中级计量课后习题参考答案(第九章)

第九章参考答案1、表面不相关回归的含义是，所涉及的各个回归似乎不相关，但实际上相关。各个回归方程分别写出，这使得它们似乎不相关，但是它们有共同点。在本章的例子中，四个回归中的每一个关系到一个不同的制造产业，但它们都会受到宏观经济条件变动（如衰退）的影响。一般来说，影响一个回归结果的事件也很可能影响其他回归的结果，这个事实表明，表面不相关回归中的各回归之间存在相关。这种相关在数学上表现为扰动项跨方程相关。表面不相关回归的步骤是：（1）用ols法分别估计每个方程，计算和保存回归中得到的残差；（2）用这些残差来估计扰动项方差和不同回归方程扰动项之间的协方差；（3）上一步估计的扰动项方差和协方差被用于执行广义最小二乘法，得到各方程系数的估计值。2、在不同的横截面种类的截距之间的差异被认为是固定的而不是随机的情况下，应采用固定效应模型。如果横截面个体是随机地被选择出来代表一个较大的总体，则采用随机效应模型比较合适。随机效应模型与固定效应模型一样，允许不同横截面种类的截距不同，但这种不同被认为是随机的，而不是固定的。3、随机影响模型的扰动项不再满足普通最小二乘法各期扰动项相互独立的假设，扰动项的一个分量在各期都相同。4、并不总是。尽管将数据合在一起将增加自由度，但有时采用混合数据也是不合适的。如果不同横截面种类的斜率系数不同的话，则最好是分别回归。如果试图通过使用斜率虚拟变量来解决不同横截面种类不同斜率系数的问题，需要假定扰动项方差为常数。而采用分别回归，每个回归的扰动项方差可以不同，也就是每个产业或每个横截面种类的扰动项方差不同。5、随机系数模型即每个横截面个体的解释变量对被解释变量的影响在横截面个体之间的差异的变动时随机的。有滞后因变量做自变量的动态模型就是动态面板数据模型。6、（1）对钢铁产业用OLS法估计的结果如下：DependentVariable:Y1Method:LeastSquaresDate:12/02/10Time:10:39Sample:19802000Includedobservations:21VariableCoefficientStd.Errort-StatisticProb.C3919.1801702.6912.3017560.0335EMP131.999985.3057566.0311810.0000OTM1722.7758348.28732.0752290.0526R-squared0.674135Meandependentvar10339.75AdjustedR-squared0.637928S.D.dependentvar1653.825S.E.ofregression995.1473Akaikeinfocriterion16.77522Sumsquaredresid17825726Schwarzcriterion16.92444Loglikelihood-173.1398Hannan-Quinncriter.16.80761F-statistic18.61879Durbin-Watsonstat0.436339Prob(F-statistic)0.000041橡胶和塑料产业：DependentVariable:Y2Method:LeastSquaresDate:12/02/10Time:10:40Sample:19802000Includedobservations:21VariableCoefficientStd.Errort-StatisticProb.C-49122.543331.606-14.744400.0000EMP2135.49486.70325520.213280.0000OTM22646.5571087.2842.4340990.0256R-squared0.989264Meandependentvar80662.43AdjustedR-squared0.988071S.D.dependentvar13744.48S.E.ofregression1501.188Akaikeinfocriterion17.59746Sumsquaredresid40564183Schwarzcriterion17.74668Loglikelihood-181.7734Hannan-Quinncriter.17.62985F-statistic829.2748Durbin-Watsonstat1.590448Prob(F-statistic)0.000000SUR的估计：在主菜单选择Object-NewObject，在弹出的对话框中选择System，点击OK。在编辑框中输入：y_gt=c(1)+c(2)*emp_gt+c(3)*otm_gty_xj=c(4)+c(5)*emp_xj+c(6)*otm_xj估计结果如下：System:SUR_YEstimationMethod:SeeminglyUnrelatedRegressionDate:12/04/10Time:10:56Sample:19802000Includedobservations:21Totalsystem(balanced)observations42Linearestimationafterone-stepweightingmatrixCoefficientStd.Errort-StatisticProb.C(1)4967.3361497.3993.3173090.0021C(2)28.882224.7322826.1032320.0000C(3)539.5766308.85691.7470120.0892C(4)-51805.333007.227-17.226950.0000C(5)142.16995.88536024.156530.0000C(6)1822.018965.58521.8869580.0673Determinantresidualcovariance1.22E+12Equation:Y_GT=C(1)+C(2)*EMP_GT+C(3)*OTM_GTObservations:21R-squared0.666480Meandependentvar10339.75AdjustedR-squared0.629422S.D.dependentvar1653.825S.E.ofregression1006.768Sumsquaredresid18244464Durbin-Watsonstat0.486554Equation:Y_XJ=C(4)+C(5)*EMP_XJ+C(6)*OTM_XJObservations:21R-squared0.988661Meandependentvar80662.43AdjustedR-squared0.987401S.D.dependentvar13744.48S.E.ofregression1542.747Sumsquaredresid42841220Durbin-Watsonstat1.484711SUR估计的模型如下：SubstitutedCoefficients:=====================Y_GT=4967.33600398+28.8822169308*EMP_GT+539.576580766*OTM_GTY_XJ=-51805.3302975+142.169865054*EMP_XJ+1822.0182034*OTM_XJ估计结果说明采用SUR估计得到的斜率和用OLS法估计得到的斜率相同。（2）1.混合回归模型：SubstitutedCoefficients:=====================Y_GT=-14046.4794265+86.7390745907*EMP_GT+3170.25123496*OTM_GTY_XJ=-14046.4794265+86.7390745907*EMP_XJ+3170.25123496*OTM_XJY_SZ=-14046.4794265+86.7390745907*EMP_SZ+3170.25123496*OTM_SZY_FZ=-14046.4794265+86.7390745907*EMP_FZ+3170.25123496*OTM_FZDependentVariable:Y?Method:PooledLeastSquaresDate:12/04/10Time:11:20Sample:19802000Includedobservations:21Cross-sectionsincluded:4Totalpool(balanced)observations:84VariableCoefficientStd.Errort-StatisticProb.C-14046.483233.619-4.3438880.0000EMP?86.739072.17543439.872080.0000OTM?3170.251731.41024.3344370.0000R-squared0.953292Meandependentvar48601.24AdjustedR-squared0.952139S.D.dependentvar26268.36S.E.ofregression5746.759Akaikeinfocriterion20.18572Sumsquaredresid2.68E+09Schwarzcriterion20.27254Loglikelihood-844.8003Hannan-Quinncriter.20.22062F-statistic826.5976Durbin-Watsonstat0.097683Prob(F-statistic)0.0000002.变截距模型：SubstitutedCoefficients:=====================Y_GT=5513.35297339-23271.5488569+92.1914938172*EMP_GT+4644.35831606*OTM_GTY_XJ=4576.61080153-23271.5488569+92.1914938172*EMP_XJ+4644.35831606*OTM_XJY_SZ=-3412.39724347-23271.5488569+92.1914938172*EMP_SZ+4644.35831606*OTM_SZY_FZ=-6677.56653145-23271.5488569+92.1914938172*EMP_FZ+4644.35831606*OTM_FZDependentVariable:Y?Method:PooledLeastSquaresDate:12/04/10Time:12:03Sample:19802000Includedobservations:21Cross-sectionsincluded:4Totalpool(balanced)observations:84VariableCoefficientStd.Errort-StatisticProb.C-23271.553783.341-6.1510580.0000EMP?92.191495.37723717.144770.0000OTM?4644.358510.15469.1038250.0000FixedEffects(Cross)_GT--C5513.353_XJ--C4576.611_SZ--C-3412.397_FZ--C-6677.567EffectsSpecificationCross-sectionfixed(dummyvariables)R-squared0.986194Meandependentvar48601.24AdjustedR-squared0.985309S.D.dependentvar26268.36S.E.ofregression3183.855Akaikeinfocriterion19.03832Sumsquaredresid7.