Modelling outliers and structural breaks in dynamic linear models with a novel use of a heavy tailed prior for the variances: An alternative to the Inverted Gamma

TL;DR

This paper introduces a new heavy-tailed prior for variances in dynamic linear models to better handle outliers and structural breaks, offering an alternative to traditional Inverted Gamma priors.

Contribution

It proposes a novel heavy-tailed prior for variances that improves robustness in dynamic linear models compared to standard approaches.

Findings

Enhanced robustness to outliers and structural breaks

Better model fit with the new prior

Reduced sensitivity to prior assumptions

Abstract

Modelling outliers and structural breaks in dynamic linear models with a novel use of a heavy tailed prior for the variances: An alternative to the Inverted Gamma