Testing for structural change in spatial regions at unknown positions

Abstract

We propose a fluctuation-type procedure for detecting breaks in spatial regions. While such tests are common in the context of time series, it is not a priori clear how to apply them to spatial data as there is no natural order of the observations. We demonstrate how this order can be constructed from a spatial autoregressive model. Once such an order is derived, standard time series results apply and break points can be consistently identified.