在相依删失下限定平均寿命的差异比较

Estimating the difference of restricted mean life under dependent censor

在临床医学及流行病学等研究中,研究人员经常会关心患者经过某种治疗后的平均寿命.由于删失的存在,使得生存函数的尾部估计偏差较大,在实际问题中通常考虑限定平均寿命作为衡量处理功效的指标.本文针对非随机化分组的治疗功效差异问题,考虑生存时间同时存在独立和相依两种删失情形下的限定平均寿命差异推断问题.本文利用两种模型分别解释两种类型的混杂因子,建立比例风险模型用以解释基准协变量,建立加性风险模型用以解释依时协变量,利用逆概率删失权方法给出模型参数的估计并讨论估计量的大样本性质.通过随机模拟给出估计方法在偏度和精度方面的表现.最后,将本文给出的方法用于肝硬化患者两种治疗方法的功效差异分析.

In clinical medicine and epidemiology,the mean life of patients after a certain treatment is of interest.Due to the censor,the estimation of survival function tail is not accurate.Alternatively,the restricted mean life is usually considered.For measuring the difference of treatment,under non-randomized grouping,we estimate the restricted mean life while the independent and dependent censor data exist simultaneously.We construct the additive hazard model for explaining dependent variate and proportional hazard model for survival time.Inverse probability censoring weighted method is used to estimate the parameters of the models and the large sample of the estimator is discussed.The numerical simulation of parameter estimation is presented.Finally,the method we provide is applied to liver transplant data.