实验三 多元线性回归模型的 估计和检验.docx

1、实验三 多元线性回归模型的 估计和检验 实 验 报 告课程名称： 计量经济学 实验项目： 实验三 多元线性回归模型的 估计和检验 实验类型：综合性 设计性 验证性专业班别： 姓 名： 学 号： 实验课室： 指导教师： 实验日期： 2015-04-30 一、实验项目训练方案小组合作：是 否小组成员：无实验目的：掌握多元线性回归模型估计和检验的方法。实验场地及仪器、设备和材料实验室：普通配置的计算机，Eviews软件及常用办公软件。实验训练内容（包括实验原理和操作步骤）：【实验步骤】（一）国内生产总值的增长模型：分析广东省国内生产总值的增长，根据广东数据（数据见“表：广东省宏观经济数据-第三章.x

2、ls”文件，各变量的表示按照试验指导课本上的来表示）选择不变价GDP（GDPB）、不变价资本存量（ZC）和从业人员（RY），把GDPB作为被解释变量，ZC和RY作为两个解释变量进行二元线性回归分析。要求：按照试验指导课本，分别作：1作散点图（GDPB同ZC，GDPB同RY）（结果控制在本页）2进行因果关系检验（GDPB同ZC，GDPB同RY）（结果控制在本页）Pairwise Granger Causality TestsDate: 04/30/15 Time: 11:22Sample: 1978 2005Lags: 2Null Hypothesis:ObsF-StatisticProb.ZC

3、 does not Granger Cause GDPB263.849390.0376GDPB does not Graner Cause ZC19.07482.E-05Pairwise Granger Causality TestsDate: 04/30/15 Time: 11:23Sample: 1978 2005Lags: 3Null Hypothesis:ObsF-StatisticProb.RY does not Granger Cause GDPB252.887440.0641GDPB does not Granger Cause RY3.463090.0382从因果关系检验看，Z

4、C明显影响GDPB,RY不太明显，这是可以理解的，计划经济时期存在着隐性失业，使得劳动力的变化对产出的影响不太明显。3作GDPB同ZC和RY的多元线性回归，写出模型估计的结果，并分析模型检验是均否通过？（三个检验）（结果控制在本页）Dependent Variable: GDPBMethod: Least SquaresDate: 04/30/15 Time: 11:25Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StaisticProb.ZC0.3771700.00835545.142650

5、.0000RY0.3536890.0427578.2720280.0000C-800.5997113.7822-7.0362470.0000R-squared0.999152Mean dependent var1754.112Adjusted R-squared0.999085S.D. dependent var1683.912S.E. of regression50.94570Akaike info criterion10.80035Sum squared resid64886.61Schwarz criterion10.94309Log likelihood-148.2050Hannan-

6、Quinn criter.10.84399F-statistic14736.32Durbin-Watson stat0.443892Prob(F-statistic)0.000000所以，由上图可得，多元线性回归方GDPB=0.377170*ZC+0.353689*RY-800.5997故，估计方程的判定系数=0.999085接近1；参数显著性t检验值大于2；方程显著性F检验显著。4将建立的二元回归模型（GDPB同ZC和RY）同一元回归模型（GDPB同ZC、GDPB同RY）相比较，分析优点。（结果控制在本页）（1）GDPB同ZC一元回归模型Dependent Variable: GDPBMet

7、hod: Least SquaresDate: 04/30/15 Time: 11:34Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.ZC0.4428980.00489690.460000.0000C133.972125.570545.2393140.0000R-squared0.996833Mean dependent var1754.112Adjusted R-squared0.996711S.D. dependent var1683.912S.E. of reg

8、ression96.57302Akaike info criterion12.04722Sum squared resid242485.0Schwarz criterion12.14238Log likelihood-166.6611Hannan-Quinn criter.12.07632F-statistic8183.011Durbin-Watson stat0.167556Prob(F-statistic)0.000000（2）GDPB同RY一元回归模型Dependent Variable: GDPBMethod: Least SquaresDate: 05/02/15 Time: 14:

9、03Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.RY2.1893170.11773518.595280.0000C-5519.137400.4253-13.783190.0000R-squared0.930067Mean dependent var1754.112Adjusted R-squared0.927377S.D. dependent var1683.912S.E. of regression453.7907Akaike info criterion15.1

10、4190Sum squared resid5354077.Schwarz criterion15.23706Log likelihood-209.9866Hannan-Quinn criter.15.17099F-statistic345.7844Durbin-Watson stat0.078643Prob(F-statistic)0.000000由于建立的二元回归模型（GDPB同ZC和RY）的调整的判定系数为0.999085,故，建立的二元回归模型（GDPB同ZC和RY）同一元回归模型（GDPB同ZC、GDPB同RY）相比较，其优点是比下面的两个一元回归模型有明显改善。5结合相关的经济理论，

11、分析估计的二元回归模型的经济意义。（结果控制在本页）由于二元回归模型GDPB=0.377170*ZC+0.353689*RY-800.5997所以说，模型估计结果说明，在假定其他变量不变的情况下，当年GDPB每增加1个单位，ZC会增加0.377170个单位，RY会增加0.353689个单位。（二）宏观经济模型：根据广东数据，研究广东省居民消费行为、固定资产投资行为、货物和服务净出口行为和存货行为，分别建立居民消费模型、固定资产投资模型、货物和服务净出口模型和存货增加模型。要求：按照试验指导课本，分别作出以下模型，并对需要改进的模型进行改进。写出最终估计的模型结果，并结合相关的经济理论，分析模型

12、的经济意义。（数据见“表：广东省宏观经济数据-第三章.xls”文件，各变量的表示按照试验指导课本上的来表示。）1居民消费模型（结果控制在本页）Pairwise Granger Causality TestsDate: 05/02/15 Time: 14:22Sample: 1978 2005Lags: 2Null Hypothesis:ObsF-StatisticProb.LB does not Granger Cause XFJ267.190100.0042XFJ does not Granger Cause LB5.455160.0124从散点图看，它们具有线性关系，从因果关系检验看到它们

13、之间似乎具有双向因果关系，宏观经济中确实如此。进行一元线性回归如下：Dependent Variable: XFJMethod: Least SquaresDate: 05/02/15 Time: 14:23Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.LB0.9867020.01691658.330100.0000C-75.9966259.99073-1.2668060.2165R-squared0.992416Mean dependent var2362.277A

14、djusted R-squared0.992125S.D. dependent var2565.722S.E. of regression227.6909Akaike info criterion13.76260Sum squared resid1347921.Schwarz criterion13.85776Log likelihood-190.6765Hannan-Quinn criter.13.79169F-statistic3402.401Durbin-Watson stat0.701578Prob(F-statistic)0.000000所以，由上图可得，一元线性回归方XFJ=0.9

15、86702*LB-75.99662除劳动报酬LB外，企业盈余YY也会影响居民消费XFJ，看散点图和因果关系检验。Pairwise Granger Causality TestsDate: 05/02/15 Time: 14:34Sample: 1978 2005Lags: 1Null Hypothesis:ObsF-StatisticProb.YY does not Granger Cause XFJ274.257200.0501XFJ does not Granger Cause YY0.093580.7623从散点图看和因果关系检验看应该把YY引入方程中，进行一元线性回归如下：Depend

16、ent Variable: XFJMethod: Least SquaresDate: 05/02/15 Time: 14:36Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.LB0.7408080.03289322.521990.0000YY0.3620750.0464527.7946920.0000C46.9151336.602821.2817350.2117R-squared0.997789Mean dependent var2362.277Adjusted R-

17、squared0.997612S.D. dependent var2565.722S.E. of regression125.3710Akaike info criterion12.60139Sum squared resid392946.9Schwarz criterion12.74412Log likelihood-173.4194Hannan-Quinn criter.12.64502F-statistic5641.541Durbin-Watson stat1.122075Prob(F-statistic)0.000000显然回归得到改善，引入YY是正确的，最后得到回归方程XFJ=0.7

18、40808*LB+0.362075*YY+46.915132固定资产投资模型（结果控制在本页）进行三元回归如下：Dependent Variable: TZGMethod: Least SquaresDate: 05/02/15 Time: 14:43Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.ZJ1.1118640.2431524.5727160.0001YY0.4316920.0525668.2123520.0000CZ0.1432100.4053080.353

19、3380.7269C31.2762527.825171.1240270.2721R-squared0.997573Mean dependent var1628.997Adjusted R-squared0.997270S.D. dependent var2003.852S.E. of regression104.7010Akaike info criterion12.27166Sum squared resid263095.1Schwarz criterion12.46197Log likelihood-167.8032Hannan-Quinn criter.12.32984F-statist

20、ic3288.646Durbin-Watson stat1.298515Prob(F-statistic)0.000000现在分别去掉一个解释变量进行三个二元回归如下：Dependent Variable: TZGMethod: Least SquaresDate: 05/02/15 Time: 14:45Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.ZJ1.1918780.08699313.700910.0000YY0.4384220.0481299.1093650

21、.0000C33.6561326.520921.2690410.2161R-squared0.997561Mean dependent var1628.997Adjusted R-squared0.997366S.D. dependent var2003.852S.E. of regression102.8521Akaike info criterion12.20542Sum squared resid264463.7Schwarz criterion12.34815Log likelihood-167.8758Hannan-Quinn criter.12.24905F-statistic51

22、11.852Durbin-Watson stat1.370345Prob(F-statistic)0.000000Dependent Variable: TZGMethod: Least SquaresDate: 05/02/15 Time: 14:46Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.ZJ1.0985780.4650212.3624280.0262CZ1.3493010.7224791.8676010.0736C-45.6139450.11293-0.9

23、102230.3714R-squared0.990754Mean dependent var1628.997Adjusted R-squared0.990014S.D. dependent var2003.852S.E. of regression200.2421Akaike info criterion13.53789Sum squared resid1002422.Schwarz criterion13.68062Log likelihood-186.5304Hannan-Quinn criter.13.58152F-statistic1339.431Durbin-Watson stat0

24、.436795Prob(F-statistic)0.000000Dependent Variable: TZGMethod: Least SquaresDate: 05/02/15 Time: 14:46Sample: 1978 2005Included observations: 28VariableCoefficientStd. Errort-StatisticProb.YY0.4300930.0704536.1047090.0000CZ1.8692780.1978469.4481350.0000C20.9189337.170150.5627880.5786R-squared0.995459Mean dependent var1628.997Adjusted R-squared0.995096S.D. dependent var2003.852S.E. of regression140.3301