计量经济学实证练习作业P106第四章实证练习1⑴利用Eviews软件得到:DependentVariable:AHEMethod:LeastSquaresDate:09/21/11Time:08:44Sample:17986Includedobservations:7986VariableCoefficientStd.Errort-StatisticProb.AGE0.4519310.03352613.480220.0000C3.3241841.0022303.3167870.0009R-squared0.022254Meandependentvar16.77115AdjustedR-squared0.022131S.D.dependentvar8.758696S.E.ofregression8.661234Akaikeinfocriterion7.155842Sumsquaredresid598935.5Schwarzcriterion7.157591Loglikelihood-28571.28F-statistic181.7164Durbin-Watsonstat1.857141Prob(F-statistic)0.000000由此得出,平均每小时收入(AHE)对年龄(Age)的回归方程为:=3.324184+0.451931×Age截距的估计值是3.324184,斜率的估计值是0.451931。当工人年长一岁时,收入增加0.451931美元/小时。⑵Bob:当Age=26时,AHE=3.324184+0.451931×26=15.07439所以利用回归估计预测Bob的收入为15.07439美元/小时。Alexis:当Age=30时,AHE=3.324184+0.451931×30=16.88211所以利用回归估计预测Alexis的收入为16.88211美元/小时。⑶因为此时R-squared=0.022254比较小,所以年龄不能说明大部分的收入方差。P134第五章实证练习1=3.324184+0.451931×Age⑴原假设H0:β1=0tact=13.48022p=Pr(|Z||tact|)=2Φ(-13.48022)≈0因为p10%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p5%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p1%,所以在10%的水平下,拒绝原假设,即β1≠0。⌒⌒⌒⌒⌒⑵β1的95%的置信区间=[β1–1.96SE(β1),β1+1.96SE(β1)]=[0.451931–1.96×0.033526,0.451931+1.96×0.033526]=[-0.20518,1.109041]所以β1的95%的置信区间为[-0.20518,1.109041]。⑶只利用高中毕业生的数据重做(1),利用Eviews软件得到:DependentVariable:AGEMethod:LeastSquaresDate:10/05/11Time:09:23Sample:14346Includedobservations:4346VariableCoefficientStd.Errort-StatisticProb.AHE0.0473540.0064897.2971120.0000C29.103310.099687291.94680.0000R-squared0.012109Meandependentvar29.75725AdjustedR-squared0.011882S.D.dependentvar2.895689S.E.ofregression2.878435Akaikeinfocriterion4.952830Sumsquaredresid35991.72Schwarzcriterion4.955765Loglikelihood-10760.50F-statistic53.24784Durbin-Watsonstat1.926993Prob(F-statistic)0.000000由此得出,平均每小时收入(AHE)对年龄(Age)的回归方程为:=29.10331+0.047354×Age原假设H0:β1=0tact=7.297112p=Pr(|Z||tact|)=2Φ(-7.297112)≈0因为p10%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p5%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p1%,所以在10%的水平下,拒绝原假设,即β1≠0。⑷只利用大学毕业生的数据重做(1),利用Eviews软件得到:DependentVariable:AHEMethod:LeastSquaresDate:10/05/11Time:09:58Sample(adjusted):13640Includedobservations:3640afteradjustingendpointsVariableCoefficientStd.Errort-StatisticProb.AGE0.6903870.05368012.861140.0000C-0.2326871.604537-0.1450180.8847R-squared0.043490Meandependentvar20.30709AdjustedR-squared0.043227S.D.dependentvar9.554442S.E.ofregression9.345658Akaikeinfocriterion7.308250Sumsquaredresid317747.7Schwarzcriterion7.311656Loglikelihood-13299.01F-statistic165.4089Durbin-Watsonstat1.869725Prob(F-statistic)0.000000由此得出,平均每小时收入(AHE)对年龄(Age)的回归方程为:=-0.232687+0.690387×Age原假设H0:β1=0tact=12.86114p=Pr(|Z||tact|)=2Φ(-12.86114)≈0因为p10%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p5%,所以在10%的水平下,拒绝原假设,即β1≠0。因为p1%,所以在10%的水平下,拒绝原假设,即β1≠0。︵︵︵︵⑸βm,1–βw,1=SE(βm,1–βw,1)==0.0540707tact===-11.892448-1.96所以在5%的显著水平下年龄对收入的影响对高中毕业生和对大学毕业生是不同的。P187第七章实证练习1平均每小时收入(AHE)对年龄(Age)的回归方程为:=3.324184+0.451931×Age⑴截距估计值是3.324184,斜率估计值是0.451931。⑵利用Eviews软件得到平均小时收入(AHE)对年龄(Age),性别(Female)和教育(Bachelor)的回归:DependentVariable:AHEMethod:LeastSquaresDate:10/19/11Time:09:41Sample(adjusted):17986Includedobservations:7986afteradjustingendpointsVariableCoefficientStd.Errort-StatisticProb.BACHELOR6.8651500.17836938.488560.0000FEMALE-3.1578640.180365-17.508210.0000AGE0.4392040.03052914.386640.0000C1.8837970.9202922.0469560.0407R-squared0.189998Meandependentvar16.77115AdjustedR-squared0.189694S.D.dependentvar8.758696S.E.ofregression7.884317Akaikeinfocriterion6.968129Sumsquaredresid496180.7Schwarzcriterion6.971628Loglikelihood-27819.74F-statistic624.0988Durbin-Watsonstat1.892619Prob(F-statistic)0.000000=1.883797+0.439204×Age+6.865150×Bachelor+-3.157864×FemaleAge对收入的效应估计值是0.439204。Age系数的95%的置信区间=[β^1–1.96SE(β^1),β^1+1.96SE(β^1)]=[0.439204–1.96×0.030529,0.439204+1.96×0.030529]=[0.379367,0.499041]⑶设H0:βm,1–βw,1=0;H1:βm,1–βw,1≠0则SE(βm,1–βw,1)==0.045343tact===0.2806811.96所以不拒绝原假设,即在5%的显著水平下两个Age对AHE的效应估计没有显著差异。所以(1)中的回归没有遭遇遗漏变量偏差。⑷Bob:AHE=1.883797+0.439204×26+6.865150×0+(-3.157864)×0=13.3031所以利用回归估计预测Bob的收入为13.3031美元/小时。Alexis:AHE=1.883797+0.439204×30+6.865150×1+(-3.157864)×1=所以利用回归估计预测Alexis的收入为18.7672美元/小时。⑸在(1)中R-squared=0.022254,AdjustedR-squared=0.022131AdjustedR-squared–R-squared=0.022131–0.022254=-0.00012在(2)中R-squared=0.189694,AdjustedR-squared=0.189998AdjustedR-squared–R-squared=0.189694–0.189998=-0.00030因为(1)中的AdjustedR-squared比(2)中的AdjustedR-squared更接近1,所以(1)中的拟合效果更好。(2)中的AdjustedR-squared和R-squared很相似是因为n比较大,导致接近于1,所以AdjustedR-squared和R-squared的值很相似。⑹Female:H0:βFemale=0;H1:βFemale≠0t=-17.50821-1.96所以在5%的显著水平下拒绝原假设,即回归中不可以剔除Female。Bachelor:H0:βBachelor=0;H1:βBachelor≠0t=38.488561.96所以在5%的显著水平下拒绝原假设,即回归中不可以剔除Bachelor。Female&Bachelor:H0:βFemale+βBachelor=0;H1:βFemale+βBachelor≠0SE(βFemale+βBachelor)===0.253667tact===14.614771.96所以在5%的显著水平下拒绝原假设,即回归中不可以同时剔除Female和Bachelor。所以性别和教育是收入的决定因素。⑺多元回归遭遇遗漏变量偏差的两个条件:①至少有一个回归变量必须与遗漏变量相关;②遗漏变量必须是因变量Y的一个决定因素。遗漏变量从业时间,在这里与回归变量年龄有关,也与因变量平均小时收入有关。所以这个多元回归遭遇了遗漏变量偏差。P227第八章实证练习1⑴=1.954832+0.437093×Age+6.869575×Bachelor+-3.172725×Female若Age从25岁增加到26岁,则预期收入变化0.437093美元每小时;若Age从33岁增加到34岁,则预期收入变化0.437093美元每小时。⑵利用Eviews软件建立平均小时收入对数ln(AHE)对Age,Femal