缺失数据下基于众数回归的变系数部分非线性模型的估计

Estimation for Varying-Coefficient Partially Nonlinear Model with Missing Response Based on Modal Regression

针对响应变量随机缺失的变系数部分非线性模型,提出了一种稳健的基于众数回归的估计方法.采取逆概率加权方法,利用QR正交分解技术,分别得到了未知参数和变系数函数的众数回归估计量.在一定条件下,证明了估计量的渐近性质.通过数值模拟和实际数据分析,说明了所提估计方法的有效性.

In this paper,we propose a robust estimation procedure for varying-coeficient partially nonlinear model with missing response based on modal regression.With the inverse probability weighed method and QR decomposition technique,the estimators of parameter and coefficient function are developed based on modal regression.Under some mild conditions,asymptotic properties of the resulting estimators are established.Furthermore,simulation studies and a real example show that the performance of the proposed estimation procedure is effective for finite samples.