固定效应部分线性可加面板数据模型的惩罚二次推断估计  

Penalized Quadratic Inference Estimation of Fixed Effects Partially Linear Additive Panel Data Model

作　　者：陈建宝 丁飞鹏[2] Jian Bao CHEN;Fei Peng DING(College of Mathematics and Informatics,Fujian Normal University, Fuzhou 350117,P.R.China;College of Mathematics and Information Science,Jiangxi Normal University, Nanchang 330022,P.R.China)

机构地区：[1]福建师范大学数学与信息学院,福州350117 [2]江西师范大学数学与信息科学学院,南昌330022

出　　处：《数学学报（中文版）》2019年第1期103-122,共20页Acta Mathematica Sinica：Chinese Series

基　　金：国家社会科学基金(16BTJ018);教育部人文社会科学重点研究基地重大项目(15JJD790029);教育部人文社会科学研究项目(13YJA9100002);福建省自然科学基金(2017J01396);福建师范大学创新团队项目(IRTL1704)

摘　　要：具有较强解释力和灵活性的部分线性可加面板数据模型在各学科领域应用广泛.针对个体内存在相关结构的固定效应部分线性可加面板数据模型,本文在结合幂样条函数和最小二乘虚拟变量(LSDV)法的基础上,利用惩罚二次推断函数(PQIF)法对模型进行估计,在一定的正则条件下,证明了参数估计的渐近正态性和非参数估计的收敛性,Monte Carlo数值模拟显示所述估计方法具有良好的有限样本表现,同时,我们还将估计技术应用于实际数据分析中.Partially linear additive panel data models with strong interpretability and flexibility have been widely used in a variety of research fields. Considered a fixed effects partially linear additive panel data model with correlation structure within subjects, we derived the estimators by using penalized quadratic inference functions method under the basis of combining exponential spline function and LSDV method;the asymptotic normality of parametric estimators and convergence of nonparametric estimators were proved under suitable regular conditions;Monte Carlo simulations show that our estimates have good performances in small sample cases;meanwhile, the estimation techniques were used to analyse a real data set.

关 键 词：固定效应部分线性可加面板数据模型 惩罚二次推断函数 相关结构 

分 类 号：O212.1[理学—概率论与数理统计]