Bayesian Analysis for Inference of an Emerging Epidemic: Citrus Canker in Urban Landscapes

Neri, Franco M.; Cook, Alex R.; Gibson, Gavin J.; Gottwald, Tim R.; Gilligan, Christopher A.

There were clear trends in both the expectation and the precision of estimates for the secondary transmission rate, b. As in the case of a, the posterior distribution for b had a large variance when based upon data for the first three months, and adding extra monthly snapshots decreased the variance of the posterior (cf Figures 2B,E) . In contrast with the case of a, there was also a trend in the posteriors for b to decrease as time progressed. The trend in b is more appropriately characterised by the sliding windows ( Figure 2E ), in which estimates are averaged over successive but overlapping six 30-day intervals (cf. model M DT slid in Table 1 , with DT = 6 months). Similar results were obtained for e ( Figures 2C,F) , suggesting that both forms of transmission were driven by environmental variables. Epidemics were dominated by secondary over primary infection: the forces of infection corresponding to b were much greater than those for e. Hence, in the following we will focus our analysis of environmental trends on the time dependence of b.