BAYESIAN CHANGEPOINT MODELS: A REVIEW

 

  • Prof. Leonhard Held – Department of Statistics – Ludwing-Maximilians – University Munich
  • FCUL – Campo Grande – Bloco C/6 – Piso 2 – Sala 6.2.53 -14:30 – 15h 45m
  • Sexta-feira, 1 de Abril de 2005
 

Regular Variation and Financial Time Series Models: A Review

In this talk I will review methods to analyse Poisson, binomial or Gaussian time series data with Bayesian changepoint models. I will focus on the case where the number of changepoints is also unknown and will first describe some interesting properties of this class of models. For statistical inference I will use reversible jump Markov chain Monte Carlo. Despite the apparent complexity of the model, the algorithmic details are extremely simple under suitable prior assumptions. I will illustrate the method through a number of examples.