Higher-order crossover design outlier detection

The purpose of this article is to introduce a new procedure for detecting subject outlier(s) in data from a fully replicated crossover design. The suggested procedure is an adjustment to a test based on the order statistics of the two-sample Hotelling T-2, introduced by Liu and Weng. Results from an empirical power study indicate that the proposed procedure is, on average, as powerful as, for example, the equivalent procedure for detecting outliers in data from a standard crossover design.