King's College London

In this chapter we discuss the use of Bayesian nonparametric methods for time series analysis. First developed by [Freguson (1973)] these methods focus on how a stochastic process can be used as a prior over probability measures as well as a prior on the underlining mixing measure in a mixture model. The empirical examples of the chapter centre on financial and macroeconomic time series, and demonstrate that volatility, long memory and vector autoregressive models underpinned by Bayesian nonparametric methods have superior out-of-sample predictive performance compared to other competitive models.