基于Pairwise惩罚方法的面板数据模型结构变点估计  

The Estimation of Panel Data Model with Multiple Structural Breaks Based on Pairwise Penalized Method

作　　者：王纬 任燕燕[1] 刘光宗 WANG Wei;REN Yan-yan;LIU Guang-zong(School of Economics,Shandong University,Jinan 250022,China;Hengtai Securities Limited Liability Company,Beijing 100033,China)

机构地区：[1]山东大学经济学院,山东济南250100 [2]恒泰证券股份有限公司,北京100033

出　　处：《数理统计与管理》2020年第1期129-138,共10页Journal of Applied Statistics and Management

摘　　要：面板数据模型在经济、生物、统计等领域有着广泛的应用。经典的面板数据模型假设解释变量系数不随时间变化。然而在现实中,解释变量系数可能会因多种因素的影响而存在多重未知的结构变点。本文假设交互固定效应面板数据模型中含有多重未知的结构变点。研究发现通过Pairwise惩罚的参数估计方法在目标函数中增加对相邻时间解释变量系数的惩罚项,能够同时进行参数估计和结构变点诊断。蒙特卡洛模拟结果显示,不管是否存在同方差假设,该方法估计的解释变量系数均偏差较小且结构变点诊断错误率低。Panel data models are widely used in economic,biological,and statistical fields.The classic panel data model assumes that the coefficients of explanatory variables do not change over time.However,in practice,due to a variety of factors,there are multiple unknown structural changes in the coefficients of explanatory variables.This paper assumes that the panel data model with interactive fixed effects contains multiple unknown structural change points.The parameter estimation method based on Pairwise penalty increases the penalty term for the adjacent time explanatory variable coefficients in the objective function,and thus can simultaneously estimate parameters and diagnose structural change points.The results of Monte Carlo simulation show that the proposed method has a small deviation for the explanatory variable coefficient and a low diagnostic error rate of the structural change point even thought under heteroscedasticity setting.

关 键 词：交互固定效应 面板数据 结构变点 ADMM算法 

分 类 号：O212[理学—概率论与数理统计]