Circular regression using Bayesian unwrapping

Circular data arise in a number of biological and physical applications. Circular regression refers to the study of the dependence of a circular response variable on a collection of explanatory variables. In this paper the circular response variable is modelled as a wrapped Gaussian process. Previously, estimation with wrapped processes has been performed used complicated iterative optimisation or random sampling techniques. The recursive Bayesian algorithm proposed here is simple to implement and computationally economical by comparison. The proposed algorithm is applied to phase parameter estimation.