Bayes on the Beach 2010

Presentations

Adrian Barnett
Chris Strickland
David Elston
Gael Martin
James McGree
Nicole White
Sajjad Haider
Scott Sisson
Xioahui Zhang
Xibin Zhang

4th - 5th October 2010
Vibe Hotel, Surfers Paradise
Programme:(pdf 85kB)
Contact

QUT and the Statistical Society of Australia are pleased to announce that Bayes on the Beach 2010 will be held on the Gold Coast this October. The conference will provide a forum for discussion on developments and applications of Bayesian statistics. The format includes seminars, a poster session, tutorials and workshops.

International keynote speaker
David Elston is the director of Biomathematics and Statistics Scotland and honorary chair in the School of Biological Sciences at Aberdeen University.

Australian keynote speaker
Scott Sisson is a Queen Elizabeth II Research Fellow in the School of Mathematics and Statistics at the University of New South Wales and the chair of the Bayesian Statistics Section of the Statistical Society of Australia.

Keynote Speaker's Abstracts

Adaptive optimal scaling of Metropolis-Hastings algorithms

Scott Sisson
University of New South Wales

In Metropolis-Hastings algorithms it is common to manually adjust the scaling parameter of the proposal distribution so that the sampler achieves a reasonable overall acceptance probability. Some theoretical results suggest that the overall acceptance probability should be around 0.44 for univariate and 0.234 for multivariate proposal distributions. However, manually tuning the scaling parameter(s) to obtain this can be time-consuming, and impractical in high dimensions.

I'll present an adaptive method for the automatic scaling of Random-Walk Metropolis-Hastings algorithms. This method will adaptively update the scaling parameter of the proposal distribution to achieve a pre-sepecified overall acceptance probability. Our approach relies on the use of the Robbins-Monro search process, whose performance is determined by an unknown steplength constant, for which we give a very simple estimator. I'll demonstrate how to incorporate the Robbins-Monro process into Metropolis-Hastings algorithms and demonstrate its effectiveness through simulated and real data examples. The algorithm is a quick robust method for finding the scaling parameter that yields a specified acceptance probability.

Some ecological and environmental applications of Bayesian methods

David Elston
Biomathematics & Statistics Scotland

My use of Bayesian methods is motivated by the utility of MCMC to perform simulation-based inference to address the statistical issues raised by a range of ecological applications. I will describe some of these applications, and the generic issues that they raise. Examples will include the following. Firstly, a compositional analysis of streamwater on the basis of observations on chemical marker data, in which each pair of data items observed brings eight additional unknowns into the model. Secondly, a power calculation for a seabird monitoring scheme that allows for uncertainty in the estimates of parameters in the model of random variation, leading to the need to add an additional probability (the probability that the intended power is achieved) into the design specification. Thirdly, the analysis of both stock and change estimates of hedgerow lengths in the UK to provide annual estimates which can be used to assess evidence for whether change in hedgerow lengths is likely to be a principal driver for changes in population size of the yellowhammer, a farmland bird that has shown a marked decline in the UK over the last 30 years. Fourthly, an analysis of between-year variation in a common asymptotic growth curve model for length-at-date of sandeels fitted to data from two different sources, united by a model of selectivity, in which the focus of attention is on trends across years, particularly the extent to which the observed variation between years can be ascribed to environmental variables.

The QUTPay website is: https://qutpay.qut.edu.au/Cart/Default.aspx
Conference payment information is available via the Faculty of Science and Technology link on the QUTPay homepage.
Conference Registration is $340 or $220 for students.