D-Optimal Design for a Five-Parameter Logistic Model
We explore the D-optimal design for a five-parameter logistic model, which includes a shape parameter to handle asymmetries, and two threshold parameters to account for situations where the asymptotes are not at 0 and 1. The optimal design is five points, including points at -∞ and ∞ representing the thresholds. We compare the efficiencies of the optimal designs arising from the two- and five- parameter models. We find a significant loss of efficiency when the two-parameter model is used on data generated from the five-parameter model.
KeywordsOptimal Design Information Matrix Directional Derivative Markov Chain Monte Carlo Method Optimal Design Point
Unable to display preview. Download preview PDF.
The authors thank two excellent referees for their detailed comments. One of the referees found a major mistake, and the authors are deeply appreciative.
- Kalish, L. A. and J. L. Rosenberger (1978). Optimal designs for the estimation of the logistic function. Technical Report 33, Pennsylvania State University.Google Scholar
- Manukyan, Z. (2009). Sequential Designs for Estimating Toxicity and Efficacy in a Dose-Response Setting. Ph. D. thesis, George Mason University.Google Scholar