Department of Statistics, North Carolina State University, Raleigh, NC 27650, USA
Abstract:
For certain non-singularly estimable full-rank appropriately invariant problems of linear inference in the setting of block designs, some complete classes of experiments have been characterized through the relation between the relevant C-matrices and their g-inverses (of the Moore-Penrose type) in regard to the specific invariance criterion discussed here. It follows that for a -invariant non-singularly estimable full-rank problem, a complete class of experiments formally consists only of -invariant designs.