Title: Bayesian diagnostics for chain event graphs
Authors: Jim Smith - Warwick University (United Kingdom)
Rachel Wilkerson - University of Warwick (United Kingdom) [presenting]
Abstract: The class of chain event graphs has now been established as a practical Bayesian graphical tool for modeling a variety of processes. However, although a number of techniques for estimating this and performing model selection on this class have now been developed no bespoke methods of diagnostically checking representatives within this family have been yet developed. We rectify this situation and provide a number of new Bayesian diagnostics that parallel those available for the more restrictive class of Bayesian network models. These are designed to check the continued validity of the selected model as data about a population continues to be collected.