【正文】 DurbinWatson statProb(Fstatistic)得∑e2i2=3）根據(jù)GoldfeldQuanadt檢驗，F(xiàn)統(tǒng)計量為：F=∑e2i2 /∑e1i2 =在α=，分子分母的自由度均為11，（11,11）=，因為F= （11,11）=，所以拒絕原假設，此檢驗表明模型存在異方差。HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid. dependent var. of regressionMean dependent varAdjusted RsquaredXCRsquaredDurbinWatson statProb(Fstatistic)得∑e1i2=2）當定義區(qū)間為113時，由軟件分析得：Dependent Variable: YMethod: Least SquaresDate: 12/11/14 Time: 12:21Sample: 22 34Included observations: 13VariableCoefficientStd. ErrortStatisticProb.HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid. dependent var. of regressionMean dependent varAdjusted RsquaredXCRsquaredDurbinWatson statProb(Fstatistic)得回歸模型為：Y= X+b)檢驗是否存在異方差：①用GoldfeldQuanadt檢驗如下：1）當定義區(qū)間為113時，由軟件分析得：Dependent Variable: YMethod: Least SquaresDate: 12/11/14 Time: 11:47Sample: 1 13Included observations: 13VariableCoefficientStd. ErrortStatisticProb.HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid. dependent var. of regressionMean dependent varAdjusted RsquaredXCRsquared估計結(jié)果為： Y= t=（）（）R2= F= DW=經(jīng)過檢驗發(fā)現(xiàn)，用權(quán)數(shù)w1的效果最好，所以綜上可知，即修改后的結(jié)果為：Y= t=（）（）R2= F= DW=(1)a)用Eviews模型分析得：Dependent Variable: YMethod: Least SquaresDate: 12/10/14 Time: 20:16Sample: 1978 2011Included observations: 34VariableCoefficientStd. ErrortStatisticProb.DurbinWatson statProb(Fstatistic)從上可知，nR2=，比較計算的統(tǒng)計量的臨界值，因為nR2=（2）=，所以接受原假設，該模型消除了異方差。HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid+12. dependent var. of regressionMean dependent varAdjusted RsquaredCWGT^22240181.X^2*WGT^2X*WGT^2RsquaredProb. ChiSquare(3)Test Equation:Dependent Variable: WGT_RESID^2Method: Least SquaresDate: 12/10/14 Time: 21:29Sample: 1 31Included observations: 31VariableCoefficientStd. ErrortStatisticProb.Prob. ChiSquare(3)Scaled explained SSProb. F(3,27)Obs*RsquaredSum squared resid19268334DurbinWatson stat對此模型進行White檢驗得：Heteroskedasticity Test: WhiteFstatistic. dependent var. of regressionMean dependent varAdjusted RsquaredDurbinWatson statProb(Fstatistic)Unweighted StatisticsRsquaredHannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid6320554.. dependent var. of regressionMean dependent varAdjusted RsquaredXCWeighted StatisticsRsquared估計結(jié)果為： Y= t=（）（）R2= F= DW=②用權(quán)數(shù)w2=1/x2,用回歸分析得：Dependent Variable: YMethod: Least SquaresDate: 12/09/14 Time: 21:08Sample: 1 31Included observations: 31Weighting series: W2VariableCoefficientStd. ErrortStatisticProb.DurbinWatson statProb(Fstatistic)得∑e2i2=7909670.3）根據(jù)GoldfeldQuanadt檢驗，F(xiàn)統(tǒng)計量為：F=∑e2i2 /∑e1i2 =7909670./ 1772245=在α=，分子分母的自由度均為10，（10,10）=，因為F= （10,10）=，所以拒絕原假設，此檢驗表明模型存在異方差。HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid7909670.. dependent var. of regressionMean dependent varAdjusted RsquaredX1CRsquaredDurbinWatson statProb(Fstatistic)得∑e1i2=1772245.2）定義區(qū)間為2031時，由軟件分析得：Dependent Variable: Y1Method: Least SquaresDate: 12/10/14 Time: 16:36Sample: 20 31Included observations: 12VariableCoefficientStd. ErrortStatisticProb.HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid1772245.. dependent var. of regressionMean dependent varAdjusted RsquaredX1CRsquared②GoldfeldQuanadt檢驗1）定義區(qū)間為112時，由軟件分析得：Dependent Variable: Y1Method: Least SquaresDate: 12/10/14 Time: 11:34Sample: 1 12Included observations: 12VariableCoefficientStd. ErrortStatisticProb.回歸方程是顯著的。DurbinWatson statProb(Fstatistic)得∑e2i2=3）根據(jù)GoldfeldQuanadt檢驗，F(xiàn)統(tǒng)計量為：F=∑e2i2 /∑e1i2 =在α=，分子分母的自由度均為4，（4,4）=，因為F= （4,4）=，所以接受原假設，此檢驗表明模型不存在異方差。HannanQuinn criter.FstatisticSchwarz criterionLog likelihoodAkaike info criterionSum squared resid. dependent var. of regressionMean dependent varAdjusted RsquaredTX