4逐步回归

14、逐步回归（1）后退法输入／移去的变量b模型输入的变量移去的变量方法1x6,x3,x4,x1,x5,x2.输入2.x4向后（准则:F-to-remove=.100的概率）。3.x3向后（准则:F-to-remove=.100的概率）。4.x6向后（准则:F-to-remove=.100的概率）。模型汇总模型RR方调整R方标准估计的误差1.998a.996.995191.847422.998b.996.995185.392233.998c.996.995183.431684.998d.996.995183.13396其中，模型1是全模型，从模型2到模型4依次剔除变量x4，x3，x6；最优回归子集模型4的复决定系数2R=0.996，调整的复决定系数2aR=0.995，全模型的复决定系数2R=0.996，调整的复决定系数2aR=0.995。Anovae模型平方和df均方FSig.1回归1.365E8622745165.498617.984.000a残差515276.0631436805.433总计1.370E8202回归1.365E8527294142.977794.121.000b残差515554.1631534370.278总计1.370E8203回归1.364E8434111978.5301013.814.000c残差538354.9281633647.183总计1.370E82024回归1.364E8345472040.7581355.835.000d残差570146.7741733538.046总计1.370E820系数a模型非标准化系数标准系数tSig.B标准误差试用版1(常量)1348.2252211.467.610.552x1-.641.167-1.125-3.840.002x2-.317.204-1.305-1.551.143x3-.413.548-.270-.752.464x4-.002.024-.007-.087.932x5.671.1283.7065.241.000x6-.008.008-.020-.928.3692(常量)1158.071313.3423.696.002x1-.650.129-1.140-5.031.000x2-.304.129-1.250-2.352.033x3-.422.519-.276-.814.428x5.664.0943.6667.060.000x6-.008.007-.021-1.074.3003(常量)1157.413310.0273.733.002x1-.630.126-1.106-5.019.000x2-.377.092-1.551-4.102.001x5.662.0933.6567.118.000x6-.007.007-.018-.972.3454(常量)874.586106.8668.184.000x1-.611.124-1.073-4.936.000x2-.353.088-1.454-3.994.001x5.637.0893.5167.142.0004个模型的F检验都很显著，说明y与整体的线性关系显著。最优回归子集模型4的回归方程为y=874.586-0.611x1-0.353x2+0.637x5，即y财政收入与x1农业增加值、x2工业增加值、x5社会消费总额有显著影响；其中y与x1农业增加值、x2工业增加值成反比，与x5社会消费总额成正比。3（2）逐步回归法输入／移去的变量a模型输入的变量移去的变量方法1x5.步进（准则:F-to-enter的概率=.050，F-to-remove的概率=.100）。2x1.步进（准则:F-to-enter的概率=.050，F-to-remove的概率=.100）。3x2.步进（准则:F-to-enter的概率=.050，F-to-remove的概率=.100）。模型汇总模型RR方调整R方标准估计的误差1.994a.989.988285.675772.996b.992.991247.769973.998c.996.995183.13396依次引入变量x5、x1、x2，最优回归模型3的复决定系数2R=0.996，调整后的复决定系数2aR=0.995。Anovad模型平方和df均方FSig.1回归1.354E811.354E81659.534.000a残差1550602.2441981610.644总计1.370E8202回归1.359E8267940624.8811106.706.000b残差1105019.2871861389.960总计1.370E8203回归1.364E8345472040.7581355.835.000c残差570146.7741733538.046总计1.370E8204系数a模型非标准化系数标准系数tSig.B标准误差试用版1(常量)710.36090.8887.816.000x5.180.004.99440.737.0002(常量)1011.893136.8977.392.000x5.311.0491.7186.374.000x1-.414.154-.726-2.694.0153(常量)874.586106.8668.184.000x5.637.0893.5167.142.000x1-.611.124-1.073-4.936.000x2-.353.088-1.454-3.994.0013个模型都通过了F检验。其中，最优回归模型3的回归方程为y=874.586-0.611x1-0.353x2+0.637x5,说明了y财政收入与x1农业增加值、x2工业增加值、x5社会消费总额有显著影响；其中y与x1农业增加值、x2工业增加值成反比，与x5社会消费总额成正比。