Statistica Sinica 16(2006), 789-803

LOCALLY $\mbox{\boldmath $D$}$-OPTIMAL DESIGNS FOR EXPONENTIAL

REGRESSION MODELS

Holger Dette, Viatcheslav B. Melas and Weng Kee Wong

Ruhr-Universität Bochum, St. Petersburg State University and

University of California at Los Angeles

Abstract: We study locally $D$-optimal designs for some exponential models that are frequently used in the biological sciences. The model can be written as an algebraic sum of two or three exponential terms. We show that approximate locally $D$-optimal designs are supported at a minimal number of points and construct these designs numerically.

Key words and phrases: Approximate designs, compartmental models, equivalence theorem, information matrix.