Title: Modeling bivariate count series through dynamic factor models
Authors: Magda Monteiro - University of Aveiro (Portugal) [presenting]
Isabel Pereira - University of Aveiro (Portugal)
Manuel Scotto - IST-University of Lisboa (Portugal)
Abstract: Current research on count series modeling has its focus centered on multivariate models. These models either belong to the class of observation driven model or to the class of parameter driven model. Belonging to the former class is one of the first multivariate count model using a multivariate Poisson state space model. The work was later generalized by proposing a dynamic factor model for multivariate count data which allow for temporal and contemporaneous interaction between series. The aim is to present a dynamic factor model for bivariate count series whose mean vector depends on an autoregressive component beyond a common latent factor. An application of this model is made to fire activity, namely, to the monthly number of forest fires in the neighboring districts of Aveiro and Coimbra, Portugal. We use a Bayesian approach, through MCMC methods, to estimate the model parameters as well as to estimate the common factor to both series.