Alternative Distance Metrics for Enhanced Reliability of Spatial Regression Analysis of Health Data

Abstract

We present a spatial autoregressive model (SAR) to investigate the relationship between the incidence of heart disease and a pool of selected socio-economic factors in Calgary (Canada). Our goal is to provide decision makers with a reliable model, which can guide locational decisions to address current disease occurrence and mitigate its future occurrence and severity. To this end, the applied model rests on a quantitative definition of neighbourhood relationships in the city of Calgary. Our proposition is that such relationships, usually described by Euclidean distance, can be more effectively described by alternative distance metrics. The use of the most appropriate metric can improve the regression model by reducing the uncertainty of its estimates, ultimately providing a more reliable analytical tool for management and policy decision making.