The Application Of GLMM And ANOVA With Repeated Effects In Repeated Measurement Data

Objective General ANOVA , ANOVA with repeated effects and General Linear Mixed Model (GLMM) was used to analyze the repeated measurement data respectively, whose analytic results were explained and explored particularly to obtain the two objectives: 1) What results could be only using general ANOVA to analyze without considering the correlation of repeated measurement data and tried to explain the possible reason. 2) ANOVA with repeated effects and GLMM were used to analyze the example respectively, whose analytic processes were represented particularly to explore their similarity and difference in methodologies so that we can use these methods to analyze the repeated measurement data better.Methods PROC GLM in SAS6.12 was designed for the analysis of general ANOVA and ANOVA with repeated effects, and PROC MIXED was designed for the analysis of GLMM.Results The results of general ANOVA indicated there were high statistical significance in height among different age(F=12.35, P<0.001). The age can be divided two different parts including the different age of different individuals in based-measurement and the different age of the same individual using ANOVA with repeated effects and GLMM, and the results indicated there was no statistical significance in height between the different age in based-measurement (F=0.03, P>0.05) , While there were high statistical significance in height among the different age of the same individual (F=82.1967, P<0.001) . ANOVA with repeated effects and GLMM were used to analyze the example respectively. This paper compared the two methods and found that their results were almost the same: there were high statistical significance in height.between different sex, among different measure time and the interaction of them and there were no statistical significance in height between thedifferent age in based-measurement and the interaction between that and different measure time. However, there was some key difference of the two methods in analytic functions and methodologies.Conclusions The correlation between observations and the different levels of random errors should be considered sufficiently for the selection statistical methods more correctly in the analysis of repeated measurement data. ANOVA with repeated effects (GLM) and GLMM were used to analyze the example respectively, and their analytic processes were represented particularly. This paper compared the results of them and drew the conclusions that GLMM was more flexible and various than GLM in the specification of mean model and covariance structure, and more precise than GLM in the estimations and tests of covariance parameter. So it is a suggestion that we should select appropriate statistical methods to analyze data according to the type of its design and the objective of investigator.