Empirical Analysis Of Repeated Measurement Data Based On Linear Mixed-Effects Model

In medical research,the patients' vital signs are often repeatedly measured to e-valuate the safety and effectiveness of the drug.Subject is required to be treated at the regular time points,but the loss of follow-up is not uncommon,which causes a special experimental data structure,such as missing value and the correlation between obser-vations.Traditional statistical analysis methods like regression analysis and analysis of variance,based on the principle of error decomposition,can be used to describe the relationship between response variables and explanatory variables,but cannot be applied to such data due to strict assumptions.The linear mixed-effects model can make good use of the correlation between data,and has a perfect fitting effect even though the intervals of repeated measurements are not the same.Taking clinical trial of new drug development as an example,this paper intends to apply linear mixed effect model analysis method,using SAS to make program and the results of depression scale and anxiety scale to evaluate the effectiveness of a new antidepressant medication.First,the descriptive statistics before modeling and the experimental grouping are described in detail,and we make the initial evaluation of medication based on the scale score.Then,the normality hypothesis of the observa-tions are tested.We use intercept model and slope model to explore data co-effect,discuss appropriate residual variance-covariance structure,and analyze the intraclass correlation per individual.Finally,the covariates that affect the test scores are includ-ed in the model.We fit the best model,and previous medication conjecture is proved by the final evaluated results.