部分线性变系数模型的新复合分位数回归估计

A Novel Composite Quantile Regression Estimation for the Partial Linear Variable Coefficient Models

针对部分线性变系数模型的参数估计问题,提出了一种新复合分位数回归估计方法.利用复合分位数回归法估计参数部分,局部非线性复合分位数回归法估计变系数函数部分,并在若干正则条件下,证明了常系数和变系数函数估计量具有较好的渐近正态性质.通过随机模拟和实例分析,验证了所提估计方法在有限样本下的良好表现,有效的证明了所提方法的优越性.

Based on the parameter estimation problem of partial linear variable-coefficient models,a novel composite quantile regression estimation method is proposed.The parameter part is estimated by using the composite quantile regression method,the variable coefficient function part is estimated by the local nonlinear composite quantile regression method.And under some regular conditions,it is proved that the estimators of constant coefficient and variable coefficient functions have better asymptotic normal properties.Through stochastic simulation and a real data analysis,the good performance of the proposed estimation method under limited samples is verified,which effectively proves the superiority of the proposed method.