Bayesian inference for stochastic volatility models using MCMC methods highlydepends on actual parameter values in terms of sampling efficiency. While drawsfrom the posterior utilizing the standard centered parameterization break downwhen the volatility of volatility parameter in the latent state equation issmall, non-centered versions of the model show deficiencies for highlypersistent latent variable series. The novel approach ofancillarity-sufficiency interweaving has recently been shown to aid inovercoming these issues for a broad class of multilevel models. In this paper,we demonstrate how such an interweaving strategy can be applied to stochasticvolatility models in order to greatly improve sampling efficiency for allparameters and throughout the entire parameter range. Moreover, this method of"combining best of different worlds" allows for inference for parameterconstellations that have previously been infeasible to estimate without theneed to select a particular parameterization beforehand.
展开▼
