Exploring Heterogeneity in Health Outcomes and Geographic Space: Integrating Quantile and Geographically Weighted Regression

Tse-Chuan Yang, Pennsylvania State University
Stephen A. Matthews, Pennsylvania State University
Vivian Chen, Tamkang University

The past decade has experienced growth in the investigation of heterogeneous associations with independent variables across the distribution of either the dependent variable or across geographic space. The former is implemented using quantile regression (QR), whereas the latter has been popularized by geographically weighted regression (GWR). However, demographers have lagged other fields in adopting either of these methods. In this paper, we fuse QR and GWR to create an innovative approach to simultaneously explore the heterogeneity embedded in both variables and space – an approach we name geographically weighted quantile regression (GWQR). The goal of this study is to introduce GWQR to demographers. We illustrate GWQR in two studies, the first using individual-level data (obesity in Philadelphia) and the second ecological data (US county-level mortality). Significant heterogeneity across space and the distributions of body mass index and mortality indicates that heterogeneity commonly exists and should be considered in model specification.