河海大学 预测与决策 part 4 - Regression and Econometric Methods.pptVIP

6.2.1回归系数B的估计采用最小二乘法估计，设观察值与模型估计值的残差为E，则其中(6-3)根据最小平方法要求，应有即6.2估计回归参数由极值原理，根据矩阵求导法则，对B求导，并令其等于零，则得整理得回归系数向量B的估计值(6-4)6.2估计回归参数6.2.2二元线性回归方程回归系数的估计二元线性回归方程为此时得出的计算公式如下：(6-5)6.2估计回归参数(6-8)(6-7)(6-6)6.2估计回归参数5/4SimpleRegressionandtheCorrelationCoefficient5/4/1ComputationalConsiderationsRegressionModelwhereCorrelationItisasimplemattertowriteanequationlinkingtheslopeandthecorrelationcoefficient,asfollows:(5-25)5/4SimpleRegressionandtheCorrelationCoefficient5/4/2InterpretationoftheCoefficientofDeterminationThecoefficientofdetermination,R2,willalwaysbepositive,andtheintuitivemeaningcanbestatedasfollows:R2tellstheproportionofvarianceinYthatcanbeexplainedbyX.TheratioofthesetwovariancesisR2R2=varianceofthevalues/varianceoftheYvalues=explainedvarianceofY/totalvarianceofY=explainedsumofsquares/totalsumofsquares(5-26)5/4SimpleRegressionandtheCorrelationCoefficientTable5-9indicateshowtheexplainedandtotalsumofsquaresaredetermined:5/4SimpleRegressionandtheCorrelationCoefficientColumns1andcolumns2arethetrueconsumption(C)andtheestimatedconsumption()ofnaturalgas—basedonthesimplelinearregression--andthesedataaretakenfromTable5-3.Thecolumn3andcolumn4showthedeviationsfromthemeanforCand,andcolumn5and6showthesquaresofthesedeviations.Thesumofthesquareddeviationsincolumn6dividedbythesumofthesquareddeviationsincolumn5yieldR2,asfollows:whereYisthedependentvariablewithmean.Figure5-14helpstomakethisclearer.R2=explainedSS/totalSS=13006/20838=0.6245/4SimpleRegressionandtheCorrelationCoefficient5/4SimpleRegressionandtheCorrelationCoefficientThecolumn3andcolumn4showthedeviationsfromthemeanforCand,