Figure 3

Accuracy of the inference method for multiple outbreak origins. A For multiple origins (here \(M=2\)), the location accuracy \(P_{\mathrm{loc}}\) increases with sample sizes \(N_{1}\) and \(N_{2}\), especially if a similar number of samples is available for all origins (\(N_{1} \approx N_{2}\), dotted line). However, if more trajectories stem from one origin instead of the other (\(N_{1} \neq N_{2}\), off-diagonal entries), the accuracy actually decreases as the overrepresented location masks the other locations. B The accuracy of the location prediction is approximately independent of the number of outbreak locations M, while the time accuracy decreases. Sample size is \(N_{i}=4\) for all origins i. Results are shown for the dEPR dataset