Extension Of The SAEM Algorithm In Nonlinear Mixed Effects Model And Its Application In The Analysis Of Aids Clinical Data

The most suitable model is the nonlinear mixed effect model for the long-term observation of the complex unbalanced longitudinal data of AIDS clinical treatment(such as CD4?CD8?viral load,etc.),which includes measurement error,missing value and censored value.However,such models are faced with severe challenges in the processing and calculation complexity of missing value.practitioners commonly used simple linear mixed effects model,which can lead to infer that accuracy is not satisfactory.The stochastic approximation version of EM algorithm(SAEM algorithm)is a kind of maximum likelihood estimation that can be used for generalized nonlinear mixed effect models.The main advantage of the SAEM algorithm is that it can provide an estimate close to maximum likelihood estimation in very few iterations.The paper attempts to use the SAEM algorithm to build a nonlinear mixed effect model for a group of unbalanced longitudinal data in the clinical treatment of AIDS,and compares it with the linear mixed effect model commonly used in practice,in the hope of providing a good reference for the practice of modeling such complex longitudinal data with a modeling program and a reference model.For our data,we first build a linear mixed effect model based on the restricted maximum likelihood estimation method,and discuss the application of the EM algorithm in parameter estimation of nonlinear mixed effect model.Then,we extend the EM algorithm to the SAEM algorithm and apply the SAEM algorithm to parameter estimation of nonlinear mixed effect model built for our data.This study shows that the nonlinear mixed effect model can significantly better fit the clinical data of this type of AIDS than the linear mixed effect model,and the SAEM algorithm has a good effect on the parameter estimation of the nonlinear mixed effect model.