Title: A model for count time series with periodic two orders autoregressive structure
Authors: Pascal Bondon - CentraleSupelec (France) [presenting]
Paulo Prezotti - Federal University of Espirito Santo (Brazil)
Valderio Anselmo Reisen - DEST-CCE-UFES (Brazil)
Marton Ispany - University of Debrecen (Hungary)
Faradiba Sarquis - Federal University of Espirito Santo (Brazil)
Abstract: A new model for count time series with conditional Poisson or geometric distribution with a periodic two orders autoregressive structure is introduced. This model is an extension of the Periodic Integer Autoregressive model of order 1 (PINAR(1) model). Stochastic properties of the model such as mean, variance, marginal and joint distributions are discussed. Moment-based and conditional maximum likelihood estimates of the parameters are presented. An alternative numerical estimation procedure, which involves less computational effort, is proposed and its performance is investigated through Monte Carlo simulations. The usefulness of the model is illustrated by an application to real data set.