Linear Mixed Model In The Aids Clinical Prediction

Analysis of variance and regression analysis are the primary means of statistical modeling. However,the basic assumptions of these methods are that residual or error is independently and identically distributed, Normality and homogeneity of variance. Mixed model is a new important method of statistical modeling, which allows relaxing the constraint of the independence assumption and permits more complex data structures such as repeated measure data, longitudinal data.To serve this purpose, this paper systematically summaries the basic theory of linear mixed model. With the aid of SAS program, random coefficient model and multilevel model are adopted to fit the data respectively in evaluating the treatments effects and other relating problerms.This paper mainly includes the following:(1) The linear mixed model(LMM) is outlined mainly the development random coefficient model and multilevel model. The superiority of LMM over general linear model is illustrated by examples; Estimation and testing are introduced, emphasizing the necessity of using SAS-Package.(2) Random coefficient model and multilevel model with suitably modification, are proposed to model regression of efficacy on treatments, patient ages, cure times. From this model, stopping time of every treatment, the best treatment and the comparison of different treatments on different ages are obtained.This paper highlights the flexible use of the linear mixed model in modeling longitudinal data. This paper only gives few of many candidate models in the analysis of AIDS treatment. Further comparison of these models to other reasonable model is needed to investigated. Only by this way can the proposed models be developed to a high level.