《多行线性回归》由会员分享，可在线阅读，更多相关《多行线性回归（5页珍藏版）》请在金锄头文库上搜索。
1、2）多重共线性检验计算各解释变量的相关系数：X1X2X3X4X11.0000000.5462420.750769-0.039316X20.5462421.0000000.772231-0.296705X30.7507690.7722311.000000-0.231755X4-0.039316-0.296705-0.2317551.000000由相关系数矩阵可知，存在一定多重共线。X1: Y = 0.155033739787 + 0.542628795297*X1Dependent Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:4
2、5Sample: 1989 2008Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C0.1550340.0588942.6324250.0169X10.5426290.2828351.9185360.0710R-squared0.169772Mean dependent var0.260534Adjusted R-squared0.123648S.D. dependent var0.100732S.E. of regression0.094299Akaike info criterion-1.7900
3、63Sum squared resid0.160060Schwarz criterion-1.690490Log likelihood19.90063Hannan-Quinn criter.-1.770626F-statistic3.680779Durbin-Watson stat0.554459Prob(F-statistic)0.071045X2：Y = 0.121822172818 + 0.019975730801*X2Dependent Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:45Sample: 1989 2008
4、Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C0.1218220.0496172.4552320.0245X20.0199760.0066103.0221710.0073R-squared0.336614Mean dependent var0.260534Adjusted R-squared0.299759S.D. dependent var0.100732S.E. of regression0.084293Akaike info criterion-2.014407Sum squared resi
5、d0.127894Schwarz criterion-1.914834Log likelihood22.14407Hannan-Quinn criter.-1.994969F-statistic9.133515Durbin-Watson stat0.356058Prob(F-statistic)0.007323X3: Y = 0.231379413507 + 0.483860396139*X3Dependent Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:46Sample: 1989 2008Included observat
6、ions: 20VariableCoefficientStd. Errort-StatisticProb.C0.2313790.0279378.2822820.0000X30.4838600.2944591.6432160.1177R-squared0.130441Mean dependent var0.260534Adjusted R-squared0.082133S.D. dependent var0.100732S.E. of regression0.096506Akaike info criterion-1.743779Sum squared resid0.167642Schwarz
7、criterion-1.644205Log likelihood19.43779Hannan-Quinn criter.-1.724341F-statistic2.700158Durbin-Watson stat0.487674Prob(F-statistic)0.117691X4: Y = -0.000318968014166 + 1.00250813034*X4Dependent Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:47Sample: 1989 2008Included observations: 20Variab
8、leCoefficientStd. Errort-StatisticProb.C-0.0003190.110690-0.0028820.9977X41.0025080.4182952.3966550.0276R-squared0.241912Mean dependent var0.260534Adjusted R-squared0.199796S.D. dependent var0.100732S.E. of regression0.090108Akaike info criterion-1.880965Sum squared resid0.146152Schwarz criterion-1.
9、781392Log likelihood20.80965Hannan-Quinn criter.-1.861527F-statistic5.743955Durbin-Watson stat0.535810Prob(F-statistic)0.027616综上可见，X2 是最重要的解释变量，所以我们选定两个基本回归方程作为初始的回归模型。X2、X1:Y = 0.104815956231 + 0.0174262180527*X2 + 0.178527771824*X1Dependent Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:
10、53Sample: 1989 2008Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C0.1048160.0584321.7938020.0907X20.0174260.0080402.1673080.0447X10.1785280.3075500.5804830.5692R-squared0.349507Mean dependent var0.260534Adjusted R-squared0.272979S.D. dependent var0.100732S.E. of regression0.0
11、85889Akaike info criterion-1.934034Sum squared resid0.125408Schwarz criterion-1.784674Log likelihood22.34034Hannan-Quinn criter.-1.904878F-statistic4.567019Durbin-Watson stat0.388566Prob(F-statistic)0.025856X1引入后R2增加，但是X1的t检验不显著。X2、X3:Y = 0.0994616103059 + 0.0256975568967*X2 - 0.28831232589*XDepende
12、nt Variable: YMethod: Least SquaresDate: 07/02/10 Time: 22:55Sample: 1989 2008Included observations: 20VariableCoefficientStd. Errort-StatisticProb.C0.0994620.0595611.6699130.1132X20.0256980.0105532.4350690.0262X3-0.2883120.410635-0.7021130.4921R-squared0.355308Mean dependent var0.260534Adjusted R-squared0.279462S.D. dependent var0.100732S.E. of regression0.085505Akaike info criterion-1.942992Sum s
