Abstract
The objective of investigating theoretically optimal designs is to provide a tool for identifying efficient and practical designs. However, these two lines of research have remained as separate entities. In an attempt to pull them together, this paper examines optimal design results when applied to actual practice. This paper quantifies efficiencies of strongly balanced designs relative to balanced designs. We show that, in two-period experiments for comparing more than three treatments, balanced designs perform almost as well as strongly balanced designs when the subject effects are treated as random. However, for typical clinical trials where this may not be the case, some compromise between strongly balanced designs and balanced designs is inevitable.