A0313
Title: Robust score-driven filters and smoothers
Authors: Debbie Dupuis - HEC Montreal (Canada) [presenting]
Luca Trapin - University of Bologna (Italy)
Abstract: Score-driven models are quickly gaining attention as simple methods to obtain fast and accurate approximate filters and smoothers for the latent states of state-space models. A class of robust score-driven filters and smoothers able to reduce the size of the approximation error is developed. A large simulation study suggests that our robust approach can provide large efficiency gains in terms of mean squared and absolute error compared to the standard score-driven filters and smoothers in several non-linear state-space examples. Two financial applications illustrate the benefits of robustification on real data.
