面板数据的自适应惩罚分位回归方法

Adaptive Penalty Quantile Regression for Panel Data

传统的面板数据是从均值角度进行研究,但这会受经典假设条件的约束.而考虑面板数据的分位回归模型,可以更加全面地描述响应变量条件分布的全貌.文章引入自适应惩罚函数构造了自适应惩罚的分位回归面板数据方法,并证明所提出的估计量具有大样本性质.蒙特卡洛模拟结果显示该方法相对于均值回归更具优势,是处理面板数据的有效手段.文章最后对我国居民交通通讯消费进行案例分析,得到了有利于决策的参考信息.

The study of traditional panel data is mainly based on the conditional mean regression methods, which cannot describe the variable characteristics on the different quantiles, resulting in the loss of much information. However, we can obtain a fully description of data by considering the quantile regression model for panel data. This paper presents a method to study panel data based on adaptive penalty quantile regression and a proof of its large sample properties. Monte Carlo simulation study shows that the proposed method is better than mean regression methods, which in- dicates that the proposed method is an effective way to deal with the panel data. Finally, we use the proposed method to analyze the case-study regarding the resi- dents＇ demand for transportation and communications in China, and we draw some interesting conclusions, which make great sense for the related policymakers to make decisions.