Fellows Research Meetings

Experimental Design for Generalized Linear Models

Ken Russell (University of Wollongong)

Suppose that one plans to collect information from an experiment in which the response variable will be analyzed using a Generalized Linear Model (GLM), and the population parameter of interest is thought to depend on one or more continuous predictor variables. At what values of the predictor variable(s) should observations be taken?

After a very brief look at the theory for the general linear model, I will consider the case of GLMs. The problem is made difficult because the optimal design depends on the values of the parameters that one is trying to estimate. Progress in finding efficient designs is being made through the application of asymptotic results, the derivation of exact results in limited situations, and the development of computational methods. The problem is by no means solved. Some recent results will be described.

Last reviewed: 10 June, 2016

Upcoming Events

The Questionnaire Revealed, Dr. Pamela Campanelli
1 - 3 November 2017

Past Events

International Conference on Robust Statistics, July 2017