Meeting Statistics seminar 2019 - “Modeling directional (circular) time series"

Relatore
Andrew Harvey - University of Cambridge

Abstract
Circular observations pose special problems for time series modeling. This article shows how the score-driven approach, developed primarily in econometrics, provides a natural solution to the difficulties and leads to a coherent and unified methodology for estimation, model selection and testing. The new methods are illustrated with hourly data on wind direction.

KEYWORDS: Circular data, diurnal e¤ects; dynamic conditional
score model; von Mises distribution; wind direction.

Organizzatore
Alessandra Luati