自相关的检验与修正

附件二：实验报告格式（首页）山东轻工业学院实验报告成绩课程名称计量经济学指导教师实验日期2013-5-25院（系）商学院专业班级实验地点二机房学生姓名学号同组人无实验项目名称自相关的检验与修正一、实验目的和要求掌握Eviews软件的操作和自相关的检验与修正二、实验原理Eviews软件的操作和自相关的检验与修正，图表法，DW检验，运用迭代法三、主要仪器设备、试剂或材料Eviews软件，计算机四、实验方法与步骤（1）准备工作：建立工作文件，并输入数据：CREATEEX-6-1A19782000;TATACINSUMINCOMEPRICE;(2)相关图分析：GENRY=CONSUM/PRICE;GENRX=INCOME/PRICE;SCATXY;LSYCX;(3)自相关检验1）图示法LINRRESID;SCATRESID(-1)RESID;2)观察结果窗口，由DW统计量，查表，与DL,DU比较得出结论；3）LM检验在方程窗口中点击View—residualtest–seriescorrelationLMtest;(4)自相关的修正GENRGDY=Y-0.7*Y(-1);GENRGDX=X-0.7*X(-1);LSGDYCGDX;(5)再次检验自相关是否存在，用1),2),3）之一检验；五、实验数据记录、处理及结果分析（1）建立工作组，输入数据如下：1978344.88388.3211979385.2425.41.011980474.72526.921.0621981485.88539.521.0751982496.56576.721.0811983520.84604.311.0861984599.64728.171.0161985770.64875.521.251986949.081069.611.33619871071.041187.491.42619881278.871329.71.66719891291.091477.771.91219901440.471638.921.9719911585.711844.982.17119921907.172238.382.41819932322.192769.262.84419943301.373982.133.52619954064.14929.534.06619964679.615967.714.43219975204.296608.564.56919985471.017110.544.54619995851.537649.834.49620006121.078140.554.478（2）相关图分析Scatxy,得到关于X和Y的散点图如下：2004006008001000120014000400800120016002000XY从上图可知，X和Y存在线性关系。再者，Eviews估计结果如下图：DependentVariable:YMethod:LeastSquaresDate:05/25/13Time:10:47Sample:19782000Includedobservations:23VariableCoefficientStd.Errort-StatisticProb.C112.608116.974476.6339660.0000X0.7108740.01679042.339840.0000R-squared0.988421Meandependentvar771.4916AdjustedR-squared0.987870S.D.dependentvar295.2138S.E.ofregression32.51395Akaikeinfocriterion9.884157Sumsquaredresid22200.30Schwarzcriterion9.982896Loglikelihood-111.6678F-statistic1792.662Durbin-Watsonstat0.582546Prob(F-statistic)0.000000得到估计的回归方程：ˆ111.440.7118ttYX(6.5)(42.1)R2=0.9883s.e=32.8DW=0.60T=23回归方程拟合得效果比较好，但是DW值比较低。（3）自相关检验1）图示法：由scatresid(-1)resid得到下图：-60-40-20020406080100-60-40-20020406080100RESID(-1)RESID-60-40-20020406080100788082848688909294969800RESID由上图可知，方程存在序列相关性。2）已知DW=0.60，若给定a=0.05,查表得DW临界值dL=1.26，dU=1.44。因为DW=0.06＜1.26，认为误差项存在严重自相关。3）LM检验Breusch-GodfreySerialCorrelationLMTest:F-statistic7.584071Probability0.003791Obs*R-squared10.21031Probability0.006065TestEquation:DependentVariable:RESIDMethod:LeastSquaresDate:05/25/13Time:10:54Presamplemissingvaluelaggedresidualssettozero.VariableCoefficientStd.Errort-StatisticProb.C3.90350913.386240.2916060.7737X-0.0056610.013323-0.4249120.6757RESID(-1)0.5966480.2317782.5742270.0186RESID(-2)0.1405640.2365540.5942150.5594R-squared0.443926Meandependentvar-2.32E-14AdjustedR-squared0.356125S.D.dependentvar31.76641S.E.ofregression25.48994Akaikeinfocriterion9.471215Sumsquaredresid12345.00Schwarzcriterion9.668693Loglikelihood-104.9190F-statistic5.056047Durbin-Watsonstat1.765655Prob(F-statistic)0.009645LM﹙BG﹚自相关检验辅助回归方程式估计结果是：10.67903.17100.0047tttteeXv﹙3.9）﹙0.2﹚﹙-0.4﹚R2=0.43DW=2.00LM==23×0.43=9.89因为，LM=9.89﹥3.84，所以LM检验结果也说明误差项存在一阶正相关。（4）自相关的修正DependentVariable:GDYMethod:LeastSquaresDate:05/25/13Time:11:00Sample(adjusted):19792000Includedobservations:22afteradjustmentsVariableCoefficientStd.Errort-StatisticProb.C46.5422111.876453.9188640.0009GDX0.6746730.03282720.552380.0000R-squared0.954792Meandependentvar269.7844AdjustedR-squared0.952532S.D.dependentvar103.3914S.E.ofregression22.52611Akaikeinfocriterion9.153735Sumsquaredresid10148.51Schwarzcriterion9.252921Loglikelihood-98.69108F-statistic422.4005Durbin-Watsonstat2.310424Prob(F-statistic)0.000000令GDY=Y-0.70*Y(-1)GDX=X-0.70X(-1)以GDY、GDX、（1979-2000）为样本再次回归，得GDY=45.2489+0.6782GDX(3.7)(20.0)R2=0.95s.e=23.2DW=2.31,T=22(1979~2000)回归方程拟合优度依然较好，且DW=2.13.查表得1.26Ld，1.43Ud。因为DW=2,13﹤2.57，依据判断规则，误差项已经消除自相关。（5）再次检验自相关是否存在Breusch-GodfreySerialCorrelationLMTest:F-statistic0.660806Probability0.528521Obs*R-squared1.504816Probability0.471230TestEquation:DependentVariable:RESIDMethod:LeastSquaresDate:05/25/13Time:11:02Presamplemissingvaluelaggedresidualssettozero.VariableCoefficientStd.Errort-StatisticProb.C0.07870712.089110.0065110.9949GDX-0.0002740.033404-0.0081990.9935RESID(-1)-0.1361810.232983-0.5845130.5661RESID(-2)0.2047850.2330330.8787800.3911R-squared0.068401Meandependentvar3.42E-14AdjustedR-squared-0.086866S.D.dependentvar21.98323S.E.ofregression22.91814Akaikeinfocriterion9.264701Sumsquaredresid9454.345Schwarzcriterion9.463072Loglikelihood-97.91171F-statistic0.440537Durbin-Watsonstat1.940615Prob(F-statistic)0.726816由上图知D.W约为2，误差项已消除自相关。六、讨论、心得通过本次试验，我进一步熟悉了掌握了Eviews软件的操作，并且掌握了自相关的检验以及修正方法。我认为在输出图像，图表的同时，分析解释非常重要。保持一颗细心，耐心的态度，是做好事情的基础。也锻炼了我的严谨的能力，我相信我一定会把计量经济学学好，相信自己行，一定我能行！附件三：实验报告附页山东轻工业学院实验报告（附页）