高维纵向数据的惩罚expectile估计

Penalized Expectile Estimation for High-Dimensional Longitudinal Data

基于期望分位数(expectile)回归理论,提出高维纵向数据的惩罚expectile(PGEEE)估计,在正则条件下,建立了估计量的Oracle性质.数值模拟和实证结果表明,PGEEE估计在实现变量选择的同时,提供了模型回归系数的相合估计,并且该方法可以有效识别异方差,刻画数据的异质结构,挖掘数据中更丰富的信息.

Based on the expectile regression theory,the penalty expectile(PGEEE)estimation of high-dimensional longitudinal data has been proposed,and the Oracle property of the estimator has been established under some regular conditions.Numerical simulation and empirical results demonstrate that the PGEEE estimation provides a consistent estimation of the regression coefficients of the model while realizing the variable selection.Moreover,this method can effectively identify heteroscedasticity,characterize the heterogeneous structure of the data,and mine the richer information in the data.