We consider a cubic regression model for mixture experiments and discuss the improvement of designs in terms of increasing symmetry (Kiefer ordering) as well as obtaining a larger moment matrix under the standard Loewner ordering. The key problem is the characterization of Loewner com-parability of invariant moment matrices. This problem is solved using concepts from representation theory. Our investigation yields two results on complete classes of designs relative to the Kiefer ordering.
展开▼
