Balance in designed experiments with orthogonal block structure (Q1059352)

This paper is concerned with \textit{J. A. Nelder}'s ''general balance'' in designed experiments [see Proc. roy. Soc. London, Ser. A 283, 147-178 (1965; Zbl 0124.107)], which brings three aspects together, i.e. usefulness of eigenvectors of associated information matrices, combinatorial and statistical notions of balance, and the simple combinability of information from different sources in the experiment. The authors show that general balance is not only a sufficient but also a necessary condition (assuming known stratum variances) for the simple recovery of all information on every contrast from every stratum in which it is estimable. Nelder's definition of general balance is broadened to accomodate unequal treatment replications with the same sort of consequences. Several examples are provided to illustrate the developments.