Efficient Bayesian estimation for GARCH-type models via Sequential Monte Carlo
share
This paper exploits the advantages of sequential Monte Carlo (SMC) to develop parameter estimation and model selection methods for GARCH (Generalized AutoRegressive Conditional Heteroskedasticity) style models. This approach provides an alternative method for quantifying estimation uncertainty relative to classical inference. We demonstrate that even with long time series, the posterior distribution of model parameters are non-normal, highlighting the need for a Bayesian approach and an efficient posterior sampling method. Efficient approaches for both constructing the sequence of distributions in SMC, and leave-one-out cross-validation, for long time series data are also proposed. Finally, we develop an unbiased estimator of the likelihood for the Bad Environment-Good Environment model, a complex GARCH-type model, which permits exact Bayesian inference not previously available in the literature.
READ FULL TEXT