038641 0.067791 15.32120 0.0000

R-squared 0.813505       Mean dependent var 12.60872
Adjusted R-squared 0.806961       S.D. dependent var 1.872096
S.E. of regression 0.822528       Akaike info criterion 2.495839
Sum squared resid 38.56350       Schwarz criterion 2.600556
Log likelihood -71.87516       Hannan-Quinn criter. 2.536799
F-statistic 124.3187       Durbin-Watson stat 1.838718
Prob(F-statistic) 0.000000

怀特检验
Heteroskedasticity Test: White

F-statistic 7.357277       Prob. F(2,57) 0.0014
Obs*R-squared 12.31094       Prob. Chi-Square(2) 0.0021
Scaled explained SS 20.23500       Prob. Chi-Square(2) 0.0000

Test Equation:
Dependent Variable: RESID^2
Method: Least Squares
Date: 05/07/16 Time: 13:11
Sample: 1 60
Included observations: 60

Variable Coefficient Std. Error t-Statistic Prob.

C 7.607670 1.976573 3.848919 0.0003
DIS^2 -0.045332 0.022877 -1.981533 0.0524
GDP^2 -0.005782 0.001849 -3.126407 0.0028

R-squared 0.205182       Mean dependent var 0.642725
Adjusted R-squared 0.177294       S.D. dependent var 1.237007
S.E. of regression 1.122004       Akaike info criterion 3.116817
Sum squared resid 71.75696       Schwarz criterion 3.221535
Log likelihood -90.50452       Hannan-Quinn criter. 3.157778
F-statistic 7.357277       Durbin-Watson stat 2.238072
Prob(F-statistic) 0.001438

扫码加我拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

全部回复

胖胖小龟宝

2016-5-7 17:48:42

第一种是在回归前先将变量进行对数处理，能够很好的减少数据波动以及异方差的问题；第二种是加权最小二乘法（wls）,quick-estimate equation界面,点击option,权重那块,填入所选择的权重就可以了。权重的选择上，可选用1/√X,1/X, 1/X^2,X,√X,X^2（生成变量W分别等于这些权重），之后将加权后的方程再进行怀特检验，选一个最好的权重就ok啦。