A Spatial Mixture Model to Account for Risk Discontinuities: Analyzing Attempted Suicide in Waterloo Region

What:
Presentation
When:
Friday Sep 30   10:50 AM to 11:10 AM (20 minutes)
Where:
Tags:
spatial analysisDisease mappingBayesian modelingSpatial structureSpatial smoothingHealth
Discussion:
0
Bayesian spatial analysis of small-area data often models spatially structured random effects via the intrinsic conditional autoregressive prior distribution (ICAR). For outcomes exhibiting abrupt variations in risk between adjacent areas, spatial smoothing imposed by ICAR may challenge the identification of high risk areas. This research explores a mixture model that weights spatially structured and unstructured random effects, increasing sensitivity to risk discontinuities. The outcome analyzed is attempted suicide in the Region of Waterloo, Ontario, Canada. Compared to a conventional non-mixture Bayesian spatial model, diagnostics suggest that the mixture model has superior fit. Maps of spatially structured random effects from mixture and non-mixture models highlight specific small-areas where the mixture model captures greater levels of residual spatial structure.
Participant
University of Waterloo
Participant

My Schedule

Add to Your Schedule