以AR(3)-GARCH(2,1)模型为例：首先在主窗口输入LSRRRR（-1）（-2）（-3）得出VariableCoefficientStd.Errort-StatisticProb.RR(-1)0.0076060.0590140.1288830.8975RR(-2)0.0580050.0585490.9907070.3227RR(-3)0.1211100.0589852.0532450.0410然后在点estimate在下拉选项中选择ARCH在命令窗口中再次输入LSRRRR（-1）（-2）（-3）并在ARCH出填入2，GARCH处为1，得出结果Variancebackcast:ONGARCH=C(4)+C(5)*RESID(-1)^2+C(6)*RESID(-2)^2+C(7)*GARCH(-1)CoefficientStd.Errorz-StatisticProb.RR(-1)0.0133920.0568630.2355140.8138RR(-2)0.1204810.0621461.9386710.0525RR(-3)0.0959210.0560701.7107430.0871VarianceEquationC0.0001273.59E-053.5533270.0004RESID(-1)^2-0.0439070.029463-1.4902530.1362RESID(-2)^20.2486250.0788553.1529600.0016GARCH(-1)0.0797690.2119420.3763720.7066R-squared0.003674Meandependentvar0.001397AdjustedR-squared-0.017908S.D.dependentvar0.013305S.E.ofregression0.013423Akaikeinfocriterion-5.819411Sumsquaredresid0.049910Schwarzcriterion-5.729472Loglikelihood833.3564Durbin-Watsonstat1.974819RR是上证综合指数的周收益，用此AR（3）-GARCH（2，1）是用残差来检验超额收益的。
