Abstract
Motivated by the need for flexible and interpretable models for circular data, this paper introduces a semiparametric regression model for a circular response that allows both linear and circular covariates in its parametric and nonparametric components. Instead of assuming a specific parametric distribution for the error term, a circular quasi-likelihood function is adopted, enabling inference when the underlying distribution is unknown. The asymptotic properties of the resulting estimators are studied and a backfitting algorithm is proposed for model estimation. Simulation studies assess the finite-sample performance of the method, and an application to the genetic effects on the migratory patterns of willow warblers illustrates its practical advantages. The results provide insights into how specific genomic elements may influence migratory behaviour.
