删失数据下部分线性测量误差模型的统计推断

Instrumental variables in partially linear measurement error model with right-censored data

本文重点研究了当响应变量为随机右删失数据时部分线性测量误差模型的统计推断,在假定线性测量误差的前提下,引入工具变量后通过最小二乘法来估计参数,用局部多项式估计来近似拟合非参数部分.通过数值模拟,比较了使用工具变量和其他方法对参数估计结果的影响,以及与忽略测量误差时非参数函数图像的对比.最后通过实例数据应用,展示了此方法的实际样本表现.

This paper focus on the statistical inference of the partially linear model when the response variable is random right-censored data.Under the assumption of linear measurement error,an instrumental variable is introduced and the parameter is estimated by least squared method.Local polynomial estimation is used to approximately fitting the non-parametric part.Through numerical simulation,the effect of using instrumental variable and other method on the estimation of the parameter part is compared,and the comparison of nonparametric function images when ignoring the measurement error is given.Finally,an actual sample performance is shown through real data analysis.