相依区间删失数据下可加可乘模型的半参数分析  

Semiparametric analysis of additive-multiplicative hazards model with informative interval-censored data

作　　者：王淑影 汪童 马睿 WANG Shuying;WANG Tong;MA Rui(School of Mathematics and Statistics,Changchun University of Technology,Changchun 130002,Jilin,China;School of Mathematics and Statistics,Northeastern University at Qinhuangdao,Qinhuangdao 066000,Hebei,China)

机构地区：[1]长春工业大学数学与统计学院,吉林长春130002 [2]东北大学秦皇岛分校数学与统计学院,河北秦皇岛066000

出　　处：《云南大学学报(自然科学版)》2025年第2期201-213,共13页Journal of Yunnan University(Natural Sciences Edition)

基　　金：吉林省自然科学基金优秀青年基金(20230101371JC);吉林省“长白英才计划”青年博士后项目(202442110)。

摘　　要：区间删失数据是生存分析中常见的数据类型,且在实际问题中感兴趣的失效时间与观测时间之间具有相依性,忽略这种相依性可能会带来有偏的分析结果.在相依Ⅱ型区间删失下考虑可加可乘模型的半参数回归分析,引入了一个脆弱项来刻画观测时间和失效时间之间的相关性,并分别对失效时间和观测时间构建相应的可加可乘脆弱模型.为了估计参数,提出了一种基于伯恩斯坦多项式的Sieve极大似然估计方法,并结合EM算法对感兴趣的参数进行估计.然后,大量的模拟研究表明所提出的方法表现良好.最后将所提出的方法应用于一项年龄相关性眼病研究中产生的真实数据,分析结果表明模型拟合效果优良.Interval-censored data is a common data type in survival analysis,and the failure time we areinterested in is dependent on the observation time in practical problems.Ignoring the dependence may lead tobiased analysis results.We consider the semi-parametric regression analysis of additive-multiplicative hazardsmodels with informative CaseⅡinterval-censored data.We introduce a frailty to characterize the correlationamong the observation time and the failure time,and construct the corresponding additive-multiplicative frailtymodel for the failure time and the observation time,respectively.In order to estimate the parameters,we propose aSieve maximum likelihood estimation method based on Bernstein polynomials,and combine it with EM algorithmto estimate the parameters of interest.Then,an extensive simulation study suggests that the proposed methodperforms well.Finally,the proposed method is applied to the real data generated in the age-related eye diseasestudy,and the analysis revealed that the model exhibited excellent fitting performance.

关 键 词：相依区间删失 可加可乘风险模型 Sieve极大似然估计 EM算法 伯恩斯坦多项式 

分 类 号：O212.1[理学—概率论与数理统计]