Some optimal designs for repeated measurements with autoregressive errors

Abstract

A number of individuals are observed at equally spaced time points; different individuals behave independently but the observations on the same individual have the autocorrelation structure of a first-order autoregressive process. We are concerned with the optimal allocation of treatments, assuming that at each time point a new selection must be made. Conditions for optimality are given for models without nuisance parameters, and also with covariates and with block effects.