Testing stability in a spatial unilateral autoregressive model

TL;DR

This paper investigates the asymptotic behavior of the least squares estimator for the stability parameter in a spatial unilateral autoregressive model, revealing normality in the unstable case with a specific scaling, contrasting with classical AR models.

Contribution

It demonstrates that the least squares estimator of the stability parameter in a spatial unilateral autoregressive process is asymptotically normal in the unstable case with a scaling of n^{5/4}, unlike traditional AR models.

Findings

Asymptotic normality of the estimator in the unstable case

Scaling factor n^{5/4} for the asymptotic distribution

Contrast with classical AR(p) models' behavior

Abstract

Least squares estimator of the stability parameter $\varrho := |\alpha| + |\beta|$ for a spatial unilateral autoregressive process $X_{k,\ell}=\alpha X_{k-1,\ell}+\beta X_{k,\ell-1}+\varepsilon_{k,\ell}$ is investigated. Asymptotic normality with a scaling factor $n^{5/4}$ is shown in the unstable case, i.e., when $\varrho = 1$, in contrast to the AR(p) model $X_k=\alpha_1 X_{k-1}+... +\alpha_p X_{k-p}+ \varepsilon_k$, where the least squares estimator of the stability parameter $\varrho :=\alpha_1 + ... + \alpha_p$ is not asymptotically normal in the unstable, i.e., in the unit root case.