| Summary: | The writers suggest a new Bayesian Markov chain Monte Carlo (MCMC) methodology to estimate a wide class of multidimensional jump-diffusion models. They base their approach on the closed-form (CF) likelihood approximations of Ait-Sahalia. They report that the CF likelihood approximation does not integrate to 1, being very close to 1 when in the center of the distribution but potentially differing markedly from 1 when far in the tails. Proposing an MCMC algorithm that addresses the problems that arise when the CF approximation is applied in a Bayesian context, they demonstrate the efficacy of their approach in a simulation study of the Cox-Ingersoll-Ross and Heston models and apply it to two well-known datasets.
|
|---|