Partially observed branching processes for stochastic epidemics.

At the offset of a (stochastic) epidemic, it is of importance to have a mathematical model that will assist in the making of an informed judgement on whether the epidemic will explode, or will be minor and die out. In this paper, we consider probabilistic inferences related to the event of extinction of a discrete time branching process when this cannot be directly observed. Instead, we are able to observe only a random "trace" of the process, which not only trails the latter, but also directly affects it (in terms of interventions). A simple model is proposed that provides tractability, preserves a marginal branching property, and gives reasonable closed form expressions.