REPEATED MEASURES DESIGN WITH ERROR SERIAL DEPENDENCE UNDER VIOLATION OF HOMOGENEITY ASSUMPTION

Abstract

Two analital procedures for a randomized experimental design with several consecutive measures taken through time in each one of the experimental units, are examined in this paper. One of them consists in facing data applying AVAR mixed model with modeled error structure by AR processes. On the other side, the problem is also faced from a more general perspective, using a repeated measures multivariate approach. Data were simulated by Monte Carlo procedures to investigate the effect that no satisfaction of independence, sphericity and homogeneity assumptions have on the bias degree of estimated parameters, on the empirical robability to make type I errors, and on the power of statistical test of the two procedures.