CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1162
Title: Sequential multi-objective planning of factorial experiments Authors:  Olga Egorova - King's College London (United Kingdom) [presenting]
Steven Gilmour - KCL (United Kingdom)
Abstract: In numerous experimental settings, especially in the ones pursuing multiple objectives and/or with the lack of preliminary knowledge about the process under study, laying out the whole design at the very beginning is a big decision to make -- and in an extensive amount of practical situations, such rigidity is neither necessary nor beneficial. Whether there is a need to run a pilot study first, followed by a series of more focused experimentations, or a situation with the model uncertainty involving combining various, often contradicting optimality criteria. Ideally, we would want to break down the decision-making process, such that each stage is planned (a) making use of the previously obtained data, and (b) accounting for the primary objectives of the following stages. In this most recent work, we explore optimal sequential planning for a series of factorial experiments in the presence of potential model misspecification, so that the quality of the inference from the fitted model is optimised together with minimising the lack-of-fit. We investigate the issues of updating the criteria, models and ensuring the coherence of the results obtained at different stages.