实验三多元时间序列分析方法1.实验目的了解协整理论及协整检验方法;掌握协整的两种检验方法:E-G两步法与Johansen方法;熟悉向量自回归模型VAR的应用;掌握误差修正模型ECM的含义及检验方法;掌握Granger因果关系检验方法。2.实验仪器装有EViews7.0软件的微机一台。3.实验内容【例6-2】时间与M2之间的关系首先用单位根检验是否为平稳序列。原假设为H0:非平稳序列H1:平稳序列。用Eviews软件解决该问题,得到如下结果:NullHypothesis:M2hasaunitrootExogenous:NoneLagLength:3(Automatic-basedonSIC,maxlag=13)t-StatisticProb.*AugmentedDickey-Fullerteststatistic5.6811691.0000Testcriticalvalues:1%level-2.5790525%level-1.94276810%level-1.615423*MacKinnon(1996)one-sidedp-values.AugmentedDickey-FullerTestEquationDependentVariable:D(M2)Method:LeastSquaresDate:04/16/13Time:10:36Sample(adjusted):1991M052005M01Includedobservations:165afteradjustmentsVariableCoefficientStd.Errort-StatisticProb.M2(-1)0.0135140.0023795.6811690.0000D(M2(-1))-0.4902800.074458-6.5846110.0000D(M2(-2))0.0706180.0837900.8427970.4006D(M2(-3))0.3870860.0737885.2459350.0000R-squared0.480147Meandependentvar1440.037AdjustedR-squared0.470461S.D.dependentvar1509.489S.E.ofregression1098.447Akaikeinfocriterion16.86513Sumsquaredresid1.94E+08Schwarzcriterion16.94042Loglikelihood-1387.373Hannan-Quinncriter.16.89569Durbin-Watsonstat1.965242从上图我们可以看出t-statistic的值是5.681169,大于临界值,p>a,故不能拒绝被检验的指数序列是非平稳的原假设。因此一阶差分序列进行ADF检验,结果如下图显示。NullHypothesis:D(M2)hasaunitrootExogenous:NoneLagLength:8(Automatic-basedonSIC,maxlag=13)t-StatistiProb.*cAugmentedDickey-Fullerteststatistic0.9881830.9143Testcriticalvalues:1%level-2.5795875%level-1.94284310%level-1.615376*MacKinnon(1996)one-sidedp-values.AugmentedDickey-FullerTestEquationDependentVariable:D(M2,2)Method:LeastSquaresDate:04/16/13Time:10:37Sample(adjusted):1991M112005M01Includedobservations:159afteradjustmentsVariableCoefficientStd.Errort-StatisticProb.D(M2(-1))0.0536160.0542570.9881830.3247D(M2(-1),2)-1.5260690.096352-15.838520.0000D(M2(-2),2)-1.5196490.149134-10.189810.0000D(M2(-3),2)-1.2256230.184003-6.6608690.0000D(M2(-4),2)-1.2374450.196285-6.3043190.0000D(M2(-5),2)-0.9720240.197161-4.9300930.0000D(M2(-6),2)-0.8100980.185290-4.3720600.0000D(M2(-7),2)-0.6050690.144997-4.1729830.0001D(M2(-8),2)-0.3337810.080550-4.1437810.0001R-squared0.801713Meandependentvar16.07001AdjustedR-squared0.791137S.D.dependentvar2352.919S.E.ofregression1075.320Akaikeinfocriterion16.85356Sumsquaredresid1.73E+08Schwarzcriterion17.02727Loglikelihood-1330.858Hannan-Quinncriter.16.92410Durbin-Watsonstat1.970407从上图我们可以看出t-statistic的值是0.988183,大于临界值,p>a,故不能拒绝被检验的指数序列是非平稳的原假设。因此二阶差分序列进行ADF检验,结果如下图显示NullHypothesis:D(M2,2)hasaunitrootExogenous:NoneLagLength:7(Automatic-basedonSIC,maxlag=13)t-StatisticProb.*AugmentedDickey-Fullerteststatistic-9.2231320.0000Testcriticalvalues:1%level-2.5795875%level-1.94284310%level-1.615376*MacKinnon(1996)one-sidedp-values.AugmentedDickey-FullerTestEquationDependentVariable:D(M2,3)Method:LeastSquaresDate:04/16/13Time:10:38Sample(adjusted):1991M112005M01Includedobservations:159afteradjustmentsVariableCoefficientStd.Errort-StatisticProb.D(M2(-1),2)-8.9007550.965047-9.2231320.0000D(M2(-1),3)6.4311290.9246726.9550380.0000D(M2(-2),3)4.9702860.8335415.9628610.0000D(M2(-3),3)3.8024320.7007735.4260550...