Flexible model comparison of unobserved components models using particle Gibbs with ancestor sampling

- Unobserved components models with stochastic volatility (SV) effects are widely used to model inflation rates.
- However, formal model comparison using Gibbs sampling is difficult.
- We show that PG-AS provides a flexible framework for estimation and model comparison.
- We provide applications using US and UK data, comparing different models.
- The model with time-varying SV in mean effects performs best.

In this paper, we show that particle Gibbs with ancestor sampling (PG-AS) provides a unified and flexible framework for estimation and model comparison of unobserved components models with time-varying volatility effects, which are widely used in inflation rate modeling.