ANALYSIS OF REPEATED-MEASUREMENTS DATA FROM RANDOMIZED BLOCK EXPERIMENTS

In this paper we consider the problem where repeated measurements are taken on each experimental unit of a randomized experimental design. If the design is a randomized complete block, then traditionally a split-block analysis is used. Here we consider extensions of the traditional split-block analysis where we allow for several orthogonal autocorrelated error components depending on the definitions of fixed and random effects. Our approach is based on time series models applied to separate orthogonal components of the data and it can also be used for more complex designs such as Latin squares.