Ioannis Ntzoufras

Publications Page


Bayesian Hypothesis Testing for the Distribution of Insurance Claim Counts Using the Gibbs Sampler

Athanassios Katsis and Ioannis Ntzoufras (2005)

Journal of Computational Methods in Science and Engineering, 5, 201-214.

 

    This paper presents an MCMC algorithm used to test the alternative hypotheses concerning three distributions commonly used to model the marginal claim counts in actuarial science. The proposed methodology involves advanced techniques of Bayesian modeling making use of Gibbs sampling in a similar manner as Gibbs variable selection algorithm of Dellaportas et al. (2002).  The advantage of this approach in favor of the alternative  reversible jump algorithm (Green, 1995) is the straightforward implementation using the MCMC language tool of  WINBUGS (Spiegelhalter et al. 2003). Results are presented for real data sets.

Keywords: Generalized Poisson,  Gibbs Sampling,  Hypothesis Tests,  Lagrangian Poisson,  Markov Chain Monte Carlo;  Mixed Poisson, Negative Binomial, Poisson, Reversible Jump, WINBUGS.

[Download paper from journal's site]

[Download associated WINBUGS code files here]

[Download associated WINBUGS results  here]



Back to Research PageAll Contents Copyright.
Last revised: 23/11/2011