Statistical analysis and research conducted by the Centre for Applied Statistics (CAS), UWA 2018

The Challenge

Project 1) formed a Masters thesis for one of CAS’ students, and was largely theoretical. The broad aim of the thesis was to propose a new probabilistic interpretation for regularisation in Bayesian modelling by way of constraints.

Project 2) is an ongoing project utilising CAS’ allocation over a number of years. The broad aim of the project is to continue to improve Bayesian modelling methods in modelling and predicting daily precipitation and temperature across Australia.

The Solution

Solutions to the project aims will generally involve developing and validating improved statistical methods to achieve the require modelling. This is particularly the case in purely theoretical contexts, such as in the Masters thesis constituting project 1) above.

The Outcome

The resources provided by Pawsey are invaluable for this project, since they allow improved management of the size of the datasets examined (as in the rainfall and precipitation modelling) as well as the complexity of the modelling conducted. The latter was particularly important in 2018, due to the Bayesian nature of the modelling conducted in both projects. Bayesian statistical methods are typically extremely computationally intensive, since they involve complex approximation based on repeated data simulation.

List of Publications

Stochastic nets & Bayesian regularisation from constraints. Master’s thesis. Author: Joshua Bon. Supervisors: Berwin Turlach, Kevin Murray, Christopher Drovandi. University of Western Australia.

Analysing sensitive data from dynamically generated overlapping contingency tables. In preparation. Joshua J. Bon (UWA), Bernard Baffour (ANU),
Melanie Spallek (ACU), and Michele Haynes (ACU).