A functional-algebraic determination of D-optimal designs for trigonometric regression models on a partial circle

Download

Description/Abstract

We investigate the D-optimal design problem in the common trigonometric regression model, where the design space is a partial circle. The task of maximizing the criterion function is transformed into the problem of determining an eigenvalue of a certain matrix via a differential equation approach. Since this eigenvalue is an analytic function of the length of the design space, we can make use of a Taylor expansion to provide a recursive algorithm for its calculation. Finally, this enables us to determine Taylor expansions for the support points of the D-optimal design.