Documentos de trabajo
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
Autor
Fúquene, Jairo
Pérez, María Eglée
Pericchi Guerra, Luis R.
Institución
Resumen
In this paper we propose a new wider class of hypergeometric heavy tailed priors that are
given as the convolution of a Student-t density for the location parameter and a Scaled Beta2
prior for the variance. These priors have heavier tails than Student-t prior, and the variances
have a sensible behavior both at the origin and at the tail, making it suitable for objective
analysis. Since the representation of our proposal is a scale mixture, it is suitable to detect
sudden changes in the model. Finally we propose a Gibbs sampler using this new family of
priors for modeling outliers and structural breaks in Bayesian dynamic linear models. It is
clearly more suitable than the almost universal use of Inverted Gamma’s for the variances 1) NIH 2) College of Natural Sciences, UPR