(完整word版)南京財(cái)經(jīng)大學(xué)計(jì)量經(jīng)濟(jì)學(xué)RESET檢驗(yàn)和異方差(考試)(完整word版)南京財(cái)經(jīng)大學(xué)計(jì)量經(jīng)濟(jì)學(xué)RESET檢驗(yàn)和異方差(考試)住房?jī)r(jià)格方程研究問(wèn)題:住房?jī)r(jià)格和住房自身特征間的關(guān)系（根據(jù)Wooldrige計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論P(yáng)219，259，288改編)數(shù)據(jù)集：hprice1變量：price：住房?jī)r(jià)格lotsize：住房占地面積（加上院子sqrft：住房面積bdrms：臥室數(shù)量1。RESET檢驗(yàn)用變量的水平值建立模型price0

lotsize1

sqrft

bdrms3SourceSSdfMSNumberofobs=88F(3,84)=57.46Model617130.7013205710.234Prob>F=0.0000Residual300723.805843580.0453R-squared=0.6724AdjR-squared=0.6607Total917854.5068710550.0518RootMSE=59.833lotsize.0020677.00064213.220.002.0007908.0033446sqrft.1227782.01323749.280.000.0964541.1491022bdrms13.852529.0101451.540.128-4.06514131.77018_cons-21.7703129.47504-0.740.462-80.3846636.84405priceCoef.priceCoef.Std.Err.tP>|t|[95%Conf.Interval]RamseyRESETtestusingpowersofthefittedvaluesofpriceHo:modelhasnoomittedvariablesF(3,81)= 4.26Prob>F= 0.0076.estatovtest,rhsRamseyRESETtestusingpowersoftheindependentvariablesHo:modelhasnoomittedvariablesF(9,75)= 6.10Prob>F= 0.0000結(jié)果說(shuō)明函數(shù)設(shè)定錯(cuò)誤，可能漏掉了自變量高次方項(xiàng)，也有可能需要對(duì)變量做變換。這個(gè)檢驗(yàn)不能得到可能存在異方差的結(jié)論。2。異方差檢驗(yàn).estathettest,iidBreusch-Pagan/Cook-WeisbergtestforheteroskedasticityHo:ConstantvarianceVariables:fittedvaluesofpricechi2(1)=9.65Prob>chi2=0.0019.estathettest,rhsiidBreusch-Pagan/Cook-WeisbergtestforheteroskedasticityHo:ConstantvarianceVariables:lotsizesqrftbdrmschi2(3)=14.09Prob>chi2=0.0028White'stestforHo:homoskedasticityagainstHa:unrestrictedheteroskedasticitychi2(9)=33.73Prob>chi2=0.0001Cameron&Trivedi'sdecompositionofIM-testSourceSourcechi2dfpHeteroskedasticity8.142.919310.00010.04320.0882Total44.78130.0000兩種檢驗(yàn)都說(shuō)明存在異方差,注意這種異方差可能是變量間函數(shù)關(guān)系設(shè)定錯(cuò)誤造成的,也可能是真正的異方差。3.對(duì)變量做對(duì)數(shù)變換，除bdrms外，因它是離散變量。擬合模型ln(price)0

ln(lotsize)1

ln(sqrft)

bdrms3SourceSSdfMSNumberofobs=88F(3,84)=50.42Model5.1550402831.71834676Prob>F=0.0000Residual2.8625632484.034078134R-squared=0.6430AdjR-squared=0.6302Total8.0176035287.092156362RootMSE=.1846llotsize.1679667.03828124.390.000.0918404.244093lsqrft.7002324.09286527.540.000.5155597.8849051bdrms.0369584.02753131.340.183-.0177906.0917074_cons-1.297042.6512836-1.990.050-2.592191-.001893lpriceCoef.lpriceCoef.Std.Err.tP>|t|[95%Conf.Interval]F(3，81)= 2.45Prob〉F= 0。0692RamseyRESETtestusingpowersoftheindependentvariablesHo:modelhasnoomittedvariablesF（9,75)= 2.01Prob〉F= 0。變換后的模型存在設(shè)定錯(cuò)誤的證據(jù)較.Breusch—Pagan/Cook-WeisbergtestforheteroskedasticityHo：ConstantvarianceVariables:fittedvaluesoflpricechi2（1）=1。39Prob>chi2=0。2379Breusch-Pagan/Cook—WeisbergtestforheteroskedasticityHo：ConstantvarianceVariables:llotsizelsqrftchi2（3） = 。22Prob>chi2= 。2383White'stestforHo:homoskedasticityagainstHa:unrestrictedheteroskedasticitychi2(9)=9.55Prob>chi2=0.3882Cameron&Trivedi'sdecompositionofIM-testSourceSourcechi2dfpHeteroskedasticity9.551.692.579310.38820.63810.1090Total13.81130.3872這些異方差檢驗(yàn)說(shuō)明，對(duì)數(shù)變換后的模型不存在異方差。至此，說(shuō)明：(1）最初的模型存在設(shè)定錯(cuò)誤；(2）最初模型的異方差也是由于函數(shù)關(guān)系錯(cuò)誤設(shè)定造成的。對(duì)因變量和自