Fig. 3.

Cases 3 and 4. a) Unexplained relative risk as a function of the relative susceptibility of the high- to low-risk individuals ($\mathrm{a}$). The different lines correspond to different levels of assortative mixing. Median and 95% CIs based on 500 simulations. b) 95% CIs for the effect of ethnicity $\exp ({\beta }_e)$ from the fitted Poisson regression model for varying levels of assortative mixing. The high-risk individuals are defined by a larger susceptibility than the low-risk individuals (case 3). There are 100 simulations for each assortativity level. c) Same as b), except the high-risk individuals are defined by a larger total number of contacts (case 4)