响应缺失下部分线性变系数模型的期望回归  

Expectile regression of partially linear varying coefficient models with missing response

作　　者：周锦涛 罗双华[1] ZHOU Jin-tao;LUO Shuang-hua(College of Science,Xi'an Polytechnic University,Xi'an 710048,Shaanxi,China)

机构地区：[1]西安工程大学理学院,陕西西安710048

出　　处：《宝鸡文理学院学报(自然科学版)》2024年第4期8-14,共7页Journal of Baoji University of Arts and Sciences(Natural Science Edition)

基　　金：陕西省自然科学基金项目(2024JC-YBMS-007)。

摘　　要：目的对响应变量随机缺失情况下的部分线性变系数模型进行期望回归估计研究。方法利用逆概率加权方法处理缺失数据,B样条近似模型的变系数函数部分,期望回归模型对未知参数进行估计。结果得到了完全数据期望估计和逆概率加权期望估计。结论模拟结果和实例分析都表明逆概率加权期望回归估计表现较好。Purposes—To estimate partially linear varying coefficient models with random missing responses by using expectile regression.Methods—The inverse probability weighting method was used to deal with missing data,the varying coefficient function part of the model is approximated with B-splines and the expectile regression model is used to estimate the unknown parameters.Results—Complete data expectile estimates and inverse probability weighted expectile estimates were obtained.Conclusions—Both the simulation results and the example analyses show that the inverse probability weighted expectile regression estimates perform better.

关 键 词：缺失数据 期望回归 B样条 逆概率加权 

分 类 号：O212.1[理学—概率论与数理统计]