Zurück zu den Preprints des Jahres 2000


2000-20

Totally Positive Regression: E-Optimal Designs

by Heiligers, B..


Series: 2000-20, Preprints

MSC:
62K05 Optimal designs

Abstract:
E-optimality of approximate designs in linear regression
models is paired with a dual problem of nonlinear Chebyshev
approximation. When the regression functions form a totally
positive system, then the information matrices of designs
for subparameters turn out to be ´´almost`` totally positive,
a property which, together with the duality, allows to
solve the nonlinear Chebyshev problem. Thereby we obtain
explicit formular for E-optimal designs in terms of
equi-oscillating generalized polynomials. The considerations
unify and generalize known results on E-optimality, found
in particular regression setups.

Keywords:
Approximate design, scalar optimality, Chebyshev approximation, Chebyshev system, E-optimality, equi-oscillation, polynomial splines, total positivity, weighted polynomial regression.