Bayesian inference for Markov processes with diffusion and discrete components

摘要: SUMMARY Data arising in certain radio-tracking experiments consist of both a continuous spatial component and discrete related to behaviour. This leads naturally stochastic models with state space which is product components. We consider class such time, can be thought as diffusions random environments. They are switching diffusion or hidden Markov models, but observations made on components at time points, so that neither completely 'hidden'. describe illustrate an approach fully Bayesian inference for these general models. The algorithm used hybrid chain Monte Carlo method. parameters, the environment parameters sample path process itself updated separately, sequence, individual steps mixture Gibbs walk MetropolisHastings types. Some implementation model checking issues discussed, example using data from experiment described.