|
Activity Number:
|
306
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #303363 |
|
Title:
|
Sequential Design of Experiments for GLMs
|
|
Author(s):
|
David M. Steinberg*+ and Hovav Dror
|
|
Companies:
|
Tel Aviv University and Tel Aviv University
|
|
Address:
|
Department of Statistics and OR, Tel Aviv, International, 69978, Israel
|
|
Keywords:
|
Factorial design ; Sequential design ; Binary response ; Bayesian design ; Robbins-Monro ; Optimal design
|
|
Abstract:
|
Many experiments involve binary or count responses that don't follow a normal distribution. Efficient designs depend on the unknown parameter values, so sequential design is an attractive option. We present an effective and practical algorithm for sequential design that achieves a high degree of robustness to initial assumptions about the model parameters. The method has a Bayesian flavor, exploiting prior belief about the coefficient values. Our algorithm uses a sampling and weighting approach, rather than direct computation, to represent the posterior. We compare the performance of the algorithm with the Robbins-Monro method via simulation and with the classic Bruceton method on an actual sensitivity test conducted at an industrial plant.
|
