|
Abstract:
|
Bayesian optimal designs have received increasing attention in recent years. Chaloner and Larntz(1989) investigated the logistic model using Nelder-Mead algorithm. However, since Nelder-Mead algorithm is a local-optimization method, the choice of starting design points has great influence on the performance of the procedure in getting to the global optimum. We propose a hybrid computational method to find the Bayesian optimal design efficiently even for multi-variable generalized linear models. More specially, logistic and Poisson models are studied. Furthermore, the Bayesian two stage optimal designs are developed by implementing the hybrid computational method for generalized linear models, especially logistic and Poisson models. The Bayesian two stage optimal design approach is better in finding the optimal design points based on the data information observed in first stage responses.
|
