Delineamento cross-over sob o enfoque de modelos com efeitos mistos

Abstract

The crossover is one of the designs used in Experimental Statistics. The main characteristic of this design is the application of different treatments to the same experimental unit, but in different periods of time, in order to verify if there is a significant difference among treatments. However, the effect of a previous treatment can influence the subsequent treatment and change its effect. Thus, periods without treatment between two of them are sometimes used. This period is called by washout. Therefore, there is an untreated period between them to ensure that the effect of a previous treatment does not change the effect of the subsequent treatment. In this work, we present some types of cross-over designs, namely: completely randomized blocks design with washout, Latin square design with washout, grouping of Latin squares with washout, 2x2 crossover design and a balanced design, known as William’s design. The main objective of this work is to present the description of these designs, to define how the analyzes of variances are made, including mean squares expectation. For the design which was considered the carry-over effect, we will present a practical example to illustrate the analysis of variance.