偏态NLME和GLMM联合模型的贝叶斯推断  

Bayesian Inference of Skewed NLME and GLMM Joint Models

作　　者：宁佳男 段西发 NING Jia-nan;DUAN Xi-fa(School of Applied Sciences,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：《太原科技大学学报》2022年第3期247-253,共7页Journal of Taiyuan University of Science and Technology

基　　金：国家青年科学基金(11701406)。

摘　　要：在HIV纵向数据中通常采用联合模型来更好的描述数据,大多数文献采用线性混合效应(Linear mixed effects models,LME)和非线性混合效应(Non-linear mixed effects models,NLME)联合建模,这种联合建模忽略了对离散数据的模拟,为了更准确的描述数据,采用非线性混合效应和广义线性混合效应(Generalized linear mixed effects models GLMM)进行联合建模。另外,同时考虑病毒载入量具有偏态和左删失问题,CD4具有测量误差问题。因此,采用非线性混合效应模型拟合具有左删失和偏态误差的协变量,采用广义线性混合效应模型拟合具有测量误差的响应变量,在贝叶斯框架下,对该联合模型的参数进行参数估计。将提出的方法应用于实际HIV数据中,经验证该方法具有更好的拟合效果和更可靠的参数估计。Longitudinal data is often used in the HIV describe the data,most of the literatures use linear mixed effects(LME)and nonlinear mixed effect model(NLME),the joint model ignores the simulation of discrete data.In order to more accurately describe the data,this paper uses the nonlinear mixed effects and generalized linear mixed effects(GLMM)on joint modeling.In addition,considering that the virus load measuring tool has skewness and left-delete loss problem,CD4 with measurement error problem.Therefore,the nonlinear mixed effect model fitting with left-delete covariate and skewness error is used,as well as generalized linear mixed effects model fitting with measurement error.Based on the Bayesian framework,the parameters of the joint model are estimated,the proposed method is applied to the actual HIV data,and the proposed method is verified to have better fitting effect and more reliable parameter estimation.

关 键 词：广义线性混合效应模型 非线性混合效应模型 贝叶斯推断 左删失 正偏态分布 

分 类 号：O212.1[理学—概率论与数理统计]