On Schur Optimality.
Abstract
Schur-optimality is defined (in the general setting of a linear model) as a generalization of the well-known D-, A- and E-optimality criteria. Techniques to establish Schur-optimality are outlined, based chiefly on a process of averaging information matrices and on vector majorization. A design with a completely symmetric information matrix of maximal trace and exactly two distinct nonzero eigenvalues is proved Schur-better than a large class of designs. One description of a subcollection of designs over which Schur-optimality holds is given only in terms of the diagonal elements of the information matrices. Consequences of this are then examined in the setting on block designs. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1980
- Accession Number
- ADA092185
Entities
People
- Gregory M. Constantine
Organizations
- Indiana University Bloomington