Optimality of neighbour balanced designs when neighbour effects are neglected

The purpose of this paper is to study optimality properties of neighbour balanced designs when neighbour effects are not taken into account in the analysis model. Two types of designs for t treatments are considered: designs with t-1 blocks of length t and designs with t blocks of length t-1. When the designs are not randomized, new criteria on the bias matrix are introduced. When they are randomized, bias and variance are used in the comparison. In both cases neighbour balanced designs are shown to be optimal. (C) 1997 Elsevier Science B.V.