Getting the Bigger Picture Through Spatial Modeling of DHS Data

A tweet today from Springer Publishing noted that the latest issue of Spatial Demography is "on the newsstands" and I am co-author on a paper in this issue. I wouldn't necessarily toot my horn on something like this, but the methods we used in the paper presaged a new effort by the DHS Spatial Data Repository team to create spatial models of data from the DHS surveys. Our work was led by one of our PhD students, Stephen Crook, and utilized Empirical Bayesian Kriging, the same method employed by the DHS folks. We focused on a four-region area of southern Ghana and examined the spatial patterns of obesity in that country. DHS data have consistently shown an increase in obesity over time in Ghana (and many other African countries), particularly in the urban areas where there is increasing access to Western-style processed foods. Here is what our map looks like, using data from the 2008 DHS (the 2014 data weren't out yet when we started working on this):

The hot spots are especially Accra (the capital and largest city in the country) in the south, and Kumasi to the northwest of Accra. Kumasi is the nation's second largest city. In particular, the data suggest that the suburban and peri-urban areas of Accra--where the wealthiest residents live--are especially prone to obesity. If you just analyze the DHS data without reference to the spatial patterning, you might get hints of what's going on, but the maps genuinely enhance our understanding.