计量经济学 庞皓 第三版课后答案

第二章简单线性回归模型（1）①首先分析人均寿命与人均GDP的数量关系，用Eviews分析：DependentVariable:YMethod:LeastSquaresDate:12/27/14Time:21:00Sample:122Includedobservations:22VariableCoefficientStd.Errort-StatisticProb.CX1R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)有上可知，关系式为y=+②关于人均寿命与成人识字率的关系，用Eviews分析如下：DependentVariable:YMethod:LeastSquaresDate:11/26/14Time:21:10Sample:122Includedobservations:22VariableCoefficientStd.Errort-StatisticProb.CX2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)由上可知，关系式为y=+③关于人均寿命与一岁儿童疫苗接种率的关系，用Eviews分析如下：DependentVariable:YMethod:LeastSquaresDate:11/26/14Time:21:14Sample:122Includedobservations:22VariableCoefficientStd.Errort-StatisticProb.CX3R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)由上可知，关系式为y=+（2）①关于人均寿命与人均GDP模型，由上可知，可决系数为，说明所建模型整体上对样本数据拟合较好。对于回归系数的t检验：t（β1）=>(20)=，对斜率系数的显着性检验 表 关于同志近三年现实表现材料材料类招标技术评分表图表与交易pdf视力表打印pdf用图表说话 pdf 明，人均GDP对人均寿命有显着影响。②关于人均寿命与成人识字率模型，由上可知，可决系数为，说明所建模型整体上对样本数据拟合较好。对于回归系数的t检验：t（β2）=>(20)=，对斜率系数的显着性检验表明，成人识字率对人均寿命有显着影响。③关于人均寿命与一岁儿童疫苗的模型，由上可知，可决系数为，说明所建模型整体上对样本数据拟合较好。对于回归系数的t检验：t（β3）=>(20)=，对斜率系数的显着性检验表明，一岁儿童疫苗接种率对人均寿命有显着影响。（1）①对于浙江省预算收入与全省生产总值的模型，用Eviews分析结果如下：DependentVariable:YMethod:LeastSquaresDate:12/03/14Time:17:00Sample(adjusted):133Includedobservations:33afteradjustmentsVariableCoefficientStd.Errort-StatisticXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)Prob.②由上可知，模型的参数：斜率系数，截距为—③关于浙江省财政预算收入与全省生产总值的模型，检验模型的显着性：1）可决系数为，说明所建模型整体上对样本数据拟合较好。2）对于回归系数的t检验：t（β2）=>(31)=，对斜率系数的显着性检验表明，全省生产总值对财政预算总收入有显着影响。④用规范形式写出检验结果如下：Y=—t=（）R2=F=n=33⑤经济意义是：全省生产总值每增加1亿元，财政预算总收入增加亿元。（2）当x=32000时，①进行点预测，由上可知Y=—，代入可得：Y=Y=*32000—=②进行区间预测：∑x2=∑（Xi—X）2=δ2x(n—1)=x(33—1)=当Xf=32000时，将相关数据代入计算得到：≤即Yf的置信区间为（—,+）(3)对于浙江省预算收入对数与全省生产总值对数的模型，由Eviews分析结果如下：DependentVariable:LNYMethod:LeastSquaresDate:12/03/14Time:18:00Sample(adjusted):133Includedobservations:33afteradjustmentsVariableCoefficientStd.Errort-StatisticProb.LNXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①模型方程为：lnY=由上可知，模型的参数：斜率系数为，截距为③关于浙江省财政预算收入与全省生产总值的模型，检验其显着性：1）可决系数为，说明所建模型整体上对样本数据拟合较好。2）对于回归系数的t检验：t（β2）=>(31)=，对斜率系数的显着性检验表明，全省生产总值对财政预算总收入有显着影响。④经济意义：全省生产总值每增长1%，财政预算总收入增长%（1）对建筑面积与建造单位成本模型，用Eviews分析结果如下：DependentVariable:YMethod:LeastSquaresDate:12/01/14Time:12:40Sample:112Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.XCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)由上可得：建筑面积与建造成本的回归方程为：Y=（2）经济意义：建筑面积每增加1万平方米，建筑单位成本每平方米减少元。（3）①首先进行点预测，由Y=得，当x=，y=②再进行区间估计：由上表可知，∑x2=∑（Xi—X）2=δ2x(n—1)=x(12—1)=当Xf=时，将相关数据代入计算得到：即Yf的置信区间为（—,+）（1）①对百户拥有家用汽车量计量经济模型，用Eviews分析结果如下：DependentVariable:YMethod:LeastSquaresDate:11/25/14Time:12:38Sample:131Includedobservations:31VariableCoefficientStd.Errort-StatisticProb.X2X3X4CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)②得到模型得：Y=+X4③对模型进行检验：1）可决系数是，修正的可决系数为，说明模型对样本拟合较好2）F检验，F=>F（3,27）=，回归方程显着。3）t检验，t统计量分别为，，，，均大于t（27）=，所以这些系数都是显着的。④依据：1）可决系数越大，说明拟合程度越好2）F的值与临界值比较，若大于临界值，则否定原假设，回归方程是显着的；若小于临界值，则接受原假设，回归方程不显着。3）t的值与临界值比较，若大于临界值，则否定原假设，系数都是显着的；若小于临界值，则接受原假设，系数不显着。（2）经济意义：人均ＧＤＰ增加１万元，百户拥有家用汽车增加辆，城镇人口比重增加１个百分点，百户拥有家用汽车减少辆，交通工具消费价格指数每上升１，百户拥有家用汽车减少辆。（3）用EViews分析得：DependentVariable:YMethod:LeastSquaresDate:12/08/14Time:17:28Sample:131Includedobservations:31VariableCoefficientStd.Errort-StatisticProb.X2LNX3LNX4CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)模型方程为：Y=LNX4+此分析得出的可决系数为>，拟合程度得到了提高，可这样改进。（１）对出口货物总额计量经济模型，用Eviews分析结果如下：：DependentVariable:YMethod:LeastSquaresDate:12/01/14Time:20:25Sample:19942011Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.X2X3CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid8007316.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①由上可知，模型为：Y=+-②对模型进行检验：1）可决系数是，修正的可决系数为，说明模型对样本拟合较好2）F检验，F=>F（2,15）=，回归方程显着3）t检验，t统计量分别为X2的系数对应t值为，大于t（15）=，系数是显着的，X3的系数对应t值为，小于t（15）=，说明此系数是不显着的。（2）对于对数模型，用Eviews分析结果如下：DependentVariable:LNYMethod:LeastSquaresDate:12/01/14Time:20:25Sample:19942011Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.LNX2LNX3CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①由上可知，模型为：LNY=+LNX2+LNX3②对模型进行检验：1）可决系数是，修正的可决系数为，说明模型对样本拟合较好。2）F检验，F=>F（2,15）=，回归方程显着。3）t检验，t统计量分别为，，，均大于t（15）=，所以这些系数都是显着的。（3）①（1）式中的经济意义：工业增加1亿元，出口货物总额增加亿元，人民币汇率增加1，出口货物总额增加亿元。②（2）式中的经济意义：工业增加额每增加1%，出口货物总额增加%，人民币汇率每增加1%，出口货物总额增加%（1）对家庭书刊消费对家庭月平均收入和户主受教育年数计量模型，由Eviews分析结果如下：DependentVariable:YMethod:LeastSquaresDate:12/01/14Time:20:30Sample:118Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.XTCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①模型为：Y=+②对模型进行检验：1）可决系数是，修正的可决系数为，说明模型对样本拟合较好。2）F检验，F=>F（2,15）=，回归方程显着。3）t检验，t统计量分别为，，均大于t（15）=，所以这些系数都是显着的。③经济意义：家庭月平均收入增加1元，家庭书刊年消费支出增加元，户主受教育年数增加1年，家庭书刊年消费支出增加元。（2）用Eviews分析：①DependentVariable:YMethod:LeastSquaresDate:12/01/14Time:22:30Sample:118Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.TCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)②DependentVariable:XMethod:LeastSquaresDate:12/01/14Time:22:34Sample:118Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.TCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid4290746.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)以上分别是y与T，X与T的一元回归模型分别是：Y=-X=+（3）对残差进行模型分析，用Eviews分析结果如下：DependentVariable:E1Method:LeastSquaresDate:12/03/14Time:20:39Sample:118Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.E2CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)模型为：E1=+参数：斜率系数α为，截距为（3）由上可知，β2与α2的系数是一样的。回归系数与被解释变量的残差系数是一样的，它们的变化规律是一致的。（1）预期的符号是X1，X2,X3,X4,X5的符号为正，X6的符号为负（2）根据Eviews分析得到数据如下：DependentVariable:YMethod:LeastSquaresDate:12/04/14Time:13:24Sample:19942011Includedobservations:18VariableCoefficientStd.Errort-StatisticProb.X2X3X4X5X6CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①与预期不相符。②评价：1）可决系数为，数据相当大，可以认为拟合程度很好。2）F检验，F=>F（）=3,89，回归方程显着3）T检验，X1，X2,X3,X4,X5，X6系数对应的t值分别为：，，，，，均小于t（12）=，所以所得系数都是不显着的。（3）根据Eviews分析得到数据如下：DependentVariable:YMethod:LeastSquaresDate:12/03/14Time:11:12Sample:19942011Includedobservations:18VariableCoefficientStd.Errort-StatisticX5X6CProb.R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①得到模型的方程为：Y=X6+②评价：1）可决系数为，数据相当大，可以认为拟合程度很好。2）F检验，F=>F（）=3,89，回归方程显着3）T检验，X5系数对应的t值为，大于t（12）=，所以系数是显着的，即人均GDP对年底存款余额有显着影响。X6系数对应的t值为，小于t（12）=，所以系数是不显着的。（1）根据Eviews分析得到数据如下：DependentVariable:LNYMethod:LeastSquaresDate:12/05/14Time:11:39Sample:19852011Includedobservations:27VariableCoefficientStd.Errort-StatisticProb.LNGDPLNCPICR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)得到的模型方程为：LNY=LNGDPtLNCPIt（2）①该模型的可决系数为，可决系数很高，F检验值为，明显显着。但当α=时，t（24）=，LNCPI的系数不显着，可能存在多重共线性。②得到相关系数矩阵如下：LNGDP，LNCPI之间的相关系数很高，证实确实存在多重共线性。（3）由Eviews得：a）DependentVariable:LNYMethod:LeastSquaresDate:12/03/14Time:14:41Sample:19852011Includedobservations:27VariableCoefficientStd.Errort-StatisticProb.LNGDPCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)b)DependentVariable:LNYMethod:LeastSquaresDate:12/03/14Time:14:41Sample:19852011Includedobservations:27VariableCoefficientStd.Errort-StatisticProb.LNCPICR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)c)DependentVariable:LNGDPMethod:LeastSquaresDate:12/05/14Time:11:11Sample:19852011Includedobservations:27VariableCoefficientStd.Errort-StatisticProb.LNCPICR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)①得到的回归方程分别为1）LNY=LNGDPt2）LNY=LNCPIt3）LNGDPt=LNCPIt②对多重共线性的认识：单方程拟合效果都很好，回归系数显着，判定系数较高，GDP和CPI对进口的显着的单一影响，在这两个变量同时引入模型时影响方向发生了改变，这只有通过相关系数的分析才能发现。（4）建议：如果仅仅是作预测，可以不在意这种多重共线性，但如果是进行结构分析，还是应该引起注意的。（1）按照 设计 领导形象设计圆作业设计ao工艺污水处理厂设计附属工程施工组织设计清扫机器人结构设计 的理论模型，由Eviews分析得：DependentVariable:CZSRMethod:LeastSquaresDate:12/03/14Time:11:40Sample:19852011Includedobservations:27VariableCoefficientStd.Errort-StatisticCZZCGDPSSZECR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid2866884.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)Prob.从回归结果可见，可决系数为，校正的可决系数为，模型拟合的很好。F的统计量为，说明在α=,水平下，回归方程回归方程整体上是显着的。但是t检验结果表明，国内生产总值对财政收入的影响显着，但回归系数的符号为负，与实际不符合。由此可得知，该方程可能存在多重共线性。（2）得到相关系数矩阵如下：由上表可知，CZZC与GDP，CZZC与SSZE，GDP与SSZE之间的相关系数都非常高，说明确实存在多重共线性。方差扩大因子均大于10，存在严重多重共线性。并且通过以上分析，两两被解释变量之间相关性都很高。（4）解决方式：分别作出财政收入与财政支出、国内生产总值、税收总额之间的一元回归。（1）①用图形法检验绘制e2的散点图，用Eviews分析如下：由上图可知，模型可能存在异方差，②Goldfeld-Quanadt检验1）定义区间为1-7时，由软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/10/14Time:14:52Sample:17Includedobservations:7VariableCoefficientStd.Errort-StatisticTXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2得∑e1i=2）定义区间为12-18时，由软件分析得：Prob.DependentVariable:YMethod:LeastSquaresDate:12/10/14Time:13:50Sample:1218Includedobservations:7VariableCoefficientStd.Errort-StatisticProb.TXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2得∑e2i=3）根据Goldfeld-Quanadt检验，F统计量为：F=∑e2i2/∑e1i2==在α=水平下，分子分母的自由度均为4，查分布表得临界值F（4,4）=，因为F=<（4,4）=，所以接受原假设，此检验表明模型不存在异方差。（2）存在异方差，估计参数的方法：①可以对模型进行变换②使用加权最小二乘法进行计算，得出模型方程，并对其进行相关检验③对模型进行对数变换，进行分析（3）评价：所得结论是可以相信的，随机扰动项之间不存在异方差。回归方程是显着的。(1)由Eviews软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/10/14Time:16:00Sample:131Includedobservations:31VariableCoefficientStd.Errort-StatisticProb.XCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)由上表可知，2007年我国农村居民家庭人均消费支出（x）对人均纯收入（y）的模型为：Y=+（2）①由图形法检验由上图可知，模型可能存在异方差。②Goldfeld-Quanadt检验1）定义区间为1-12时，由软件分析得：DependentVariable:Y1Method:LeastSquaresDate:12/10/14Time:11:34Sample:112Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.X1CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid1772245.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2得∑e1i=1772245.2）定义区间为20-31时，由软件分析得：DependentVariable:Y1Method:LeastSquaresDate:12/10/14Time:16:36Sample:2031Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.X1CR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid7909670.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2得∑e2i=7909670.3）根据Goldfeld-Quanadt检验，F统计量为：F=∑e2i2/∑e1i2=7909670./1772245=在α=水平下，分子分母的自由度均为10，查分布表得临界值F（10,10）=，因为F=>（10,10）=，所以拒绝原假设，此检验表明模型存在异方差。（3）1）采用WLS法估计过程中，①用权数w1=1/X,建立回归得：DependentVariable:YMethod:LeastSquaresDate:12/09/14Time:11:13Sample:131Includedobservations:31Weightingseries:W1VariableCoefficientStd.Errort-StatisticProb.XCWeightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid8352726.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)UnweightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvarDurbin-Watsonstat对此模型进行White检验得：HeteroskedasticityTest:WhiteF-statisticProb.F(2,28)Obs*R-squaredProb.Chi-Square(2)ScaledexplainedSSProb.Chi-Square(2)TestEquation:DependentVariable:WGT_RESID^2Method:LeastSquaresDate:12/10/14Time:21:13Sample:131Includedobservations:31CollineartestregressorsdroppedfromspecificationVariableCoefficientStd.Errort-StatisticProb.C1045682.WGT^21173622.X*WGT^2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid+13SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)从上可知，nR2=，比较计算的统计量的临界值，因为nR2=<（2）=，所以接受原假设，该模型消除了异方差。估计结果为：Y=（）（）R2=F=DW=②用权数w2=1/x2,用回归分析得：DependentVariable:YMethod:LeastSquaresDate:12/09/14Time:21:08Sample:131Includedobservations:31Weightingseries:W2VariableCoefficientStd.Errort-StatisticXCWeightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid6320554.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb.Prob(F-statistic)UnweightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvarDurbin-Watsonstat对此模型进行White检验得：HeteroskedasticityTest:WhiteF-statisticProb.F(3,27)Obs*R-squaredProb.Chi-Square(3)ScaledexplainedSSProb.Chi-Square(3)TestEquation:DependentVariable:WGT_RESID^2Method:LeastSquaresDate:12/10/14Time:21:29Sample:131Includedobservations:31VariableCoefficientStd.Errort-StatisticProb.CWGT^22240181.X^2*WGT^2X*WGT^2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid+12SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)从上可知，nR2=，比较计算的统计量的临界值，因为nR2=<（2）=，所以接受原假设，该模型消除了异方差。估计结果为：Y=（）（）R2=F=DW=③用权数w3=1/sqr（x）,用回归分析得：DependentVariable:YMethod:LeastSquaresDate:12/09/14Time:21:35Sample:131Includedobservations:31Weightingseries:W3VariableCoefficientStd.Errort-StatisticProb.XCWeightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid9990985.SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)UnweightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvarDurbin-Watsonstat对此模型进行White检验得：HeteroskedasticityTest:WhiteF-statisticProb.F(2,28)Obs*R-squaredProb.Chi-Square(2)ScaledexplainedSSProb.Chi-Square(2)TestEquation:DependentVariable:WGT_RESID^2Method:LeastSquaresDate:12/09/14Time:20:36Sample:131Includedobservations:31CollineartestregressorsdroppedfromspecificationVariableCoefficientStd.Errort-StatisticC1212308.2141958.WGT^21301839.X^2*WGT^2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid+13SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)Prob.估计结果为：Y=（）（）R2=F=DW=经过检验发现，用权数w1的效果最好，所以综上可知，即修改后的结果为：Y=（）（）R2=F=DW=(1)a)用Eviews模型分析得：DependentVariable:YMethod:LeastSquaresDate:12/10/14Time:20:16Sample:19782011Includedobservations:34VariableCoefficientStd.Errort-StatisticXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)得回归模型为：Y=X+b)检验是否存在异方差：①用Goldfeld-Quanadt检验如下：1）当定义区间为1-13时，由软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/11/14Time:11:47Sample:113Includedobservations:13VariableCoefficientStd.Errort-StatisticXCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)Prob.Prob.得∑e1i2=2）当定义区间为1-13时，由软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/11/14Time:12:21Sample:2234Includedobservations:13VariableCoefficientStd.Errort-StatisticProb.XCR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2得∑e2i=3）根据Goldfeld-Quanadt检验，F统计量为：F=∑e2i2/∑e1i2==在α=水平下，分子分母的自由度均为11，查分布表得临界值F（11,11）=，因为F=>（11,11）=，所以拒绝原假设，此检验表明模型存在异方差。②White检验用EViews软件分析得：HeteroskedasticityTest:WhiteF-statisticProb.F(2,31)Obs*R-squaredProb.Chi-Square(2)ScaledexplainedSSProb.Chi-Square(2)TestEquation:DependentVariable:RESID^2Method:LeastSquaresDate:12/11/14Time:12:56Sample:134Includedobservations:34VariableCoefficientStd.Errort-StatisticCXX^2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresid+11SchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)从上图中可以看出，nR2=，比较计算的nR2=>异方差。用以上两种方法，可以检验模型是存在异方差的。c)修正模型1）用加权二乘法修正异方差现象步骤如下：①当权数w1=1/x时，用软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/11/14Time:13:22Sample:134Includedobservations:34Weightingseries:W1VariableCoefficientStd.Errort-StatisticXCWeightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.Prob.统计量的临界值，因为（2）=，所以拒绝原假设，不拒绝备择假设，表明模型存在Prob.F-statisticDurbin-WatsonstatProb(F-statistic)UnweightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionSumsquaredresid1489089.Durbin-Watsonstat得方程模型为：Y=（）（）R2=F=DW=对此模型进行White检验如下：HeteroskedasticityTest:WhiteF-statisticProb.F(2,31)Obs*R-squaredProb.Chi-Square(2)ScaledexplainedSSProb.Chi-Square(2)TestEquation:DependentVariable:WGT_RESID^2Method:LeastSquaresDate:12/11/14Time:11:20Sample:134Includedobservations:34CollineartestregressorsdroppedfromspecificationVariableCoefficientStd.Errort-StatisticProb.CWGT^2X*WGT^2R-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)2从上图中可以看出，nR=，比较计算的统计量的临界值，因为nR2=<响。（2）=，所以接受原假设，即该模型消除了异方差的影②当权数w2=1/x2时，用软件分析得：DependentVariable:YMethod:LeastSquaresDate:12/11/14Time:13:27Sample:134Includedobservations:34Weightingseries:W2VariableCoefficientStd.Errort-StatisticXCWeightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.F-statisticDurbin-WatsonstatProb(F-statistic)UnweightedStatisticsR-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionSumsquaredresidDurbin-Watsonstat得方程模型为：Y=+t=（）（）R2=F=DW=用White检验模型得：HeteroskedasticityTest:WhiteF-statisticProb.F(3,30)Obs*R-squaredProb.Chi-Square(3)Sc