Medical Longitudinal Data Research Based On Variable Parameter Models And Mixed Models

Aim：Application of Variable Paremeter Models and Linear Mixed Models inmedical longitudinal data, have a variety of advantages compare to traditianal models.Method：Through the selective standards, estimation of paremeters and fittingprocess of the models,to illustrate the models the way being using.Result：As to the Variable Paremeter Models, we choose the vary coefficientfixed model through F test and Hausman test, the result show that symptom score oftreatment group is significantly different from the control group（t=-13.45,P<0.0001）,which means the new drug effects when treatment ofrheumatism. As to Linear Mixed Model, we choose the Unstructured model ofconvariance structure model through the evaluation criterion of-2LL,AIC,AICC,BIC,the result show that mesu is significantly different from drug groups（F=38.71，P<0.0001), and different time point make a significantly different（F=42.21，P<0.0001）, and the interaction of drug and time is significantly different, too（F=12.97，P<0.0001）. Compared with General Linear Model, the VariableParemeter Models and Linear Mixed Models could enlarge the application area andimprove the determination coefficent,make the model more suitable.Conclution： There is a number of advantages of application of VariableParemeter Models and Linear Mixed Models in medical longitudinal data.