|
Abstract:
|
For many diseases, there are geographic patterns known as spatial clusters that can indicate areas of elevated or reduced disease risk. Such areas may be indicative of an outbreak or harmful environmental exposures and identifying these clusters can help guide public health interventions. In this work, we develop a stacking approach to identify spatial clusters by using overlapping circular windows to create a set of single-cluster models. We use likelihood-based weights to stack single-cluster relative risk estimates into a meta-model, where the optimal number of parameters/space-time clusters is identified using information criteria. This enables us to calculate confidence bounds for cells inside the cluster using model-averaged tail area intervals, which we compare to several other methods using coverage and confidence bound widths. Our proposed method is further illustrated using data on female breast cancer incidence at the municipality level in Japan.
|
