The final BRAG meeting for this year approaches! SPECIAL BRAG will be held this Thursday 30/11 at 1:30 pm via Zoom/GP-Y801. Note that we will be starting 30 minutes later than usual.
What’s a SPECIAL BRAG? In these sessions, presenters will have 10-15 minutes to discuss their research, with a focus on work that is close or already published. Following the presentation, an external reviewer/assessor/panelist/colleague (the discussant) will have an additional 10-15 minutes to summarize the research and highlight relevant discussion points and questions. What makes SPECIAL BRAG unique is that the discussant may not be an expert in the presenter’s field, offering fresh perspectives and an opportunity for both the presenter and discussant to develop their skills in scholarly communication.
This week we will have a presentation by myself (@James) with @Aiden as the discussant.
Zoom Link: https://qut.zoom.us/j/7952714168?omn=87181364211
Title: Summarising area level cancer risk factors – using the Bayesian model ‘hammer’ to hit the `index` nail
Abstract: The increasing granularity of small area level health data enables detailed spatial analysis but also presents challenges, particularly in creating composite indicators or indices that summarize this multidimensional data. However, existing methods to generate such indices often overlook both spatial autocorrelation and measurement error. Furthermore, they typically rely on a single index, which may inadequately capture the complexity of the data. We extend the Bayesian Shared Component Model (SCM), commonly used for creating indices, to accommodate heteroscedastic measurement error and spatial priors. This generalization allows for the development of multiple, more meaningful, and robust area-level health indices. The proposed Generalized SCM bridges the gap between conventional SCMs and Bayesian spatial factor models, a connection that has yet to be acknowledged. Implemented in Stan, our proposed model features the first practical use of the Leroux spatial prior. We apply the proposed model to Australian cancer risk factor data, creating the first area-level cancer risk index, which reveals spatial disparities and aids in targeted interventions.
Thanks,
Jamie & Scott
BRAG Co-Chairs
Bayesian Research & Applications Group 