Abstract: Most governments, companies and researchers are now aware of the need for data-focused decision-making. Often, however, these data are held by multiple data custodians who are unwilling or unable to share for reasons of privacy, commercial confidentiality, security or management. An appealing option is to create alternative data assets that can be made open with mild caveats. Moreover, although this might address the issue of data access, organisations face a second challenge of making sense of overwhelming amounts of data, particularly if the focus is on local decision-making.
In this presentation, we consider a number of remedies to these challenges, based on Bayesian statistical models. These include robust, privacy-preserving spatial models, probabilistic insights and novel visualisations. We discuss these in the context of the Australian Cancer Atlas, the first interactive digital atlas of small-area estimates of incidence and survival for around 20 cancers. The first release of the ACA focused on geographic inequity. The current update is focusing on spatio-temporal extensions with increased decision support. We will also touch on extensions that will support our work, including federated learning for hierarchical spatial models.
Bio: Helen Thompson is an Associate Professor of Statistics in the School of Mathematical Sciences and the Centre for Data Science at QUT. She specialises in statistical modeling and machine learning. With expertise in high-dimensional data analysis, space-time modeling, and optimum experimental design, she has made significant contributions to various fields including health, environment, and social sciences. She has published extensively in leading journals and her work provides valuable insights into complex datasets, uncovering hidden patterns and informing optimal decision-making processes in projects including Optimal Resource Extraction with BHP, Emergency Department Demand Modelling with Queensland Metro South Health and Hospital Services, Great Barrier Reef monitoring programs, and the Australian Cancer Atals.