← Research overview

Bayesian research / Priors

Priors for circular models

The concentration parameter controls how tightly a circular distribution gathers around its mean direction. A prior on that parameter expresses what we assume about directional variation before observing the data.

THE RESEARCH

Penalized complexity priors

Penalized complexity (PC) priors express a preference for a simpler base model and penalize departures from it. For von Mises concentration, two base models lead to different constructions: the circular uniform distribution and a point mass. Calibration through mean resultant length or angular spread makes the prior interpretable in terms of directional variation.