《计量经济学-eviews实验报告(共12页).doc》由会员分享,可在线阅读,更多相关《计量经济学-eviews实验报告(共12页).doc(12页珍藏版)》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、精选优质文档-倾情为你奉上计量经济学作业院系:商学院国贸三班教室:高辉姓名:吴谧学号:2INDEX问题2模型设定3检验异方差4图形检验4Glejser检验5White检验6调整异方差6习题5.8表5.13给出的是1998年我国重要制造业销售收入与销售利润的数据表5.13 我国重要制造工业1998年销售利润与销售收入情况行业名称销售利润销售收汝行业名称销售利润销售收入食品加工业187.253180.44医药制造业238.711264.1食品制造业111.421119.88化学纤维制品81.57779.46饮料制造业205.421489.89橡胶制品业77.84692.08烟草加工业183.871
2、328.59塑料制品业144.341345纺织业316.793862.9非金属矿制品339.262866.14服装制品业157.71779.1黑色金属冶炼367.473868.28皮革羽绒制品81.731081.77有色金属冶炼144.291535.16木材加工业35.67443.74金属制品业201.421948.12家具制造业31.06226.78普通机械制造354.692351.68造纸及纸品业134.41124.94专用设备制造238.161714.73印刷业90.12499.83交通运输设备511.944011.53文教体育用品54.4504.44电子机械制造409.833286.1
3、5石油加工业194.452363.8电子通讯设备508.154499.19化学原料纸品502.614195.22仪器仪表设备72.46663.68试完成以下问题:1) 求销售利润与销售收入的样本回归函数,并对模型进行经济意义检验和统计检验;2) 分别用图形法、Glejser方法、White方法检验模型是否存在异方差;3) 如果模型存在异方差,选用适当的方法对异方差性进行修正。1)假定销售利润与销售收入之间满足线性约束,则理论模型设定为Yi = 1 + 2X I + ui其中,Yi表示销售利润,表示销售收入。Dependent Variable: YMethod: Least SquaresDa
4、te: 12/26/09 Time: 14:45Sample: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C12.0356419.517790.0.5428X0.0.12.366700.0000R-squared0.Mean dependent var213.4650Adjusted R-squared0.S.D. dependent var146.4895S.E. of regression56.90368Akaike info criterion10.98935Sum squared
5、resid84188.74Schwarz criterion11.08450Log likelihood-151.8508Hannan-Quinn criter.11.01844F-statistic152.9353Durbin-Watson stat1.Prob(F-statistic)0.图1估计结果为i = 12.03564 + 0.X i(0.61665)(12.3667)R2 = 0.8547,F = 152.94括号内为t统计量值。a) 经济意义检验:所估计得参数1=0.,说明销售收入(X)没相差1亿元,可导致销售利润(Y)相差0.亿元。随着销售收入的增加,销售利润的平均水平是不断
6、提高的,符合经济意义。b) 拟合优度的度量:由此估计参数课件,该模型R2=0.8547可绝系数很高,说明所建模型整体上对样本数据拟合较好,即解释变量“销售收入(X)”对被解释变量“销售利润(Y)”的绝大部分作出解释。c) 对于回归系数的t检验:在给定显著性水平=0.05,在t分布表中查处自由度为n-2=26的临界值t0.025(26)=2.056,由于t(0)=0. t0.025(26)=2.056,对X的系数t检验显著影响。这表明,销售收入(X)对销售利润(Y)有显著影响。2)检验异方差l 图形分析检验观察销售利润(Y)与销售收入(X)的相关图(图1):SCAT X Y图2从图中可以看出,随
7、着销售收入的增加,销售利润的平均水平不断提高,但离散程度也逐步扩大。这说明变量之间可能存在递增的异方差性。残差分析图3图3显示回归方程的残差分布有明显的扩大趋势,即表明存在异方差性。l Glejser检验建立回归模型(结果同图1所示)。生成新变量序列GENR E=ABS(RESID)分别建立新残差序列(E)对各解释变量(X/X2/X(1/2)/X(1)/ X(2)/ X(1/2))的回归模型:LS E C X,回归结果如图4、5、6、7、8、9所示。Dependent Variable: EMethod: Least SquaresDate: 12/26/09 Time: 17:29Sampl
8、e: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C12.2380810.618521.0.2596X0.0.3.0.0026R-squared0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regression30.95804Akaike info criterion9.Sum squared resid24918.40Schwarz criterion9.Log likel
9、ihood-134.8065Hannan-Quinn criter.9.F-statistic11.05084Durbin-Watson stat1.Prob(F-statistic)0.图4Dependent Variable: EMethod: Least SquaresDate: 12/26/09 Time: 17:32Sample: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C27.057248.3.0.0029X22.74E-061.02E-062.0.0123R-squared
10、0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regression32.68642Akaike info criterion9.Sum squared resid27778.46Schwarz criterion9.Log likelihood-136.3276Hannan-Quinn criter.9.F-statistic7.Durbin-Watson stat1.Prob(F-statistic)0.图5Dependent Variable: EMethod: Least
11、 SquaresDate: 12/26/09 Time: 17:35Sample: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C-15.6781817.09598-0.0.3675X(1/2)1.0.3.0.0014R-squared0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regression30.29690Akaike info criterion9.Sum squ
12、ared resid23865.45Schwarz criterion9.Log likelihood-134.2020Hannan-Quinn criter.9.F-statistic12.68554Durbin-Watson stat1.Prob(F-statistic)0.图6Dependent Variable: EMethod: Least SquaresDate: 12/26/09 Time: 17:36Sample: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C59.3887
13、68.6.0.0000X(-1)-19128.427040.292-2.0.0116R-squared0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regression32.61490Akaike info criterion9.Sum squared resid27657.02Schwarz criterion9.Log likelihood-136.2663Hannan-Quinn criter.9.F-statistic7.Durbin-Watson stat1.Prob
14、(F-statistic)0.图7Dependent Variable: EMethod: Least SquaresDate: 12/26/09 Time: 17:37Sample: 1 28Included observations: 28VariableCoefficientStd. Errort-StatisticProb.C46.935017.6.0.0000X(-2)-.-1.0.0832R-squared0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regress
15、ion34.84529Akaike info criterion10.00846Sum squared resid31569.05Schwarz criterion10.10362Log likelihood-138.1185Hannan-Quinn criter.10.03755F-statistic3.Durbin-Watson stat1.Prob(F-statistic)0.图8Dependent Variable: EMethod: Least SquaresDate: 12/26/09 Time: 17:38Sample: 1 28Included observations: 28
16、VariableCoefficientStd. Errort-StatisticProb.C86.8062415.090735.0.0000X(-1/2)-1611.089496.1924-3.0.0032R-squared0.Mean dependent var41.69584Adjusted R-squared0.S.D. dependent var36.26524S.E. of regression31.17267Akaike info criterion9.Sum squared resid25265.12Schwarz criterion9.Log likelihood-134.99
17、99Hannan-Quinn criter.9.F-statistic10.54239Durbin-Watson stat1.Prob(F-statistic)0.图9由上述各回归结果可知,各回归模型中解释变量的系数估计值显著不为0且均能通过显著性检验。所以认为存在异方差性。由F值或确定异方差类型。Gleiser检验中可以通过F值或值确定异方差的具体形式。图6所示的回归方程F值()最大,可以据次来确定异方差的形式。l White检验建立回归模型,回归结果如图1。在方程窗口上点击ViewResidual TestHeteroskedastcity TestsWhite,检验结果如图10Heter
18、oskedasticity Test: WhiteF-statistic3.Prob. F(2,25)0.0420Obs*R-squared6.Prob. Chi-Square(2)0.0435Scaled explained SS7.Prob. Chi-Square(2)0.0220图10其中F值为辅助回归模型的F统计量值。取显著水平,由于20.05(2) = 5.99nR2 = 6.2706,所以存在异方差性。3) 调整异方差性1.确定权数变量根据Gleiser检验生成权数变量:GENR W1=1/X0.5另外生成:GENR W2=1/ABS(RESID)GENR W3=1/ RESID
19、22.利用加权最小二乘法估计模型在Eviews命令窗口中依次键入命令:LS(W=) Y C X或在方程窗口中点击EstimateOption按钮,并在权数变量栏里依次输入W1、W2、W3、W4,回归结果图11、12、13所示。Dependent Variable: YMethod: Least SquaresDate: 12/26/09 Time: 18:13Sample: 1 28Included observations: 28Weighting series: W1VariableCoefficientStd. Errort-StatisticProb.C8.11.187330.0.44
20、69X0.0.13.704730.0000图11Dependent Variable: YMethod: Least SquaresDate: 12/27/09 Time: 17:44Sample: 1 28Included observations: 28Weighting series: W2VariableCoefficientStd. Errort-StatisticProb.C4.3.1.0.2803X0.0.30.963150.0000图12Dependent Variable: YMethod: Least SquaresDate: 12/27/09 Time: 17:56Sam
21、ple: 1 28Included observations: 28Weighting series: W3VariableCoefficientStd. Errort-StatisticProb.C5.1.3.0.0047X0.0.54.162070.0000图133.对所估计的模型再进行White检验,观察异方差的调整情况对所估计的模型再进行White检验,其结果分别对应图11、12、13的回归模型(如图14、15、16所示)。图14、16所对应的White检验显示,P值较大,所以接收不存在异方差的原假设,即认为已经消除了回归模型的异方差性。图15对应的White检验没有显示F值和的值,这
22、表示异方差性已经得到很好的解决。Heteroskedasticity Test: WhiteF-statistic0.Prob. F(2,25)0.5038Obs*R-squared1.Prob. Chi-Square(2)0.4600Scaled explained SS1.Prob. Chi-Square(2)0.4798图14Heteroskedasticity Test: White图15Heteroskedasticity Test: WhiteF-statistic3.Prob. F(3,24)0.0437Obs*R-squared7.Prob. Chi-Square(3)0.04
23、81Scaled explained SS0.Prob. Chi-Square(3)0.8116图164.异方差修正结果估计用w2为权数,即以图12估计:Dependent Variable: YMethod: Least SquaresDate: 12/27/09 Time: 17:44Sample: 1 28Included observations: 28Weighting series: W2VariableCoefficientStd. Errort-StatisticProb.C4.3.1.0.2803X0.0.30.963150.0000Weighted StatisticsR-
24、squared0.Mean dependent var136.4194Adjusted R-squared0.S.D. dependent var107.5683S.E. of regression17.19393Akaike info criterion8.Sum squared resid7686.414Schwarz criterion8.Log likelihood-118.3404Hannan-Quinn criter.8.F-statistic958.7164Durbin-Watson stat2.i = 4. + 0.X i(1.)(30.96315)R2 = 0.9736,DW = 2.,F = 958.7164括号中为t统计量值。 可以看出消除异方差后,X系数的t检验显著,R2=0.9736大幅度提高,F检验也显著,并说明销售收入每增加一亿元,销售利润将增加0.10941亿元。专心-专注-专业