Inferring pandemic growth rates from sequence data

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

In order to assess how inferred BSPs are affected by the way in which viral sequences are collected from infected individuals, we examine two sampling schemes: uniform and log-proportional. Under uniform sampling, we randomly pick one sample per generation of simulated sequences and use these selected sequences to infer posteriors and BSPs. This would be analogous to choosing to sequence a virus from a randomly selected infected person at regular temporal intervals, perhaps set by laboratory capacity. Under log-proportional sampling, we randomly extract samples in proportion to the logarithm of the number of samples available per generation. This is closer to how viral samples are collected during an epidemic where more samples are collected if more individuals are infected, but sampling proportional to the actual number of cases is not possible for logistical reasons.