Allan Clark will present the Department of Statistical Science's seminar with a talk entitled, "Approximate Bayesian occupancy models".
Presence-absence data is often used to investigate the range and range shifts of animal populations using occupancy models that includes covariates assumed to capture the probability of species occurrence as well as covariates assumed to determine the probability of species detection. To date, likelihood as well as Bayesian methods have been developed for these types of models. The Bayesian models are often analysed using WinBugs or OpenBugs because Markov chain Monte Carlo methods are required to sample from the posterior distribution of the required parameters. We develop various variational Bayes approximations to the posterior distribution of the parameters of various occupancy models (spatial and non-spatial). The effectiveness of the proposed techniques are investigated via simulations and data extracted from the Southern African Bird Atlas Project (SABAP) is used to illustrate the usefulness of the techniques