Variable Selection Method For Linear Mixed Effects Models And Its Application In The Analysis Of AIDS Clinical Data

In this paper,a linear mixed effect model was constructed based on a group of unbalanced longitudinal data set from AIDS Clinical Trial Group 193 A Study in the United States to compare the therapeutic effects of different therapies in the clinical trial.Two variable selection methods are used to construct the mixed effect model and are compared with each other.Firstly,we investigate the application of the penalized likelihood approach,proposed by Fan et al.for linear models,robust regression models and generalized linear models in the linear mixed effects models in 2001.Then,we derive the iterative formula of the penalized likelihood estimation equation for the parameters of the linear mixed effect model,discuss the goodness of the penalized likelihood estimation,and construct a new linear mixed effect model.Afterwards,we refer to a new variable selection method proposed by Chen Jiading in 2015,which aims to provide a simple variable selection method for Multiple Regression model.For the new method,we try to extend it to the linear mixed effects model and compare it with the penalized likelihood method proposed by Fan et al.The results of this paper are as follows:In the course of AIDS clinical practice,we believe that the efficacy of fourth treatment is better than other therapies.For variable selection method,penalized likelihood method can do variable selection and parameter estimation at the same time,but the calculation process is relatively complex,the new method by Chen is simple in calculation process,and for the slightly complex model its goodness-of-fit is slightly superior to penalized likelihood method,but variable selection and parameter estimation cannot be performed at the same time.All calculations of this paper have been completed by the statistical software R.