Eviews——AIC信息准则

Eviews——AIC信息准则董志勇博士教授北京大学经济学院Email:[e-mail]电话:133-6603-5111010-62759760AIC信息准则AIC(AkaikeInformationCriterion)/赤池信息准则用于选择模型通过AIC可以说明回归的好坏，所取得变量使得AIC越小则越好AIC的计算-(1)计算公式：AIC=-2l/n+2k/n其中：l是对数似然函数值，n为样本容量，k为系数个数对数似然函数值(LogLikelihood):2l=-[l+ln(2)+ln(RSS/n)]nAIC的计算-(2)2/knRSSAICen其中2k/n为惩罚因子有些教材直接定义AIC为ln形式2lnln()kRSSAICnnAIC准则的原理kAICkRSSAIC直接：间接：2lnln()kRSSAICnnAIC的替代方法AdjustedR-squared(调整后的判定系数R2)该指标越大越好SC（SchwarzCriterion，施瓦兹准则）:SC=-2l/n+(klogn)/nSchwarz准则引入了对增加稀疏的更大的惩罚，SC越小则越好AIC标准和SC标准有时候不统一，可以选一个AIC准则的应用(LI2-4)DependentVariable:YMethod:LeastSquaresDate:03/25/07Time:15:26Sample:19871998Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.C-180859.942190.69-4.2867250.0016X2220364.442122.475.2315180.0004R-squared0.732397Meandependentvar38953.65AdjustedR-squared0.705637S.D.dependentvar24406.49S.E.ofregression13241.81Akaikeinfocriterion21.97116Sumsquaredresid1.75E+09Schwarzcriterion22.05198Loglikelihood-129.8269F-statistic27.36878Durbin-Watsonstat0.716502Prob(F-statistic)0.000383DependentVariable:YMethod:LeastSquaresDate:03/25/07Time:15:24Sample:19871998Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.C-327701.575644.37-4.3321340.0019X2146213.849110.472.9772420.0155X3573.3049260.08402.2043070.0550R-squared0.826219Meandependentvar38953.65AdjustedR-squared0.787601S.D.dependentvar24406.49S.E.ofregression11248.17Akaikeinfocriterion21.70612Sumsquaredresid1.14E+09Schwarzcriterion21.82734Loglikelihood-127.2367F-statistic21.39463Durbin-Watsonstat1.247999Prob(F-statistic)0.000380DependentVariable:YMethod:LeastSquaresDate:03/25/07Time:15:28Sample:19871998Includedobservations:12VariableCoefficientStd.Errort-StatisticProb.C-306717.489148.80-3.4405110.0088X2142124.451901.062.7383720.0255X3570.9199271.56822.1023080.0687X4-4222.6768323.401-0.5073260.6256R-squared0.831636Meandependentvar38953.65AdjustedR-squared0.768499S.D.dependentvar24406.49S.E.ofregression11743.08Akaikeinfocriterion21.84112Sumsquaredresid1.10E+09Schwarzcriterion22.00275Loglikelihood-127.0467F-statistic13.17199Durbin-Watsonstat1.425642Prob(F-statistic)0.001839董志勇博士教授北京大学经济学院Email:[e-mail]电话:133-6603-5111010-62759760