Work

A novel random environment INAR(1) model with generalized signed thinning for regime-switching integer-valued time series

Year: 2026
Type: preprint
Abstract: Count time series in criminology, epidemiology and economics often exhibit regime-switching behavior and may take negative values after temporal or cross-series differencing. This paper proposes a novel random environment INAR model with generalized signed thinning operator, in which the innovation term is governed by a finite-state Markov chain. The proposed model accommodates \(\mathbb{Z}\)-valued observations and captures regime-switching dynamics induced by unobserved environmental changes. We establish the Markovian property and ergodicity of the bivariate process and derive its conditional moments and autocovariance structure. Based on the ergodicity of the bivariate Markov chain, the strong consistency and asymptotic normality of the Yule-Walker estimators are established by applying the additive functional ergodic theorem, without requiring a closed-form marginal distribution. A Monte Carlo simulation study is conducted to assess the finite-sample performance of the estimators. An empirical application to the monthly motor vehicle theft differences between two police beats in Pittsburgh confirms the model's ability to capture regime-dependent dynamics and handle negative-valued count series. \noindent\textbf{Keywords:} Regime switching; Random environment; Generalized sign thinning operator; Integer-valued time series; Strong consistency; Asymptotic normality \noindent\textbf{Mathematics Subject Classification:} Primary 62M10; Secondary 60J10.
Language: English
Open Access status: green
OpenAlex ID: W7236470329
Cites: 21
Cited by:
Related to: 10