竞争风险下纵列持续数据随机效应模型的估计与模拟研究

The Estimation and Simulation of the Random Model for the Longitudinal Duration Data under Competing Risks

本文论证竞争风险下纵列持续数据随机效应模型属于广义线性模型的范畴,推导出用于模型估计的等级似然函数,将等级似然估计的运用由单风险扩展到竞争风险,并进行了模拟研究。模拟结果表明,对于竞争风险下的随机效应模型,等级似然估计能够给出协变量系数相当精确的估计,克服了忽略异质性影响所导致的偏差;模拟研究还表明,本文提出的估计方法同样适用于区间观测数据。

In this paper,the random effect model for duration time under competing risk is shown to belong to GLMM（Generalized Linear Mixed Model） class,thus hierarchical likelihood method for single risk is extended to competing risks and used for estimating the model.The simulation study indicates that the hierarchical likelihood method can overcome the bias for omitting the heterogenity and give accurate estimators of regression coefficients.Simulation results also showed that this method can be used for interval observed data.