固定效应部分线性单指数面板模型的惩罚分位数回归  

Penalized Quantile Regression for Partially Linear Single Index Panel Models with Fixed Effect

作　　者：丁飞鹏[1] 

机构地区：[1]江西师范大学数学与信息科学学院

出　　处：《统计研究》2017年第2期101-109,共9页Statistical Research

摘　　要：分位数回归是均值回归的有益补充,该方法毋须对分布函数的具体形式做出假设,且对具有异常值或厚尾分布的数据仍具有稳健性。当前,对部分线性单指数面板模型估计方法的研究主要集中于均值回归,基于此,本文考虑了固定效应部分线性单指数面板分位数回归模型,结合B-样条函数、SCAD惩罚函数和迭代加权最小二乘法,构建了模型的估计方法,证明了估计方法的一致性和渐近正态性,同时利用Monte Carlo模拟评价了所述方法在有限样本下的表现。最后,将估计方法应用于分析碳排放的影响因素。Quantile regression is a useful supplement to mean regression. The method needn＇ t assume specification of distribution and is robust to outlie or heavy tall distribution of data. Most previous studies on partially linear single index models concentrated on mean regression. Based on this, this paper investigated quantile regression of partially linear single index panel models with fixed effect. We estimated the model based on the combination of B spline function, SCAD penalized function and reweighted least squares algorithm, and proved the consistence and asymptotic normality of estimators. Also, we evaluated the finite sample performance of the proposed method via Monte Carlo simulation. Furthermore, the proposed method is illustrated in factor analysis of carbon emissions.

关 键 词：分位数回归 部分线性单指数面板模型 样条函数 

分 类 号：C81[社会学—统计学]