Hierarchical Bayesian analysis of the carryover effect in two-period crossover designs

In this paper, hierarchical Bayesian methods for the analysis of crossover designs are developed. First, we consider two hierarchical prior structures yielding posterior quantities that can be obtained via numerical integration. We illustrate this method with an analysis of two well-known data sets. Then we consider more flexible prior structures and a corresponding Markov chain Monte Carlo analysis of the resulting posterior distribution. We illustrate this method using data from a study conducted at Eli Lilly and Company. Throughout, we concentrate on inferences involving the carryover effect.