摘要
近年来,研究预测变量带测量误差的分位数回归模型的参数估计问题已逐渐成为统计学中的一大热点问题,但由于带测量误差的分位数回归模型的参数估计十分复杂,所以相关研究很少,而目前的研究主要集中于误差分布为球形对称分布的正交回归法和误差分布自由的校正损失函数法等。在回归误差分布为正态分布的假设下,提出修正因子得分法(CFS),即在因子得分法的基础上进行参数估计,并对估计偏差进行修正得到最终估计;通过模拟研究,比较修正因子得分法相对正交回归估计(H-L估计)的优劣,并对修正因子得分法进行综合评价。
A quantile regression estimation method is developed in this paper,when the predictor is measured with error.Measurement error model has gradually become a hot issue in statistics recent years,but the parameter estimation of quantile regression model with measurement error is so complex that there are few studies on it.Concerning the quantile measurement error model,the present study mainly includes the orthogonal regression method which requires the error distribution to be spherically symmetric,and the corrected loss function method with the error distribution free,etc.Under the assumption that the regression error distribution is a normal distribution,the corrected factor-score(CFS)method is proposed,which is based on parameter estimation of factor score method,and the estimation of deviation correction to get the final estimate.Finally,through simulation studies,we compare the corrected factor-score method with the orthogonal regression estimation(H-L estimation),and then a comprehensively evaluation is given for the proposed corrected factor-score method.
作者
关静
杨香云
GUAN Jing;YANG Xiang-yun(School of Mathematics,Tianjin University,Tianjin 300354,China)
出处
《统计与信息论坛》
CSSCI
北大核心
2018年第4期13-19,共7页
Journal of Statistics and Information
