Title: The fractional sinusoidal waveform process
Authors: Tommaso Proietti - University of Roma Tor Vergata (Italy)
Federico Maddanu - University of Rome Tor Vergata (Italy) [presenting]
Abstract: A novel model for time series displaying persistent cycles, the fractional sinusoidal waveform process, is proposed. It is based on the simple idea of allowing the parameters that regulate the amplitude and the phase of a cycle to evolve according to a fractional noise process. While the autoregressive polynomial of the reduced form is a Gegenbauer polynomial, the main advantage of our formulation is that the autocovariance function is available in closed form. This opens the way for estimation of the parameters by exact maximum likelihood, evaluated with the support of the Durbin-Levinson algorithm.