部分线性变系数模型的随机约束岭估计

A Stochastic Restricted Ridge Regression Estimator in Partially Linear Varying Coefficient Models

作为变系数模型和部分线性模型的推广,部分线性变系数模型近年来得到越来越多的关注.本文考虑该模型在线性部分自变量存在多重共线性并且参数分量附加有随机约束条件时的估计问题.基于profile最小二乘技术以及岭估计和混合估计方法,构造参数分量的profile混合岭估计,并且研究所提估计量的渐近性质.最后利用数值模拟验证所提估计方法的有效性.

As a generalization of varying coefficient model and partially linear model, partially lin- ear varying coefficient model has attracted more and more attention in recent years. This paper considers estimation of the semiparametric model when the covariates in the linear part are suffered with multi- collinearity and some additional stochastic linear restrictions on the parametric component are available. Based on the profile least-squares approach, ridge regression and mixed regression method, we propose a profile mixed estimator for the parametric component, and give its asymptotic properties. Finally, some simulation studies are conducted to illustrate the proposed procedure and the results are satisfactory.