Acuracy of the univariate model to analyze repeated measurements through multivariate simulation

Abstract

In experiments using repeated measurements, when split plot designs are utilized the F test in relation to the subplot will have an exact F distribution only if the matrix of covariance satisfies the sphericity condition. The objective of this article was to verify throught simulation the accuracy of the analysis throught the observation of some cases when the matrix of covariance satisfies or not the sphericity condition. When the matrix of covariance was of the component type of variance and of compound symmetry, the sphericity condition was satisfied and the accuracy of the analysis adequate. This does not happen with other structures of the covariance matrix. In relation to the effects of parameters, depending on the magnitude of the values of the effects, they will interfere with the results of the tests, over estimating the effects when the amplitude of the effects is large.