Drew Conway continues his analysis of the Wikileaks data. Having concluded that the data appear legitimate (except perhaps in one region, based on a Benford's Law analysis of the numbers in the documents), Drew follows up with a spatio-temporal analysis of activity within Afghanistan, based on the datelines of the documents themselves (click to enlarge):
Each panel represents a year of documents; each dot is a document color-coded by the event type. Over time, you can see a steady increase in the number of reported incidents and also a change in their character, such as the perplexing number of neutral/unknown incidents in 2007. (If I were to request one change of Drew for this chart, it would be for a four-way, for-color segmentation of incidents as friendly, neutral, enemy and unknown, to make it easier to distinguish the event types.)
Wired's Danger Room column gives an in-depth look at Drew's analysis; you should take a look at is as well as Drew's original post below. Drew's R code for all his analysis is also available at this github repository.
Zero Intelligence Agents: Wikileaks Attack Data by Year and Type Projected on Afghanistan Regional Map
Comments