A note on reproducible GIS by economists: it’s mostly absent. Here are some tips.
Reproducibility and GIS
Reproducibility means that the inputs and methods can be repeated by a (somewhat) knowledgeable person. For GIS, that might (should) mean code, but it definitely means at least SOME instructions. Even if they are manual….
Maps are data. While your typical Stata/Matlab/R/Julia graph is data projected into Cartesian coordinates, maps are data projected into geographic coordinates. So at a minimum, we need to know what the inputs to the map are, same as we need to know inputs to graphs.
So “data + code” + “
graph twoway scatter x y” -> 📈📉,
and “data + code” + “
maptile x, geo(state)” -> 🗺️. Or “data + code” + “instructions(ArcGIS)” -> 🗺️.
Note that for a map, “data” includes shapefiles (including provenance of the shapefile)
Checklist for reproducible maps
- ✅ data (+ provenance)
- ✅ code (manipulates data)
- ✅ shapefile (provides coordinates)
- ✅ code (preferred) or instructions (sufficient)
A few tentative resources are collected at https://social-science-data-editors.github.io/guidance/guidance-reproducible-gis-analysis.html (please suggest improvements!), with some other links below:
GeoPandas and others
ArcGIS can also be scripted (via python) and now integrates Jupyter-like notebooks
Please try to create scripted maps, but always describe what data you are mapping, and where you got the shapefiles from (note: copyrights might apply, permissions might need to be obtained!)
And for those exceptions from scripting: a 2-3 sentence description of what you did would be sufficient. And for the unscripted ArcGIS which you did before you learned how to script stuff: a 4-5 sentence description may be sufficient as well.
Follow the conversation
Slightly modified from https://twitter.com/AeaData/status/1359516297990701057