A study of the performance of 2-stage adaptive optimal designs in a logistic dose-response model

It is well-known that optimal designs for logistics regression models depend on the unknown parameter values. In practice, guess values are used as proxies. Thus the actual design implemented is a pseudo optimal design. In this article, we assess whether this problem in optimality from ill-guessed parameter values can be improved by a 2-stage optimal design. We examine the optimal allocation of resources to each stage and how the 2-stage optimal design compares to the single-stage pseudo optimal design for various departures of the guess values from the true parameter values. We restrict our study to A- and D-optimality.