Dynamic Bayesian Models via Monte Carlo - An Introduction with Examples -
G. Jóhannesson, Brian Gerald Hanley, J.J. Nitao · 2004
This report gives an introduction to a Bayesian probabilistic approach to modeling a dynamic system, with emphasis on stochastic methods for posterior inference. The Bayesian paradigm is a powerful tool to combine observed data along with prior knowledge to gain a current (probabilistic) understanding of unknown model parameters. In particular, it provides a very natural framework for updating the state of knowledge in a dynamic system. For complex systems, such updating needs to be carried out via stochastic sampling of unknown model parameters. An overview is given of the well established Markov chain Monte Carlo (MCMC) approach to achieve this and of the more recent sequential Monte Carlo (SMC) approach, which is better suited for dynamic systems. Examples are provided, including an application to event reconstruction for an atmospheric release.