1、量（万辆）度（万公里）198528.49105851323.543.7294.24198634.71107507337.836.9896.28198742.29109300356.947.1898.22198860.42111026383.464.4799.96198973.12112704381.458.35101.43199081.62114333394.451.4102.83199196.04115823427.171.42104.111992118.2117171482.3106.67105.671993155.77118517521.4129.85108.351994205.4211
2、9850543.8136.69111.781995249.96121121584.6145.27115.71996289.67122389637.8147.52118.581997358.36123626664.4158.25122.641998423.65124810701.3163127.851999512.35125836754.2174.3130.462000704.9126987786.1185.6139.162001821.64128036801.3209.5145.892002945.7130549843.1236.76152.7原始数据的分析记录在附录1由于原模型存在自相关，我
3、们决定通过改变解释变量模式从而消除由于样本数量不足带来的问题。新模型中：Y/X1Y 人均民用汽车私人拥有量X3/X1X1人均汽车产量X2X2 全国居民消费指数X4/X1X3 人均公路长度y/x1x3/x1x4/x10.0002690.0004130.000890.0003230.0003440.0008960.0003870.0004320.0008990.0005440.0005810.00090.0006490.0005180.0007140.000450.0008290.0006170.0010090.000910.0009020.0013140.0010960.0009140.0017
4、140.0011410.0009330.0020640.0011990.0009550.0023670.0012050.0009690.0028990.001280.0009920.0033940.0013060.0010240.0040720.0013850.0010370.0055510.0014620.0064170.0016360.0011390.0072440.0018140.00117三 平稳性检验Augmented dickey_fuller unit root test on D （Y1，2）ADF Test Statistic-4.437656 1% Critical Val
5、ue*-4.9893 5% Critical Value-3.8730 10% Critical Value-3.3820*MacKinnon critical values for rejection of hypothesis of a unit root.Augmented Dickey-Fuller Test EquationDependent Variable: D（Y,3）Method: Least SquaresDate: 06/15/05 Time: 20:08Sample（adjusted）: 1991 2002Included observations: 12 after
6、adjusting endpointsVariableCoefficientStd. Errort-StatisticProb. D（Y（-1）,2）-4.3758780.9860790.0044D（Y（-1）,3）2.9263250.9167543.1920510.0188D（Y（-2）,3）1.5436610.8190691.8846530.1085D（Y（-3）,3）2.2029800.5752423.8296600.0087C-0.0002910.000166-1.7485570.1310TREND（1985）5.07E-051.75E-052.9026000.0272R-square
7、d0.963765 Mean dependent var-4.17E-09Adjusted R-squared0.933569 S.D. dependent var0.000517S.E. of regression0.000133 Akaike info criterion-14.70026Sum squared resid1.07E-07 Schwarz criterion-14.45781Log likelihood94.20157 F-statistic31.91688Durbin-Watson stat2.451776 Prob（F-statistic）0.000300Augment
8、ed dickey_fuller unit root test on D （x1,1）-3.492868-4.8025-3.7921-3.3393 D（X1）18 1989 2002 14 after adjusting endpointsX1（-1）-0.9929380.2842760.0082D（X1（-1）1.0610820.2565354.1362170.0033D（X1（-2）-0.0393390.273775-0.1436910.8893D（X1（-3）0.4459430.2287311.9496380.08707.54E-051.4860210.17568.97E-052.50E
9、-053.5950670.00700.7797008.81E-050.6420130.0001026.08E-05-16.279362.96E-08-16.00547119.95555.6628332.5019180.015901 Augmented dickey_fuller unit root test on D （x2,2）-3.882038-4.7315-3.7611-3.3228 D（X2,3） 06/14/05 Time: 10:56 1988 2002 15 after adjusting endpointsD（X2（-1）,2）-1.1995810.3090080.00223.
10、89526513.935620.2795190.7846-0.2369961.283314-0.1846750.85660.5632441.4533330.49045229.4253021.004569.1042135294.3019.245823-65.281607.7376522.1520350.006941Augmented dickey_fuller unit root test on D （x3）-4.293548-4.6712-3.7347-3.3086 D（X3,2）23 1987 2002 16 after adjusting endpointsD（X3（-1）-1.15101
11、60.268080-1.36E-056.83E-06-1.9868610.06843.48E-069.74E-073.5728920.00340.5866221.56E-060.5230251.50E-051.04E-05-19.947951.40E-09-19.80309162.58369.2240932.1281250.003208可以看出，解释变量X1在0阶差分，X2在2阶差分， X3 在1阶差分，Y在2阶差分情况下通过了平稳性检验。四 Granger Causality检验Pairwise Granger Causality Tests26Sample: 1985 2002 Lags:
12、 1 Null Hypothesis:ObsF-StatisticProbability X1 does not Granger Cause Y17 1.27559 0.27770 Y does not Granger Cause X1 1.34361 0.26579滞后长度1 结论：X1Y 拒绝 Y X1 拒绝27 X2 does not Granger Cause Y 5.20861 0.03863 Y does not Granger Cause X2 1.54629 0.23411X2Y 接受 Y X2 拒绝29 2 X3 does not Granger Cause Y16 4.39
13、300 0.03960 Y does not Granger Cause X3 5.59132 0.02111滞后长度2 结论：X3Y 接受 Y X3 接受五 参数的估计 Y25 1985 2002 18-0.0193820.000828-23.416700.0000X10.6329890.3954581.6006470.1318X2-3.54E-071.33E-06-0.2655860.7944X321.988031.21339618.121060.9959340.0023200.9950630.0021950.000154-14.523553.33E-07-14.32569134.7120
14、1143.1750.9978840.000000得Y=-0.019382+0.632989X1-3.54E-07X2+21.98803X3 T （-23.41670）（1.600647）（-0.265586）（18.12106） 可决系数R2=0.995934 修正值为0.995063 从拟合图形看出（特别是残差图）做出的模型拟合效果不是非常理想，所以决定通过检验来进一步修正该模型。六 模型的检验和修正1.经济意义检验 首先在经济意义上通不过。X2的系数为负，全国居民消费水平指数的提高，全国私人汽拥有量反而在下降。这显然与经济事实不符。2统计推断检验 从回归结果来看，模型的整体拟合度较高（可决系数R2=0.995934）,F统计量的值在给定显著性水平=0.05的情况下很显著（F=1143.175），但是C，X1，X2，的T统计值均不显著（它们的T统计量的值的绝对值均小于2），说明这些变量对Y的影响不显著，或者变量之间可能存在多重共线性的影响使其T值不显著。3计量经济学检验（1）多重共线性检验X1和Y回归 06/26/05 Time: 16:00-0.004192-7.2157311.18E-051.01E-0611.735610.89
