A1033
Title: A semiparametric quasi-likelihood regression for circular responses
Authors: Anna Gottard - University of Firenze (Italy) [presenting]
Andrea Meilan-Vila - Universidade da Coruna (Spain)
Agnese Panzera - Università degli Studi di Firenze (Italy)
Abstract: Flexible and interpretable methods for circular outcomes are in high demand. A semiparametric regression framework is proposed for a circular response that accommodates both linear and circular predictors in its parametric and nonparametric components. Instead of assuming a specific error distribution, a circular quasi-likelihood is employed. The proposed semiparametric regression method bridges the gap among fully parametric circular regression, fully nonparametric kernel methods, and spline-based GAMs with cyclic smoothers, offering both flexibility and interpretability. The proposed regression method can be adopted in several fields, such as meteorology, for studying wind and wave direction, ecology, for animal movement and migration direction, and any context involving angular or periodic data. The asymptotic behavior of the estimators is established, and a backfitting algorithm is outlined for estimation.
