摘要
为避免模型出现过拟合,将自适应LASSO变量选择方法引入二元选择分位回归模型,利用贝叶斯方法构建Gibbs抽样算法并在抽样中设置不影响预测结果的约束条件‖β‖=1以提高抽样值的稳定性.通过数值模拟,表明改进的模型有更为良好的参数估计效率、变量选择功能和分类能力.
Binary quantile regression model with the adaptive LASSO penalty is proposed for overfitting problems by presenting a Bayesian Gibbs sampling algorithm to estimate parameters.In the process of sampling,the restriction on‖β‖ =1 is motivated to improve the stability of the sampling values.Numerical analysis show there are better improvements of the proposed method in parameter estimation,variable selection and classification.
引文
[1]Roger Koenker,Gilbert Bassett,Regression Quantiles[J].Econometrica,1978,46(1):33-50.
[2]Charles F.Manski,Maximum score estimation of the stochastic utility model of choice[J].Journal of Econometrics,1975,3(3):205-228.
[3]Keming Yu,Rana A.Moyeed,Bayesian quantile regression[J].Statistics&probability Letters,2001,54(4):437-447.
[4]Dries F.Benoit,Dirk Van den Poel,Binary quantile regression:a Bayesian approach based on the asymmetric Laplace distribution[J].Journal of Applied Econometrics,2012,27(7):1174-1188.
[5]Robert Tibshirani,Regression shrinkage and selection via the LASSO[J].Journal of the Royal Statistical Society,Series B,1996,58(1):267-288.
[6]Hui Zou,The adaptive LASSO and its oracle properties[J].Journal of the American Statistical Association,2006,101(476):1418-1429.
[7]Dries F.Benoit,Rahim Alhamzawi,Keming Yu,Bayesian lasso binary quantile regression[J].Computational Statistics,2013,28(6):2861-2873.
[8]Hussein Hashem,Veronica Vinciontti,Rahim Alhamzawi,Keming Yu.,Quantile regression with group lasso for classification[J].Advances in Data Analysis and Classification,2016,10(3):375-390.
[9]Yonggang Ji,Nan Lin,Baoxue Zhang,Model selection in binary and tobit quantile regression using the Gibbs sampler[J].Computational Statistics&Data Analysis,2012,56(4):827-839.
[10]Beong In Yun,Transformation methods for finding multiple roots of nonlinear equations[J].Applied Mathematics and Computation,2010,217(2):599-606.