Statistical Analysis And Comparison By Simulation Based On Repeated Measures Data In Clinical Trials

Repeated measures data is a common data type in clinical trials.One of the characteristics is that the data is often not independent and has a certain correlation.Neglecting the correlation between data can lead to misuse of methods and wrong conclusions.Besides,the data is often missing due to individual expulsion in clinical trials which can be considered as the second feature.So,This paper mainly discusses the statistical analysis of repeated measures data from these two aspects.In this paper,I have pointed out the wrong conclusions caused by using wrong method and compared ANOVA and the mixed-effect linear models by analyzing a clinical trial example.For the complete repeated measures data,the two methods get same results.In contrast,for the data with missing values,the mixed-effect linear model is more robust.In addition,it may produce unreasonable conclusions only by comparing the AIC values to select the covariance structure when using the mixed-effect linear model.For covariance structures with close AIC values,it is recommended to combine the descriptive statistics charts or the result of ANOVA to screen the most appropriate covariance structure.In addition,the complete data and the data with different missing rates of four common covariance structures are generated through the Monte Carlo simulation.Using ANOVA and the mixed-effect linear model to analyze these data sets.It is found that both of them can control the type I error at around 0.05.For the complete repeated measures data,the two methods get close powers.In contrast,for the data with missing values,the mixed-effect linear model is more robust and the power is greater,and with the increase of the missing rate,the advantage of the mixed-effect linear model is more obvious.