Ancillarity-sufficiency interweaving strategy (ASIS) for boosting MCMC estimation of stochastic volatility models

Abstract

Bayesian inference for stochastic volatility models using MCMC methods highly depends on actual parameter values in terms of sampling efficiency. While draws from the posterior utilizing the standard centered parameterization break down when the volatility of volatility parameter in the latent state equation is small, non-centered versions of the model show deficiencies for highly persistent latent variable series. The novel approach of ancillarity-sufficiency interweaving has recently been shown to aid in overcoming these issues for a broad class of multilevel models. In this paper, we demonstrate how such an interweaving strategy can be applied to stochastic volatility models in order to greatly improve sampling efficiency for all parameters and throughout the entire parameter range. Moreover, this method of "combining best of different worlds" allows for inference for parameter constellations that have previously been infeasible to estimate without the need to select a particular parameterization beforehand.

Does joint modelling of the world economy pay off? Evaluating global forecasts from a Bayesian GVAR

“…In short, this implies starting with the full model (i. e., all elements of ψ i are set equal to one) and then using a simple birthdeath sampler to explore the posterior distribution (Madigan and York, 1995). The history of log-volatilities and the corresponding parameters are sampled using an ancillarity sufficiency interweaving strategy proposed in Kastner and Frühwirth-Schnatter (2014).…”

Measuring International Uncertainty: The Case of Korea

“…In the first step we obtain forecast errors by estimating the model in Equation 1 via OLS estimation. Then, we run an MCMC algorithm (Kastner and Fruhwirth-Schnatter, 2014) to generate posterior draws for (h j,1:T , c…”

Trend Fundamentals and Exchange Rate Dynamics

“…Following Kastner and Frühwirth-Schnatter (2014) we use a non-conjugate Gamma prior on the variance of the log-volatility,…”