Title: Poisson Chain Event Graphs and Applications to Longitudinal Studies

Abstract: Chain Event Graphs (CEGs) have recently emerged as a powerful and evocative family of probabilistic graphical models. Based on a staged tree, a coloured probability tree, they are capable of modelling processes that contain asymmetries. However, up until now, they have only been explored for categorical response variables. This is a limitation in scenarios where such a natural response variable is not present, such as longitudinal studies with non-uniform observation times. In this talk, we will see CEGs extended to have Poisson response variables, with or without zero-inflation incorporated. These new methods will then be demonstrated through a study on the effects of anti-epileptic drugs on tonic-clonic seizure rates.