Bayes Linear Analysis of Imprecision in Computer Models, with Application to Understanding Galaxy Formation

Vernon, Ian R.; Goldstein, M.

Abstract

Imprecision arises naturally in the context of computer models and their relation to reality. An imprecise treatment of general computer models is presented, illustrated with an analysis of a complex galaxy formation simulation known as Galform. The analysis involves several different types of uncertainty, one of which (the Model Discrepancy) comes directly from expert elicitation regarding the deficiencies of the model. The Model Discrepancy is therefore treated within an Imprecise framework to reflect more accurately the beliefs of the expert concerning the discrepancy between the model and reality. Due to the conceptual complexity and computationally intensive nature of such a Bayesian imprecise uncertainty analysis, Bayes Linear Methodology is employed which requires consideration of only expectations and variances of all uncertain quantities. Therefore incorporating an Imprecise treatment within a Bayes Linear analysis is shown to be relatively straightforward. The impact of an imprecise assessment on the input space of the model is determined through the use of an Implausibility measure.

Citation

Vernon, I. R., & Goldstein, M. (2009, July). Bayes Linear Analysis of Imprecision in Computer Models, with Application to Understanding Galaxy Formation. Presented at Sixth International Symposium on Imprecise Probability: Theories and Applications (ISIPTA'09), Durham University

• Presentation Conference Type: Conference Paper (published)
• Conference Name: Sixth International Symposium on Imprecise Probability: Theories and Applications (ISIPTA'09)
• Publication Date: Jul 1, 2009
• Deposit Date: Mar 21, 2011
• Publicly Available Date: Oct 20, 2017
• Volume: 6
• Pages: 441-450
• Book Title: Proceedings of the Sixth International Symposium on Imprecise Probability: Theories and Applications.
• Keywords: Bayesian Inference, Computer models, Calibration, Imprecise model discrepancy, Implausibility, Galaxy Formation, Graphical Representation of Model Imprecision.
• Public URL: https://durham-repository.worktribe.com/output/1158236
• Publisher URL: http://www.sipta.org/isipta09/proceedings/061.html