Bayesian research / Circular covariate
Regression with a circular covariate
An angle can also carry information about a linear response. Wind direction and wind speed provide one example. Their relationship must account for the fact that the angular coordinate repeats after a full turn.
THE RESEARCH
Joint circular models
A joint circular model treats the angular quantity as a response in its own right and links it to a linear response through a shared latent predictor. This represents their association while carrying uncertainty about the circular component into the joint analysis. Shared and response-specific AR(2) processes describe temporal variation, as illustrated by wind direction and speed.
RELATED PAPERA Bayesian regression framework for circular models with INLA
Ye, Van Niekerk & Rue (2026)