Bayesian Modelling of Time Series of Counts
Refik Soyer · 2014
In this chapter we consider time-series of correlated counts which often arise in finance, operations and marketing applications. We present a class of parameter-driven time-series models for counts using a Bayesian state-space approach. In so doing, we consider Poisson counts and present multivariate extensions to negative-binomial time-series. We develop Bayesian inference of these models using Markov chain Monte Carlo methods and present applications such as modelling of mortgage defaults, call center arrivals, and shopping trips. 1 2 CHAPTER 1. BAYESIAN MODELLING OF TIME SERIES OF COUNTS 1.