Fig. 5

Robustness of network-based targeting. Shown is the final epidemic size (the percentage of individuals ever infected) for varying initial proportions of individuals holding pro-vaccine \(\upsilon ^{+}\) and anti-vaccine \(\upsilon ^{-}\), with the targeting method \(\mathcal {O}_s\) set to high-degree in the top panel and random in the bottom. The insets show the case of \(\upsilon ^{+} = \upsilon ^{-}\). Degree-based targeting provides more effective interventions than random placement. Each boxplot is based on 100 independent simulations