This repository is in support of a talk on geospatial data.

To recreate all of the examples, the data are available here: https://s3.amazonaws.com/gardner-datasets/PyDataBerlin/data.zip (it is about 2GB)

The exception is the Demographics and Health Survey (DHS) as this requires registering to download and use. Downloading the data into the data folder is necessary to recreate the examples. The extract_dhs script will format the DHS data into the form used in the examples. If the DHS data are not available, it will create synthetic random data that are structually similar, but the will not reveal the same relationships between cluster median asset and nighttime light intensity.

There are two examples in notebooks. The first, Nigeria Nightlights is what is discussed in the talk. The second US Nightlights performs similar analysis on U.S. data.