Title: Optimum designs for generalised linear models for bivariate response
Authors: Alexander Donev - University of Manchester (United Kingdom) [presenting]
Abstract: The development of new products often requires their characteristics to be studied experimentally. When the manufacturing process is prone to gradual changes in the performance of the equipment between calibrations, such changes can be adjusted for by incorporating their effect in the required statistical models for each of the studied characteristics. Different generalised linear models have to be estimated for each of the product characteristics. The construction of a suitable D-optimum experimental design that will be used to collect data for estimating these models requires the use of a composite design criterion. Its choice needs to be done carefully. An example based on an industrial experiment where the response is bivariate will be provided.