Universal optimality of balanced uniform crossover designs

Abstract

Kunert [Ann. Statist. 12 (1984) 1006-1017] proved that, in the class of repeated measurement designs based on t treatments, p = t periods and n = λt experimental units, a balanced uniform design is universally optimal for direct treatment effects if t ≥ 3 and λ = 1, or if t ≥ 6 and λ = 2. This result is generalized to t ≥ 3 as long as λ ≤ (t - 1)/2.