Autoregressive Models for Count Time Series


Séminaire Probabilités & Statistique

13/01/2011 - 14:00 Konstantinos Fokianos (Department of Mathematics & Statistics, University of Cyprus) Salle 1 - Tour IRMA

Count time series are observed in diverse applications, for instance consider the number of transaction per minute of some stock, or the monthly number of people with a certain disease, and so on.
For the analysis of these data, there has been developed a number of models based either on thinning operator or on GLM framework. We will
be examining the second class of models which include a feedback mechanism. Such models are expected, in general, to be more parsimonious, pretty much as is the case of GARCH models. It is
important therefore to study their statistical properties and develop algorithms for estimation. In particular we will be discussing some multivariate generalizations of previously established models in the literature.