短面板随机效应模型的分位数回归方法及模拟  

Simulation of Quantile Regression for Short Panel Random Effect Model

作　　者：王娜[1,2] 任燕燕[1] Wang Na Ren Yanyan(School of Economics, Shandong University, Jinan 250100, China Shandong Police College, Jinan 250014, China)

机构地区：[1]山东大学经济学院,济南250100 [2]山东警察学院,济南250014

出　　处：《统计与决策》2017年第9期14-18,共5页Statistics & Decision

基　　金：山东省自然科学基金资助项目(ZR2014AM014)

摘　　要：对于随机效应面板数据的分位数回归研究的难点在于如何处理截面内存在的相关性,文章借助分位数回归与ALD分布之间的关系,提出了带有Copula相关结构的分位数回归的极大似然估计法,其中Copula函数可用来表示短面板中的截面内相关性。通过数值优化算法迭代求解目标函数可得参数估计值,蒙特卡洛模拟结果显示该方法的均方误差更小,因此更为精确可靠。The key of solving random-effects model is how to represent the internal correlations of cross-sections. This paper, considering the relationship between quantile regression and ALD distribution, puts forward maximum likelihood estimation of quantile regression with Copula relational structure, in which Coupla function can be used to show the cross-section internal correlation of the short panel. Numerical iterative algorithm is used to obtain the parameter estimators, and Monte-Carlo simulation shows this new method has smaller mean square error, hence more effective.

关 键 词：短面板 随机效应 分位数回归 COPULA 

分 类 号：O212[理学—概率论与数理统计]