A0928
Title: Bayes linear analysis with uncertain covariates
Authors: Samuel Jackson - Durham University (United Kingdom) [presenting]
David Woods - University of Southampton (United Kingdom)
Abstract: Statistical models typically capture uncertainties in existing knowledge of the corresponding real-world processes. However, it is less common for this uncertainty specification to capture uncertainty surrounding the values of the covariates to the model, which are often assumed to be known. General modelling methodology is developed with uncertain covariates in the context of the Bayes linear paradigm, which involves adjustment of second-order belief specifications over all the quantities of interest only, without the requirement for probabilistic specifications. In particular, an extension of commonly employed second-order modelling assumptions is proposed for the case of uncertain covariates, with explicit implementation in the context of regression analysis, stochastic process modelling, and statistical emulation. The methodology is demonstrated in the context of extracting aluminum by electrolysis and emulation of a complex network of functions.
