Durbin-Watson stat 1.444093 Prob(F-statistic) 0.000000
Y =40826.16+ X0.758419X2+ 10.01565X3+（-0.069142）X4+(-1.760365)X6 t= (2.760067) (0.408914) (4.656038) (0.520955) (-3.414820)
R2=0.992440
可以看出个因素的T统计量都得到了不同程度的改善。
在前一模型的基础上剔出X6，拟合优度变差，但对C的t值影响很大，统计检验t=-0.799100，不显著。而且X4的系数为负，与经济意义相悖。
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 12:59
Sample: 1989 2003 Included observations: 15
Variable C X2 X3 X4 R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient Std. Error -6120.442 4.679794 6.016219 -0.286073 0.983624 0.979157 1184.370 15430045 -125.1124 1.460596 7659.170 2.043819 2.531957 0.163526 t-Statistic -0.799100 2.289730 2.376114 -1.749398 Prob. 0.4411 0.0428 0.0368 0.1080 18793.77 8203.735 17.21499 17.40380 220.2341 0.000000 Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion
F-statistic Prob(F-statistic)
剔出X2进行回归，X4不但经济意义违背而且T统计值较小，不能通过检验
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 13:01
Sample: 1989 2003 Included observations: 15 Variable C X3 X4 X6 R-squared Adjusted R-squared
Coefficient Std. Error 42613.82 10.81935 -0.019650 -1.890879 0.992313 0.990217
13585.49 0.840499 0.052361 0.389190 t-Statistic 3.136714 12.87253 -0.375287 -4.858501 Prob. 0.0095 0.0000 0.7146 0.0005 18793.77 8203.735
Mean dependent var S.D. dependent var
S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
811.4261 7242535. -119.4398 1.382173
Akaike info criterion Schwarz criterion
F-statistic Prob(F-statistic)
16.45864 16.64746 473.3484 0.000000
剔出X4进行回归虽然拟合优度略有改善，但X2的T统计值为-0.166847，通不过检验，应剔出
X2在做回归。而其他因素的统计值都较好。
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 13:00 Sample: 1989 2003 Included observations: 15 Variable C X2 X3 X6
R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
Coefficient 42252.90 -0.122707 10.78630 -1.888906
Std. Error 14046.33 0.735446 1.509109 0.437375
t-Statistic 3.008110 -0.166847 7.147459 -4.318737
Prob. 0.0119 0.8705 0.0000 0.0012 8203.735 16.46884 16.65765 468.5100 0.000000
0.992235 Mean dependent var 18793.77 0.990117 S.D. dependent var 815.5728 Akaike info criterion 7316750. Schwarz criterion -119.5163 F-statistic 1.360316 Prob(F-statistic)
综合考虑所得结果，选择含有X2 X3X6 这三个因素的模型。 再做剔出X2的模型的参数估计：
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 13:06 Sample: 1989 2003 Included observations: 15 Variable C X3 X6 R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood
Coefficient 41516.51 10.54410 -1.851909 Std. Error 12783.36 0.395538 0.361404 t-Statistic 3.247700 26.65764 -5.124202 Prob. 0.0070 0.0000 0.0003 8203.735 16.33803 16.47964 764.7022
0.992215 Mean dependent var 18793.77 0.990917 S.D. dependent var 781.8390 Akaike info criterion 7335267. Schwarz criterion -119.5352 F-statistic
Durbin-Watson stat 1.352428 Prob (F-statistic) 0.000000
可以看出拟合优度很好 F统计量的值在给定显著性水平α=0.05的情况下也较显著，C ，X3，
X6的T统计值也很显著，表明对Y的影响也很显著。
新模型估计结果：
Y =41516.51+ 10.54410X3 +(-1.851909) X6 t= (3.2477) (426.65764) (-5.124202) R2=0.992215
(2)异方差检验
①检验：
利用Goid_Quandt检验法检验模型是否存在异方差。
将时间定义为1989——1993，然后对Y CX3用OLS法求的下列结果： Y=-6225.673+5.317281X3
t= (-2.982843) (7.083533)
R2=0.943584
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 13:12 Sample: 1989 1993 Included observations: 5
Variable C X3
R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
?e21=634718.8
Coefficient -6225.673 5.317281
Std. Error 2087.161 0.750654
t-Statistic -2.982843 7.083533
Prob. 0.0585 0.0058
0.943584 Mean dependent var 8486.818 0.924779 S.D. dependent var 459.9706 Akaike info criterion 634718.8 Schwarz criterion -36.47344 F-statistic 1.632631 Prob(F-statistic)
1677.103 15.38938 15.23315 50.17643 0.005786
将时间定义为1999——2003，然后对Y CX3用OLS法求的下列结果
Y=-44209.20+16.62678X3
t= (-5.018903) (8.034508)
R2=0.955591
Dependent Variable: Y Method: Least Squares
?e22=777592.5
Date: 04/30/05 Time: 13:13 Sample: 1999 2003 Included observations: 5 Variable C X3 R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
Coefficient -44209.20 16.62678 Std. Error 8808.537 2.069421 t-Statistic -5.018903 8.034508 Prob. 0.0152 0.0040 2092.227 15.59240 15.43617 64.55332 0.004026
0.955591 Mean dependent var 26539.42 0.940788 S.D. dependent var 509.1144 Akaike info criterion 777592.5 Schwarz criterion -36.98099 F-statistic 1.966867 Prob(F-statistic)
22÷ee?2?1=777952.5/634718.8=1.22566481409 小于F0.05（4，4）=6.39 接受H0不存在异方差
将时间定义为1989——1993，然后对Y C X6用OLS法求的下列结果：
Y=-20445.55+0.864900X6
t= （-0.473093) (0.669589)
R2=0.130018
?e21=9787897
Dependent Variable: Y Method: Least Squares Date: 04/30/05 Time: 13:19 Sample: 1989 1993 Included observations: 5 Variable C X6 R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient -20445.55 0.864900 Std. Error 43216.71 1.291689 t-Statistic -0.473093 0.669589 Prob. 0.6684 0.5510 1677.103 18.12510 17.96887 0.448349 0.551047 0.130018 Mean dependent var 8486.818 -0.159975 S.D. dependent var 1806.276 Akaike info criterion 9787897. Schwarz criterion -43.31274 F-statistic 0.815079 Prob(F-statistic)
将时间定义为1999——2003，然后对Y C X6用OLS法求的下列结果
Y=126537.0+（-3.097615）X6
t= (48.84461) (-38.60686)
R2=0.997991
?e22=35171.96