Geography of obesity: Scales and analytics

There is abundant literature on the geography of obesity that discusses its spatial patterns and how it is related to socio-economic and environmental factors. With a few exceptions, most of such studies use conventional regression modelling with census tracts or counties as the units of analysis. However, this demonstration suggests that the investigation of geographic relationships could benefit from using geographically weighted regression (GWR) models and, whenever possible, the units of analysis should be refined to census block groups or even census blocks. First, the use of GWR is proposed because such models provide statistical results that allow for the analysis and mapping of how regressed relationships vary spatially. Furthermore, large variations in socio-economic and environmental attributes within a county or a census tract can result in over-simplified conclusions due to the relatively low geographic resolutions of such geographies. To demonstrate, this analysis modeled overweight/obesity among residents of Summit County, Ohio, using the heights and weights of over 400,000 records obtained from the Ohio Bureau of Motor Vehicles. Using GWR, the locations of fitness centers and non-fresh food outlets were regressed against the spatial distribution of the overweight/obese population in the county. The model demonstrates the difference in geographic resolution between using census tracts and census block groups as the units of analysis. The demonstration shows the over-simplification that can result from the use of larger geographic units and highlights the critical issue of spatial variation in the investigation of public health disparities, especially as they apply to geography.