PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Particle filter estimation of duration models
Miguel Belmonte, Omiros Papaspiliopoulos and Mike Pitt
Computational Statistics and Data Analysis 2008.


n this paper we model financial durations by discrete and continuous-time point processes in state-space form (SSF). We illustrate our analysis on a duration dataset analysed in Engle (2000). For estimation of intensity and static parameters, we resort to particle filters. We compare estimates delivered by simulation-based filters, with methodology suggested in Engle (2000) and Bauwens and Veredas (2004). We conclude that the smooth particle filter (SPF) of Pitt (2002) is an efficient method for off-line parameter estimation and on-line filtering of univariate SSF duration models. We have applied the particle MCMC of Andrieu et al. (2008) to univariate models and offer a comparison with estimates from the SPF.

EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Computational, Information-Theoretic Learning with Statistics
ID Code:5048
Deposited By:Omiros Papaspiliopoulos
Deposited On:24 March 2009