Robust D-Optimal Design for Response Functions with a Downturn

Abstract

Researchers studying dose-response relationships must allocate limited resources to design points in order to maximize the information gained from the study. D-optimal design is a well-described design that works efficiently to study model parameters. In order to find the D-optimal design, the model that describes the dose-response relationship has to be known. In cases where dose-response relationships show a downturn at high doses, scientists sometimes ignore the downturn to study only the increasing part of the response curve. Here we have two model choices; one describes the overall dose-response relationship, and the other describes only the increasing part of the response curve. The D-optimal designs for these two models will be different and the D-optimal design for one model may not work efficiently for the other model. This research studies robust D-optimal design, a design that works efficiently for both models.