Time inhomogeneous multivariate Markov chains: Detecting and testing multiple structural breaks occurring at unknown dates

Recommended citation: Damásio, B., & Nicolau, J. (2024). Time inhomogeneous multivariate Markov chains: Detecting and testing multiple structural breaks occurring at unknown dates. Chaos, Solitons & Fractals, 180, 114478.https://doi.org/10.1016/j.chaos.2024.114478

Abstract: Markov chain models are used in several applications and different areas of study. A Markov chain model is usually assumed to be homogeneous in the sense that the transition probabilities are time-invariant. Yet, ignoring the inhomogeneous nature of a stochastic process by disregarding the presence of structural breaks can lead to misleading conclusions. Several methodologies are currently proposed for detecting structural breaks in a Markov chain. However, these methods have some limitations: namely they can only test directly for the presence of a single structural break. This paper proposes a new methodology for detecting and testing the presence of multiple structural breaks in a Markov chain occurring at unknown dates.