【正文】 看出社會固定資產投資總額對陜西省建筑業(yè)總產值的影響相對比較大。（一） 系數(shù)分析 大β系數(shù) = = ==*= * ==*= *=統(tǒng)計（經濟）含義：X1（人民可支配收入）每變化一個標準差,Y（陜西省建筑業(yè)總產值）。（3） 對多重可決系數(shù)的解釋統(tǒng)計含義：剔除解釋變量數(shù)目和樣本容量的影響，%可由X1和X2來解釋，%歸因于隨機誤差項中所包含因素陜西省建筑業(yè)總產值的影響。 （2） 對多重可決系數(shù)的解釋統(tǒng)計含義：%可由X1和X2來解釋，%歸因于隨機誤差項中所包含因素對Y的影響。經濟意義：（1）當社會固定資產投資總額不變時，人民可支配收入增加或減少1億元。調整后的模型為：=++（1） 對j的解釋統(tǒng)計意義：（1）當X2保持不變時，X1增加或減少1個單位。下面檢驗異方差：x4e22x5e22大小FstatisticF臨界值有表可知：F檢驗值均小于臨界值，所以消除了異方差。下面進行異方差處理：Genr E2=abs(resid)Genr YE2=Y/E2Genr CE2=1/E2Genr X4E2=X4/E2Genr X5E2=X5/E2產生新數(shù)據如下：obsCE2YE2X4E2X5E21976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009對新數(shù)據進行回歸Dependent Variable: YE2Method: Least SquaresDate: 06/11/11 Time: 14:16Sample: 1976 2009Included observations: 34VariableCoefficientStd. ErrortStatisticProb. CE2X4E2X5E2Rsquared Mean dependent varAdjusted Rsquared . dependent var. of regression Akaike info criterionSum squared resid Schwarz criterionLog likelihood DurbinWatson stat進行異方差檢驗：x4x5大小FstatisticF臨界值如圖可知，F(xiàn)檢驗值均小于臨界值，說明沒有了異方差。故存在自相關。故舍去該三元模型，尋找新的模型。下面進行異方差檢驗：x1e32x4e32x5e32大小FstatisticF臨界值由表可知，F(xiàn)值均小于臨界值，所以不存在異方差。下面進行異方差檢驗：x1x4x5大小FstatisticF臨界值如表可知：F檢驗值均大于臨界值，所以存在異方差，下面進行異方差處理：第一次異方差處理：Genr E3=abs(resid)Genr YE3=Y/E3Genr CE3=1/E3Genr X1E3=X1/E3Genr X4E3=X4/E3Genr X5E3=X5/E3產生新數(shù)據，如下：obsCE3YE3X1E3X4E3X5E31976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009對新數(shù)據進行回歸：Dependent Variable: YE3Method: Least SquaresDate: 06/11/11 Time: 11:29Sample: 1976 2009Included observations: 34VariableCoefficientStd. ErrortStatisticProb. CE3X1E3X4E3X5E3Rsquared Mean dependent varAdjusted Rsquared . dependent var. of regression Akaike info criterionSum squared resid Schwarz criterionLog likelihood DurbinWatson stat進行異方差檢驗：x1e3x4e3x5e3大小FstatisticF臨界值如表可知，F(xiàn)值均小于臨界值，所以不存在異方差。七、 三元模型的異方差、自相關處理對Y C X1 X4 X5進行回歸Dependent Variable: YMethod: Least SquaresDate: 06/11/11 Time: 10:58Sample: 1997 2009Included observations: 13VariableCoefficientStd. ErrortStatisticProb. CX1X4X5Rsquared Mean dependent varAdjusted Rsquared . dependent var. of regression Akaike info criterionSum squared resid Schwarz criterionLog likelihood FstatisticDurbinWatson stat Prob(Fstatistic)查表可知：（,）之間的，沒有自相關。但是X2E44與X2E44的系數(shù)為負，即社會消費品零售總額和人民可支配收入與建筑業(yè)總產值之間為負相關關系，與現(xiàn)實不符。=，故存在自相關，下面進行自相關處理：第一次自相關處理：Dependent Variable: YE43Method: Least SquaresDate: 06/11/11 Time: 10:16Sample(adjusted): 1977 2009Included observations: 33 after adjusting endpointsConvergence achieved after 8 iterationsVariableCoefficientStd. ErrortStatisticProb. CE43X1E43X2E43X4E43X5E43AR(1)Rsquared Mean dependent varAdjusted Rsquared . dependent var. of regression Akaike info criterionSum squared resid Schwarz criterionLog likelihood DurbinWatson statInverted AR Roots .46Genr ye44=ye43*ye43(1)Genr ce44=ce43*ce43(1)Genr x1e44=x1e43*x1e43(1)Genr x2e44=x2e43*x2e43(1)Genr x4e44=x4e43*x4e43(1)Genr x5e44=x5e43*x5e43(1)代入數(shù)值：Genr ye44=*ye43(1)Genr ce44=*ce43(1)Genr x1e44=*x1e43(1)Genr x2e44=*x2e43(1)Genr x4e44=*x4e43(1)Genr x5e44=*x5e43(1)補數(shù)據Ye44’1=*ye43Ce44’1=*ce43x1e44’1=*x1e43x2e44’1=*x2e43x4e44’1=*x4e43x5e44’1=*x5e43產生新序列：obsCE44YE44X1E44X2E44X4E44X5E441976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009對新序列進行自相關和異方差檢驗：Dependent Variable: YE44Method: Least SquaresDate: 06/11/11 Time: 10:41Sample: 1976 2009Included observations: 34VariableCoefficientStd. ErrortStatisticProb. CE44X1E44X2E44X4E44X5E44Rsquared Mean dependent varAdjusted Rsquared . dependent var. of regression Akaike info criterionSum squared resid Schwarz criterionLog likelihood DurbinWatson stat由表可知，=,落在了（,）之間，所以不存在自相關下面進行異方差檢驗：x1e