useR! 2024

Fitting areal spatial models can be a cumbersome task, particularly when the geographical units are not well-behaved. The presence of islands, for example, gives rise to particular issues when creating neighbourhood structures based on contiguity. Further complications can arise from the presence of other natural barriers such as rivers and mountains, or man-made connectivities such as bridges, tunnels and ferry crossings. In order to create what a researcher considers to be an appropriate neighbourhood structure, incorporating all of the domain knowledge that they might have about the system, it should be simple and intuitive to add and remove connections between spatial units. Using examples from Indonesian earthquakes to London's river Thames, this session demonstrates a package which streamlines the human workflow involved in both the setting up of neighbourhood structures for spatial models, and the extraction of predictions from subsequent models. The package has a heavy emphasis on visualisation of both neighbourhood structures and model predictions and this will be reflected in the examples.