具有时间相依协变量和信息删失的复发事件数据的半参数估计

Semi-Parametric Estimation for Recurrent Event Data with Time-Dependent Covariates and Information Censoring

该文基于计数过程的强度函数,研究一类具有时间相依协变量和信息删失的复发事件数据的半参数估计问题.首先在复发事件数据中考虑一个具有时间独立和时间相依的协变量的半参数模型,该模型允许删失时间通过未观察到的脆弱性与复发事件过程相关联.然后提出一个半参数估计方法,它不依赖于脆弱变量的分布和删失时间的分布,同时不需要对复发事件过程进行泊松假设.接着研究回归参数估计量和基准累积强度函数估计量的大样本性质,并通过重采样方法估计其渐近方差.最后通过数值模拟检验半参数模型和所提出的估计方法的合理性和优良性.

This paper investigates the semi-parametric estimation problem of a class of recurrent event data with time-dependent covariates and information censoring,based on the strength function of the counting process.We first consider a semi-parametric model with time-independent and time-dependent covariates in recurrent event data,which allows the censoring time to be correlated with the recurrent event process via an unobserved frailty.Then we propose a novel semiparametric estimation method,which depends on neither the distributions of frailty variables nor the failure times,and does not require Poisson assumption for the recurrent event process.Next we study the large sample properties of the regression parameter estimators and baseline cumulative intensity function estimators,and we estimate their asymptotic variances by resampling.Finally we test by numerical simulation the rationality and optimality of the semi-parametric model and the prposed estimation method.