半参数非线性混合效应模型的多重估算法  

Multiple Estimation Method for Semi-parametric Nonlinear Mixed Effect Model

作　　者：齐培艳 华文杰 段西发 Qi Peiyan;Hua Wenjie;Duan Xifa(School of Applied Science,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学应用科学学院,太原030024

出　　处：《统计与决策》2021年第24期24-28,共5页Statistics & Decision

基　　金：国家自然科学基金青年项目(11701406)。

摘　　要：在HIV纵向数据的分析中常采用混合效应模型。模型中协变量可部分解释个体间的差异,但其往往带有测量误差。同时,由于人体耐药性或治疗方案的失败,病毒载量后期可能会出现反弹。为此,文章提出了含变点和测量误差的半参数非线性混合效应模型,其中,变点表示病毒载量的反弹时刻,反弹后的病毒载量轨迹用非参数函数描述。采用多重估算法对此联合模型的参数进行估计。实例分析表明所提方法具有更可靠的参数估计和较好的拟合效果。Mixed effect models are often used in HIV longitudinal data analysis. The covariates in the model can partly explain the differences between individuals, but they often have measurement errors. At the same time, viral load can rebound later due to drug resistance or the failure of treatment regimens in humans. For this reason, this paper proposes a semi-parametric nonlinear mixed effect model with change-points and measurement errors the change-points in the model represent the rebound time of the viral load, and the trajectory of the viral load after the rebound is described by a nonparametric function. The paper also uses multiple estimation method to estimate the parameters of this joint model. The example analysis shows that the proposed method has more reliable parameter estimation and better fitting effect.

关 键 词：半参数非线性混合效应模型 变点 协变量测量误差 多重估算法 

分 类 号：O212.1[理学—概率论与数理统计]