一类参数变系数区域分位数回归的统计推断  

Statistical inference of a class of parametric varying coefficients in a regional quantile regression model

作　　者：程丽丝 李志强 CHENG LiSi;LI ZhiQiang(College of Mathematics and Science,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]北京化工大学数理学院,北京100029

出　　处：《北京化工大学学报(自然科学版)》2025年第2期116-123,共8页Journal of Beijing University of Chemical Technology(Natural Science Edition)

基　　金：横向课题“小样本指示性参数模型建立方法服务”(H2022568)。

摘　　要：为了解决一类参数变系数区域分位数回归的统计推断问题,基于加权复合分位数回归的思想,提出一类参数变系数区域分位数回归模型参数的有效估计算法,并利用随机加权重抽样方法构造出有限样本下模型参数显著性检验的拒绝域。数值模拟结果表明,本文所提出的检验统计量能够有效地筛选出模型中的协变量。最后将本文方法应用于海外留学数据,分析了各个因素在不同分位数下对录取机会的影响,从而选取出对录取机会有显著影响的因素。In order to solve the statistical inference problem of a class of parametric varying coefficients regional quantile regression,we propose an effective estimation of model parameters based on the idea of a weighted compos-ite quantile regression.A random weighted resampling method was used to construct a rejection domain for signifi-cance tests of parameters in the model given limited samples.The numerical simulation results indicate that the proposed test statistic can effectively screen out covariates in the model.Finally,we applied the method to over-seas study data and analyzed the impact of various factors on the chance of admission at different quantiles,in order to select the factors that have a significant impact on the chance of admission.

关 键 词：区域分位数回归模型 区域分位数估计 假设检验 随机加权重抽样 

分 类 号：O212.1[理学—概率论与数理统计]