INLA: past, present and future
Esther Jones and Fergus Chadwick from the Offshore Renewables Group attended a two-day workshop at the ARC at the University of Glasgow. The event was organised by Dr Daniela Castro-Camilo and attended by users and developers of INLA. The aim of the workshop was to create a discussion forum for meeting colleagues (old and new!) to share experiences of using R-INLA, along with technical challenges and opportunities.
Integrated Nested Laplace Approximations (INLA) combined with Stochastic Differential Partial Equations (SPDE) are increasingly used to model complex spatial and temporal data across many disciplines such as ecology, epidemiology, and hydrology. The R libraries R-INLA and inlabru fit spatially-explicit hierarchical Bayesian models in continuous space to reflect spatial structural dependencies. The models are computationally efficient, allowing data with complex observation processes to be modelled.
Over the course of the workshop, key developers of R-INLA and inlabru gave a series of talks presenting the latest developments and new implementations that are due for release soon. Prof. Håvard Rue talked about the new developments being released in R-INLA, Dr David Bolin presented two new R libraries that will be released rSPDE and MetricGraph, and Prof. Finn Lindgren discussed the release of fmesher and the next implementation of inlabru. There were also interesting talks from staff and students about INLA applications such as Prof. Andrea Riebler who is investigating the link between education and health in the US, and Dr Sara Martino who is integrating data for robust species distribution models. Prof. Janine Illian gave us insight on the user perspective and practical challenges of inlabru.