Wednesday 04 Dec 2013: Bayesian inference for a stochastic growth process
Ana Paula Palacios - Plymouth University
Plymouth University, Fitzroy 210 15:00-16:00
Usually growth processes are described using discrete time models where the mean function is deterministic and a stochastic element is introduced via an additive, random noise component. An alternative approach is to consider continuous time modelling. In this talk, we introduce a new stochastic model for growth curves which can be used to include stochastic variability into any deterministic growth function via subordination. One advantage of our approach is to be able to easily deal with data that are irregularly spaced in time or dierent curves that are observed at dierent moments of time. We examine two approaches to Bayesian inference for our model: the rst based on a Gibbsdsampler and the second based on approximate Bayesian computation. Our approach is illustrated using real Listeria growth data.