第二章一元线性回归模型案例一、中国居民人均消费模型从总体上考察中国居民收入与消费支出的关系。表2.1给出了1990年不变价格测算的中国人均国内生产总值（GDPP）与以居民消费价格指数（1990年为100）所见的人均居民消费支出（CONSP）两组数据。表2.1中国居民人均消费支出与人均GDP（单位：元/人）年份CONSPGDPP年份CONSPGDPP1978395.8000675.10001990797.10001602.3001979437.0000716.90001991861.40001727.2001980464.1000763.70001992966.60001949.8001981501.9000792.400019931048.6002187.9001982533.5000851.100019941108.7002436.1001983572.8000931.400019951213.1002663.7001984635.60001059.20019961322.8002889.1001985716.00001185.20019971380.9003111.9001986746.50001269.60019981460.6003323.1001987788.30001393.60019991564.4003529.3001988836.40001527.00020001690.8003789.7001989779.70001565.9001）建立模型，并分析结果。输出结果为:DependentVariable:CONSPMethod:LeastSquaresDate:07/02/08Time.20:13Sample:19782000Includedobservations:23VariableCoefficientStd.Errort-StatisticProbC201.1189114.8840213.512410.0000GDPP0.3861800.00722263474710.0000R-squared0.992710Meandependentvar905.3304Adjustedsquared0.992363S.D.dependentvar380.6334S.E.ofregression33,26450Akaikeinfocriterion9.929SOOSumsquaredresid23237.06Schwarzcriterion10,02854Loglikelihood-112.1927F-statistic2859644Durbin-Watsonstart0.550636Prob(F-statistic)0.000000对应的模型表达式为：CONSP=201.1070.3862GDPP_2（13.51）（53,47）R=0.9927,F=2859.23,DW=0.55从回归估计的结果可以看出，拟合度较好，截距项和斜率项系数均通过了t检验。中国人均消费增加10000元，GDP增加3862元。二、线性回归模型估计表2.2给出黑龙江省伊春林区1999年16个林业局的年木材采伐量和相应伐木剩余物数据。利用该数据（1）画散点图；（2）进行OLS回归；（3）预测。表2.2年剩余物yt和年木材采伐量xt数据林业局名年木材剩余物yt（万m）年木材米伐重xt（万m）乌伊岭26.1361.4东风23.4948.3iw21.9751.8红星11.5335.9五营7.1817.86.8017.0友好18.4355.0翠面11.6932.7乌马河6.8017.0美溪9.6927.3大丰7.9921.5南岔12.1535.5带岭6.8017.0朗乡17.2050.0桃山9.5030.0双丰5.5213.8合计202.87532.00（1）画散点图FileEditObjectiewFreeQuick。卫tionWindowK*LpOorkfHe:CASE1-(dViewProcObjectPrint1DehcRange:116-16obsSample:1161BobsCresidSample.GenerateSeries.Show.,GraphEmptyGroup(EditSeries)Seri_esStatisticsGtoiirStatisticsEstimateEquation.srT=HLinegraphEargraphScatter注lineTieEstimateVAR.先输入横轴变量名，再输入纵轴变量名SeriesListIDm处如仪）&口7DOKCancel得散23.2420怕.12(2) OLS估计牌EViewsFileEditObjectiewProcQuickClRtionsWindowHelp汨SIreM弹出方程设定对话框EquationEstiaationEqu&tionspcificatianDependentvariablefollowedbylistofregressorsandFDLtermw,QE皿explicitequationliktionsettingsMethod-LeastSquarts_(NLSuidA!1ASample116确定得到输出结果如图:EVieTs-Equation:UHTITLEDlorkfile:CASEHCaselL_lFilEdit0Lj#ctiewFreeQuickOptionWindowHelpView|PrDc|objeut|Priit|NwtieFreezeE优加趾巳|Foreua5H53tMRe与汨51DependentVariable:YMethod:LeastSquaresDate:O6/28AJ8Time:18:20Sample:116Includedobservations:16VariableCoefficientStd.Errort-StatisticProbC-0.7629281220966-0.6248560.5421X0.4042800.03337712.112G60.0000R-squared0.912S90Meandependentvar12.67937AdjustedR-squared0.906668SDdependentvar6665466S.E.ofregnession2036319Akaikeinfocriterion4.376633Sumsquaredresid58.05231S匚hwdracriterion4473207Loglikelihood33.01306F-statistic14671B6Durbin-Watsonstat1401946Prob(F-statistic)0.000000由输出结果可以看出，对应的回归表达式为：%-0.76290.4043xt(-0.625)(12.11)R2=0.9129,F=146.7166,DW=1.48(3) x=20条件下模型的样本外预测方法首先修改工作文件范围ViewsFileEditObjectViewProeRuickOjtiorLSWiitd&wHelplorkfile：CASE1-(1二课件宜大vf1)DisplayFilter:*SampleMe,im应耳ObjectPrintsaved温Ik+Usho内因记RJstore口目曰te归bm：RanStSample.=Structure/ResireCurrentFige.CwAppendtoCurrentPage.XContractCurrentfigs.yReshapeCurrentPageCopy/ExtractfromCurrentPw管士SirtCurrentP4g.将工作文件范围从116改为117IXIWorkfilestructyrttypeD4tewpeeiEiestionDated-regularfrequencyStirtEndFrequencyorkfilestructure确定后将工作文件的范围改为包括17个观测值，然后修改样本范围EVievsFileEditObjectViewFroc业tickOgtioueWindowHelpforkfile:CASE1-(dtKttXdata2casel.rf1)回Vtew忸其小33次日Print5weDe3ik+/-5h口碑57115七011GerirSampleSetSample.Structurft/E&sizeCurrentF自售电.AppendtoCurrentPage.ContractCurrentFage.ReshapeCm-rentFsgeCcpy/EutractfromCurrentPageSortCurrentFag4r,.将样本范围从116改为117Samplerangepirs(orw铀pleobjecttocopy)1LTIFcndition(options!)anc1打开x的数据文件，利用Edit+/-给x的第17个观测值赋值为20回C必resid0Y需EVi.w-E(iuatian=WTITLEDlorkfile:CASE1=:CaselQofeJFileEdit0tjectViewProcuickOtionsirtdowHelp.iw;(Pro匚Jobjei叵毗岫11司曰一酬EftinnateFarecmstl./tats限sidgDependentVariable:YMethod:LeastSquaresDate:06/29/08Time:18:17Sample(adjusted):116I