Title: Circular time series analysis based on the projected normal distribution
Authors: Takayuki Shiohama - Tokyo University of Science (Japan) [presenting]
Takuto Kotsubo - Tokyo University of Science (Japan)
Hiroaki Ogata - Tokyo Metropolitan University (Japan)
Abstract: A new approach to a circular time series modeling is introduced which is based on the projected normal distributions with circular-circular regression. Some new perspectives on circular partial autocorrelation coefficients are introduced and its statistical inferences are discussed. Maximum likelihood estimation for the unknown model parameters and its asymptotic properties are investigated. Numerical simulations are provided to demonstrate the performance of the proposed approach. The resulting models are used to illustrate the higher order dependency of the wind direction time series.