摘要
该文在删失指标随机缺失下,研究了一般线性模型的加权最小二乘回归估计;基于校准、插值和逆概率三种加权方法,分别构建了参数的估计量;在适当的假设条件下,建立了这些估计量的渐近正态性,并提出了一种新的基于最小二乘加权残差(LSWR)的Bootstrap检验程序;最后通过数值模拟和实证,分析了这些估计方法和检验程序的有效性.
This article investigates the weighted least squares regression estimators of general linear models with censoring indicators missing at random.Based on three weighting methods of calibration,interpolation,and inverse probability,parameter estimators are constructed respectively.Under appropriate assumptions,asymptotic normality of these estimators has been established,and a new bootstrap testing program based on least squares weighted residual(LSWR)is proposed.Finally,the effectiveness of these estimators and testing procedures are analyzed through numerical simulations and actual data.
作者
饶珍敏
王江峰
胡康
何姗
Zhenmin Rao;Jiangfeng Wang;Kang Hu;Shan He(School of Statistic and Mathematics,Zhejiang Gongshang University,Hangzhou 310018)
出处
《数学物理学报(A辑)》
北大核心
2025年第3期919-933,共15页
Acta Mathematica Scientia
基金
国家重点研发计划项目(2024YFA1013502)
国家自然科学基金(U23A2064,12031005)
浙江省自然科学基金(LY24A010004)
国内访问学者“教师专业发展项目”(FX2023017)。
