Inferring pandemic growth rates from sequence data

de Silva, Eric; Ferguson, Neil M.; Fraser, Christophe

In general, lower values of k correspond to more heterogeneity in infectiousness, and thus epidemics are increasingly characterized by superspreaders. For k ¼ 1, the distribution is geometric, while it becomes Poisson as k ! 1. We assume that the number of offspring each sequence generates is independent of and distributed identically to the offspring generated by any other sequence. Implicitly this is equivalent to assuming random (although not homogeneous) mixing and an effectively infinite susceptible population. For many parameter combinations (especially low k and/or low R), there is a high probability of early extinction of a simulated outbreak. Because we are interested in emerging epidemics, not self-limited outbreaks, we select our simulated datasets from outbreaks which do not go extinct.