B0959
Title: Local circular regression with errors-in-variables
Authors: Marco Di Marzio - University of Chieti-Pescara (Italy)
Stefania Fensore - University of Chieti-Pescara (Italy) [presenting]
Agnese Panzera - University of Florence (Italy)
Charles Taylor - University of Leeds (United Kingdom)
Abstract: Circular data are observations consisting of directions or angles. In some contexts, which also involve circular variables, data are, for some reason, not directly observable or are measured with errors. This is the case of errors-in-variables problems. Nonparametric estimation of regression functions involving circular variables is considered in the presence of measurement errors. Kernel-based approaches are proposed within different regression problems, including the case where the response is linear or binary. The asymptotic properties of the proposed estimators are discussed, along with possible generalizations and extensions. Some numerical results are provided to illustrate the performances of the proposed methods.