An example of universal optimality in a full-rank model (Q1108719): Difference between revisions

If in the usual model for repeated measurements designs no period effect is present, then the information matrix for the estimation of residual effects no longer has row- and column-sums zero. This means that the usual criterion for universal optimality cannot be applied. However, \textit{A. Giovagnoli} and \textit{H. P. Wynn} [Linear Algebra Appl. 67, 111-135 (1985; Zbl 0562.15007)] and \textit{A. Giovagnoli, F. Pukelsheim} and \textit{H. P. Wynn} [Group invariant orderings and experimental designs. Preprint Nr. 76, Inst. Math. Univ. Augsburg (1985)] showed a criterion which ensures universal optimality in a full-rank model. The present note shows that balanced uniform repeated measurements designs fulfill this criterion.