Autoregressive processes with generalized hyperbolic innovations

Abstract

This article investigates autoregressive processes with the flexible and attractive symmetric/asymmetric and light/heavy tailed Generalized-Hyperbolic innovations. The Generalized-Hyperbolic family of distributions has an interesting stochastic representation which can be used in simulating the proposed autoregressive model and estimating its parameters via an Expectation-Maximization (EM) type algorithm. The performance of the proposed model and its estimation through a simulation study is also evaluated. The model is then applied on two real-time series datasets. © 2019 Taylor & Francis Group, LLC.