Geographical Analysis Earth-Surface Processes

When Hotspots Disagree: Reconciling Cluster Detection and Relative Risk Surface Estimation in Applied Spatial Analysis

Tilman M. Davies
Wiley · 2026-10-25 · DOI 10.1111/gean.70073

Abstract

ABSTRACT Hotspot analysis is central to many point‐referenced geospatial applications, but the term “hotspot” is often used to describe different kinds of evidence: a statistically significant cluster, a region of high estimated intensity, or an elevated relative risk surface. As a result, outputs produced by different methods are often compared within a common visual and practical language, even when they support different spatial claims. We examine this problem through a focused comparison of scan‐statistic cluster detection and kernel‐smoothed relative risk estimation for case‐control point patterns. We develop an agreement‐profile framework that separates four dimensions of comparison: Detection state, Location, Scale, and Morphology. Across repeated simulations and three real‐data examples, we show that similar overall overlap can conceal different containment, extent, and component relationships; that agreement between fitted regions is distinct from recovery of a prespecified spatial target; and that interpretation can change under alternative analytical specifications and reference measures. Rather than ranking methods, it provides a reproducible vocabulary for interpreting and communicating how hotspot outputs relate across analyses.

Read at publisher →