发布时间 : 星期六 文章计量经济学更新完毕开始阅读4a60b0cb58fafab068dc0226
Sum squared resid Log likelihood Durbin-Watson stat Unweighted Statistics R-squared
Adjusted R-squared S.E. of regression Durbin-Watson stat Dependent Variable: Y Method: Least Squares
4970900. Schwarz criterion -229.7565 F-statistic 1.084119 Prob(F-statistic)
15.04455 451.4251 0.000000
820.9407 3435565.
0.830077 Mean dependent var 1250.323 0.824217 S.D. dependent var 344.1915 Sum squared resid 0.530370
表三
Date: 05/14/15 Time: 09:11 Sample: 1 31
Included observations: 31 Weighting series: W4
Variable C X
Weighted Statistics R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Unweighted Statistics R-squared
Adjusted R-squared S.E. of regression Durbin-Watson stat Coefficient -683.7572 0.085493
Std. Error 8.557345 0.000673 t-Statistic -79.90296 127.0520 Prob. 0.0000 0.0000 1936.539 6.642377 6.734892 2666676. 0.000000 820.9407 1788998.
0.999989 Mean dependent var 396.9977 0.999989 S.D. dependent var 6.495303 Akaike info criterion 1223.480 Schwarz criterion -100.9568 F-statistic 1.830314 Prob(F-statistic) 0.911516 Mean dependent var 1250.323 0.908465 S.D. dependent var 248.3739 Sum squared resid 0.902817 表4
⒊对所估计的模型再进行White检验,观察异方差的调整情况 如下图对表4分析
White Heteroskedasticity Test: F-statistic Obs*R-squared
Test Equation:
Dependent Variable: STD_RESID^2 Method: Least Squares Date: 05/14/15 Time: 09:30 Sample: 1 31
2.213648 Probability 4.232428 Probability
0.128074 0.120487
Included observations: 31 Variable C X X^2 R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient -34405.69 6.564101 9.25E-05 Std. Error 318449.4 31.17507 0.000676 t-Statistic -0.108041 0.210556 0.136700 Prob. 0.9147 0.8348 0.8922 270423.4 27.86733 28.00610 2.213648 0.128074 0.136530 Mean dependent var 166949.1 0.074854 S.D. dependent var 260105.5 Akaike info criterion 1.89E+12 Schwarz criterion -428.9436 F-statistic 1.848441 Prob(F-statistic)
四、实践结果报告:
1、用图示法初步判断是否存在异方差:被解释变量Y随着解释变量X的增大而逐渐分散,离散程度越来越大;同样的,残差平方ei对解释变量X的散点图主要分布在图形中的下三角部分,大致看出残差平方ei随Xi的变动呈增大趋势。因此,模型很可能存在异方差。但是否确实存在异方差还应该通过更近一步的检验。
再用White检验异方差:因为nR=4.898482> ?20.05(2)=4.22 ,所以拒绝原假设,不拒绝备择假设,这表明模型存在异方差。
2、用加权最小二乘法修正异方差:
发现用权数?2t的效果最好,则估计结果为:
222?= -34405.69 + 6.564101Xi Yi(1.863374) (9.725260)
R2=0.136530
括号中的数据为t统计量值。
由上可以看出,R=0.922715,拟合程度较好。在给定?=0.0时,t=9.725260 > t0.025(26)=2.056 ,拒绝原假设,说明收入对储蓄有显著性影响。 F=94.58068 > F0.05(1,26)= 4.23 , 表明方程整体显著。
运用加权最小二乘法后,参数?2的t检验显著,可决系数提高了不少,F检验也显著,并说明收入每增长1元,储蓄平均增长0.024665元。
3、再用White检验修正后的模型是否还存在异方差:
White检验结果
2
由上看出
White Heteroskedasticity Test: F-statistic Obs*R-squared
Test Equation:
Dependent Variable: STD_RESID^2 Method: Least Squares Date: 05/14/15 Time: 09:30 Sample: 1 31
Included observations: 31 Variable C X X^2 R-squared
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat 22.213648 Probability 4.232428 Probability
0.128074 0.120487
Coefficient -34405.69 6.564101 9.25E-05 Std. Error 318449.4 31.17507 0.000676 t-Statistic -0.108041 0.210556 0.136700 Prob. 0.9147 0.8348 0.8922 270423.4 27.86733 28.00610 2.213648 0.128074 0.136530 Mean dependent var 166949.1 0.074854 S.D. dependent var 260105.5 Akaike info criterion 1.89E+12 Schwarz criterion -428.9436 F-statistic 1.848441 Prob(F-statistic) ,nR= 5.628058 ,由White检验知,在?=0,05下,查?2分布表,得临界值:
?20.05(2)=5.99147。
比较计算的?2统计量与临界值,因为nR= 5.628058 < ?20.05(2)=5.99147 ,所以接受原假设,这说明修正后的模型不存在异方差。
2