Figure 4
From: Finding disease outbreak locations from human mobility data
Estimation of the number of outbreak origins from the score distribution. A The scores \(S_{i}\) of the 10 highest-scoring locations, for a varying number of outbreak origins M. We find that the top M locations corresponding to the true outbreak origins have a distinctly high score (hollow symbols), at around the expected values of \(E(S_{i})=1/M\) (dotted lines). The scores of further locations follow an exponentially decaying distribution. Error bars indicate the standard deviation. B Based on the characteristic form of the score distribution, we define an estimation heuristic for the number of outbreak locations, \(P(\widehat{M}=M)\) (see main text). The heuristic correctly identifies the true number of outbreak sources M, where the accuracy decreases with rising number of outbreak origins and tends to overestimate the number of outbreak origins