Adaptive MCMC Methods for Inference on Discretely Observed Affine Jump Diffusion Models.

In the present paper we generalize in a Bayesian framework the inferential solution proposed by Eraker, Johannes & Polson (2003) for stochastic volatility models with jumps and affine structure. We will use an adaptive sampling methodology known as Delayed Rejection suggested in Tierney & Mira (1999) in a Markov Chain Monte Carlo settings in order to reduce the asymptotic variance of the estimates. Furthermore, the use of a particle filtering procedure allows to compute the Bayes factor.