New framework simplifies predictive density evaluation using PIT
Researchers from Banco de España propose a simplified framework for evaluating conditional predictive densities. The approach, based on the probability integral transform (PIT), accommodates a wide range of estimation schemes and applies to both stationary and non-stationary processes.
A flexible approach to density evaluation
This paper introduces a novel framework for evaluating conditional predictive densities, building on the probability integral transform (PIT).
It abstracts from parameter estimation uncertainty, accommodating both rolling and expanding window estimation schemes.
This flexibility extends to stationary and non-stationary processes, broadening the applicability of existing tests.
The framework also incorporates weighted test statistics, allowing researchers to focus on specific parts of the predictive distribution, such as the left tail, without disregarding other regions.
Stochastic volatility for US industrial output
Through Monte Carlo simulations, the authors demonstrate that the proposed tests exhibit favorable size and power properties under their new assumptions.
In an empirical application, the framework is used to forecast US industrial production growth.
The findings reveal that incorporating stochastic volatility into an unobserved components model is essential for generating correctly calibrated density forecasts of US industrial production growth, at both monthly and quarterly frequencies.