计量经济学用eviews分析数据

中国储蓄存款总额（Y，亿元）与GDP（元）数据如下表。表1-1年份GDP储蓄Y19783645.2210.619794062.6281.019804545.6399.519814889.5523.719825330.5675.419835985.6892.519847243.81214.719859040.71622.6198610274.42238.5198712050.63081.4198815036.83822.2198917000.95196.4199018718.37119.8199121826.29141.6199226937.311758.0199335260.015203.5199448108.521518.8199559810.529662.3199670142.538520.8199778060.946279.8199883024.353407.5199988479.259621.8200098000.564332.42001108068.273762.42002119095.786910.62003134977.0103617.32004159453.6119555.42005183617.4141051.02006215904.4161587.32007266422.0172534.22008316030.3217885.42009340320.0260771.72010399759.5303302.52011468562.4343635.92012516282.1399551.0数据来源：《中国统计年鉴》，2013年图1-1解：一、估计一元线性回归模型由经济理论知，储蓄存款总额受GDP影响，当GDP增加时，储蓄存款总额也随着增加，他们之间具有正向的同步变动趋势。储蓄存款总额除受GDP影响之外，还受到其他一些变量的影响及随机因素的影响，将其他变量及随机因素的影响均并到随机变量U中，根据X与Y的样本数据，作X与Y之间的散点图可以看出，他们的变化趋势是线性的，由此建立中国储蓄存款总额Y与GDP之间的一员线性回归模型。由表1-1中样本观测数据，样本回归模型为用Eviews软件估计结果：DependentVariable:YMethod:LeastSquaresDate:12/14/14Time:10:41Sample:19782012Includedobservations:35CoefficientStd.Errort-StatisticProb.C-7304.2941561.216-4.6785920.0000GDP0.7625290.00869887.662520.0000R-squared0.995724Meandependentvar78882.560100,000200,000300,000400,000500,0000200,000400,000600,000GDPYAdjustedR-squared0.995595S.D.dependentvar108096.8S.E.ofregression7174.769Akaikeinfocriterion20.64997Sumsquaredresid1.70E+09Schwarzcriterion20.73885Loglikelihood-359.3745Hannan-Quinncriter.20.68065F-statistic7684.717Durbin-Watsonstat1.224720Prob(F-statistic)0.000000即样本回归方程为：-4.67859287.66252二、对估计结果做结构分析（1）对回归方程的结构分析0.762529是样本回归方程的斜率，他表示GDP的边际增长率，说明GDP每增加1元，将有0.762529用于储蓄；-7304.294是样本回归方程的截距，他表示不受GDP影响的自发性储蓄增长。和的符号和大小，均符合经济理论及目前国家的实际情况。（2）统计检验=0.995724，说明总离差平方和的99.6%被样本回归直线解释，仅有0.4%未被解释，因此，样本回归直线对样本点的拟合优度是很高的。给出显著性水平α=0.05，查自由度v=35-2=33的t分布表，得临界值()，||＞2.03，故回归系数均显著不为零，回归模型中应摆放常数项，GDP对Y有显著影响。从以上的评价可以看出，此模型是比较好的。三、假设𝒈𝒅𝒑𝟐𝟑=568845.0，对2013年国民储蓄总额进行预测给出𝑔𝑑𝑝3=568845.0，可以得到Y𝑓3=568845-100,0000100,000200,000300,000400,000500,0001980198519901995200020052010YF±2S.E.Forecast:YFActual:YForecastsample:19782013Includedobservations:35RootMeanSquaredError6966.760MeanAbsoluteError5142.179MeanAbs.PercentError211.2178TheilInequalityCoefficient0.026295BiasProportion0.000000VarianceProportion0.001071CovarianceProportion0.998929图1-2四、对异方差进行检验用怀特检验法进行异方差检验：HeteroskedasticityTest:WhiteF-statistic0.081308Prob.F(3,28)0.9696Obs*R-squared0.276363Prob.Chi-Square(3)0.9644ScaledexplainedSS0.882741Prob.Chi-Square(3)0.8296TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/14/14Time:14:54Sample:19812012Includedobservations:32CoefficientStd.Errort-StatisticProb.C38523385206556381.8650300.0727E(-1)^2-0.0420010.172609-0.2433330.8095E(-1)*E(-2)-0.4096310.859096-0.4768160.6372E(-2)^20.0940820.2820030.3336200.7412R-squared0.008636Meandependentvar36993700AdjustedR-squared-0.097581S.D.dependentvar1.01E+08S.E.ofregression1.06E+08Akaikeinfocriterion39.91526Sumsquaredresid3.16E+17Schwarzcriterion40.09848Loglikelihood-634.6442Hannan-Quinncriter.39.97599F-statistic0.081308Durbin-Watsonstat1.876393Prob(F-statistic)0.969643提出假设：𝐻：𝛼=0，i=1,2𝐻：𝛼，𝛼中至少有一个不等于零HeteroskedasticityTest:WhiteF-statistic0.081308Prob.F(3,28)0.9696Obs*R-squared0.276363Prob.Chi-Square(3)0.9644ScaledexplainedSS0.882741Prob.Chi-Square(3)0.8296TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/14/14Time:14:54Sample:19812012Includedobservations:32CoefficientStd.Errort-StatisticProb.C38523385206556381.8650300.0727E(-1)^2-0.0420010.172609-0.2433330.8095E(-1)*E(-2)-0.4096310.859096-0.4768160.6372E(-2)^20.0940820.2820030.3336200.7412R-squared0.008636Meandependentvar36993700AdjustedR-squared-0.097581S.D.dependentvar1.01E+08S.E.ofregression1.06E+08Akaikeinfocriterion39.91526Sumsquaredresid3.16E+17Schwarzcriterion40.09848Loglikelihood-634.6442Hannan-Quinncriter.39.97599F-statistic0.081308Durbin-Watsonstat1.876393Prob(F-statistic)0.969643WT（g）=TR=0.276363当H成立时服从自由度为g的χ分布，其中g=（k+）（k+）-1=2给定显著性水平α，查临界值χ𝛼()=5.991WT（g）＜χ𝛼()则H成立，那么原模型不存在不存在异方差。五、自相关检验检验误差项是否存在自相关：已知DW=1.224720，若给定α=0.05，查表，𝐿=1.40，𝑢。因为DW=1.22＜1.40，依据判别规则，认为误差项存在严重的正自相关。残差序列见图1-3。图1-3-25,000-20,000-15,000-10,000-5,00005,00010,00015,0001980198519901995200020052010RESID𝑒=1.187𝑒−-0.479𝑒−+𝑣DependentVariable:EMethod:LeastSquaresDate:12/14/14Time:14:46Sample(adjusted):19812012Includedobservations:32afteradjustmentsCoefficientStd.Errort-StatisticProb.E(-1)0.1871180.1820951.0275810.3124E(-2)-0.4791470.188696-2.5392570.0165R-squared0.181787Meandependentvar-181.2631AdjustedR-squared0.154513S.D.dependentvar6831.637S.E.ofregression6281.715Akaikeinfocriterion20.38914Sumsquaredresid1.18E+09Schwarzcriterion20.48074Loglikelihood-324.2262Hannan-Quinncriter.20.41950Durbin-Watsonstat1.385678一阶广义差分：DependentVariable:GDYMethod:LeastSquaresDate:12/15/14Time:22:03Sample(adjusted):19792012Includedobservations:34afteradjustmentsCoefficientStd.Errort-StatisticProb.C-4804.8991508.843-3.1844930.0032GDGDP0.7672320.01256161.079980.0000R-squared0.991496Meandependentvar54110.50AdjustedR-squared0.991230S.D.dependentvar72242.89S.