《序列相关性案例.pptx》由会员分享,可在线阅读,更多相关《序列相关性案例.pptx(13页珍藏版)》请在淘文阁 - 分享文档赚钱的网站上搜索。
1、1被解释变量被解释变量人均消费支出人均消费支出Y Y解释变量解释变量居民纯收入居民纯收入X X二、模型设定二、模型设定二者之间满足线性约束,理论模型设定为:二者之间满足线性约束,理论模型设定为:第1页/共13页21985-20031985-2003年农村居民人均收入和消费年农村居民人均收入和消费 :(n=19):(n=19)年份年份全年人全年人均纯入均纯入(现价)(现价)全年人均全年人均消费性支消费性支出(现价)出(现价)消费价格指消费价格指数数(1985=1001985=100)人均实际纯收人均实际纯收入入(19851985可比价)可比价)人均实际消费人均实际消费性支出性支出(1985198
2、5可比价)可比价)19851985397.6397.6317.42317.42100100397.6397.6317.4317.419861986423.8423.8357357106.1106.1399.43399.43336.48336.4819871987462.6462.6398.3398.3112.7112.7410.47410.47353.42353.4219881988544.9544.9476.7476.7132.4132.4411.56411.56360.05360.0519891989601.5601.5535.4535.4157.9157.9380.94380.94339
3、.08339.0819911991708.6708.6619.8619.8168.9168.9419.54419.54366.96366.96数据来自统计年鉴数据来自统计年鉴397.6/100*100=397.6317.42/100*100=317.42423.8/106.1*100=399.43357/106.1*100=336.48第2页/共13页3Dependent Variable:YMethod:Least SquaresDate:06/05/11 Time:15:51Sample:1985 2003Included observations:19VariableCoefficien
4、tStd.Errort-StatisticProb.C106.757412.223678.7336610.0000X0.5997810.02139328.036710.0000R-squared0.978831 Mean dependent var437.6705Adjusted R-squared0.977586 S.D.dependent var92.57790S.E.of regression13.86023 Akaike info criterion8.195225Sum squared resid3265.801 Schwarz criterion8.294640Log likeli
5、hood-75.85464 F-statistic786.0569Durbin-Watson stat0.770478 Prob(F-statistic)0.000000三、参数估计:三、参数估计:LS Y C XLS Y C XYi=106.7574+0.599781Xi第3页/共13页4(一)图形法(一)图形法 点击点击EViewsEViews方程输出窗口方程输出窗口“ResidsResids”可得残差图可得残差图四四、检验模型的、检验模型的序列相关性序列相关性第4页/共13页5n=19n=19,k=2k=2,=5%=5%查查DWDW统计表(统计表(P378P378):):dL=1.18d
6、L=1.18,dU=1.40dU=1.40D.W.=0.770478D.W.=0.770478(二二)D.W.D.W.检验法检验法正正相相关关不不能能确确定定无自相关无自相关不不能能确确定定负负相相关关0dL1.18dU1.424-dU2.64-dL2.82第5页/共13页6对对残差残差进行回归分析,生成进行回归分析,生成e e(残差序列):(残差序列):方法:方法:(1 1)主菜单)主菜单“Quick/Generate SeriesQuick/Generate Series”或工作文件或工作文件 “Procs/Generate Procs/Generate SeriesSeries”(2 2
7、)对话框中)对话框中“e=reside=resid”(3 3)使用)使用e et t进行滞后一期的自回归:进行滞后一期的自回归:EViewsEViews命今栏中命今栏中“ls e e(-1)ls e e(-1)”得回归得回归方程:方程:五、自相关问题的处理(科克伦五、自相关问题的处理(科克伦奥克特迭代法)奥克特迭代法)第6页/共13页7Dependent Variable:EMethod:Least SquaresDate:06/05/11 Time:16:41Sample(adjusted):1986 2003Included observations:18 after adjusting
8、endpointsVariableCoefficientStd.Errort-StatisticProb.E(-1)0.4960860.1775282.7944150.0125R-squared0.302715 Mean dependent var1.546119Adjusted R-squared0.302715 S.D.dependent var12.00063S.E.of regression10.02096 Akaike info criterion7.501187Sum squared resid1707.133 Schwarz criterion7.550652Log likeli
9、hood-66.51068 Durbin-Watson stat1.326733 e et t=0.496086e=0.496086et-1t-1=0.496086=0.496086对原模型进行对原模型进行广义差分广义差分,得广义差分方程,得广义差分方程:第7页/共13页8对广义差分方程进行回归:对广义差分方程进行回归:“ls Y-0.4960*Y(-1)c X-0.4960*X(-1)ls Y-0.4960*Y(-1)c X-0.4960*X(-1)”:Dependent Variable:Y-0.4960*Y(-1)Method:Least SquaresDate:06/05/11 Tim
10、e:16:52Sample(adjusted):1986 2003Included observations:18 after adjusting endpointsVariableCoefficientStd.Errort-StatisticProb.C60.447768.9638706.7434890.0000X-0.4960*X(-1)0.5832830.02940519.836010.0000R-squared0.960925 Mean dependent var231.9284Adjusted R-squared0.958483 S.D.dependent var49.34576S.
11、E.of regression10.05460 Akaike info criterion7.558377Sum squared resid1617.520 Schwarz criterion7.657307Log likelihood-66.02539 F-statistic393.4671Durbin-Watson stat1.397521 Prob(F-statistic)0.000000y*y*t t=60.4477660.44776+0.5832830.583283x x*t tDW=1.397521,k=2n=18,=5%,dL=1.16,dU=1.39DW dU,无自相关,不必再
12、迭代无自相关,不必再迭代第8页/共13页9为保证样本数不减少,可用为保证样本数不减少,可用普莱斯普莱斯温斯腾温斯腾变换补变换补充充第一个观测值(差分后不存在了)第一个观测值(差分后不存在了)六、样本容量问题六、样本容量问题方法是:方法是:本例即为:本例即为:X X1 1=397.6 Y=397.6 Y1 1=317.4=317.4 X X1 1*=345.236 Y=345.236 Y1 1*=275.598=275.598 第9页/共13页10要补充因差分损失的第一个观测值要补充因差分损失的第一个观测值 在在EViewsEViews中要生成中要生成X X和和Y Y的差分序列的差分序列X X*
13、和和Y Y*主菜单主菜单“Quick/Generate SeriesQuick/Generate Series”对话框对话框 “a=Y-0.4960*Y(-1)a=Y-0.4960*Y(-1)”OK OK得到得到广义差分序列广义差分序列a a 同理得同理得广义差分序列广义差分序列b b此时此时a a和和b b缺少第一个观测值,需补充进去缺少第一个观测值,需补充进去345.236 275.598345.236 275.598 第10页/共13页11得到得到1919个观测值的序列个观测值的序列:ls a c bls a c bDependent Variable:AMethod:Least Squ
14、aresDate:06/05/11 Time:22:22Sample:1985 2003Included observations:19VariableCoefficientStd.Errort-StatisticProb.C59.510819.1287496.5190540.0000B0.5888850.02972619.810530.0000R-squared0.958482 Mean dependent var234.2268Adjusted R-squared0.956039 S.D.dependent var48.99078S.E.of regression10.27180 Akai
15、ke info criterion7.595982Sum squared resid1793.667 Schwarz criterion7.695396Log likelihood-70.16183 F-statistic392.4570Durbin-Watson stat1.345497 Prob(F-statistic)0.000000a=a=59.5108159.51081+0.5888850.588885b b第11页/共13页12由上面的差分方程有:由上面的差分方程有:1 1=59.51081=59.51081(1-0.49601-0.4960)=118.077=118.077最终中国农村居民消费模型为最终中国农村居民消费模型为:Y Yt t=118.077+0.588885X=118.077+0.588885Xt t中国农村居民中国农村居民边际消费倾向为边际消费倾向为0.58330.5833即每即每增加收入增加收入1 1元元,将,将增加消费支出增加消费支出0.58330.5833元元第12页/共13页13感谢您的观看!第13页/共13页
限制150内