Non-Gaussian Bridge Sampling with an Application

摘要: This paper provides a new bridge sampler that can efficiently generate sample paths, subject to some endpoint condition, for non-Gaussian dynamic models. uses companion pseudo-Gaussian as the proposal and sequentially re-simulates paths via sequence of tempered importance weights in way bearing resemblance density-tempered sequential Monte Carlo method used Bayesian statistics literature. is further accelerated by employing novel idea k-fold duplicating base set followed support boosting. We implement this on GARCH model estimated S&P 500 index series, our implementation covers both parametric non-parametric conditional distributions. Our performance study reveals far superior either simple-rejection when it applicable or other alternative samplers designed with fixed endpoint. Computing SRISK NYU-Stern Volatility Institute then demonstrate method's real-life applicability.