"The Great Blackbury Pie" ~ or ~ Focal Area Bias in Geographically Weighted Analysis

Focal and geographically weighted analyses are commonplace in GIS applications across many fields and disciplines. However, where such analyses are based on ‘dense’ datasets (e.g., a raster surface), they can suffer from an unintended bias towards the periphery of the focal zone (neighbourhood), which (counterintuitively) is exacerbated by the use of distance weighting functions. This paper serves to characterise this problem, which we call focal area bias (FAB), present a proposed correction, and point to extensive simulation-based analysis, which demonstrates both the impact that this effect can have on analyses and the efficacy of our proposed solution.