D-Efficient Bayesian Designs for a Class of Nonlinear Models

Abstract

The present paper considers Bayesian efficient designs for nonlinear (in parameters) regression models. For studying such designs a functional approach based on Taylor series representations is implemented. This approach was developed in Melas (2006) for investigating locally optimal and maximin efficient designs. Here we construct and compare designs of these types for several specific models of practical importance. The standard D-optimality criterion is used as the optimality criterion.

Key Words:

• Nonlinear regression models
• Efficient Bayesian designs
• Locally optimal designs
• Maximin efficient designs
• Equidistant designs
• D-criterion
• Functional approach
• Michaelis-Menten model
• Rational model
• Exponential model