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.

The robustness of sliding-window estimates for a to different estimation periods motivates the following assumption: environmental fluctuations affect the model only through primary and secondary infection rates, while the short-range dispersal scale a remains constant at each census site all along the epidemic. We integrated this assumption into our estimations, and fitted to the entire dataset model M DT a , with heterogeneous time scales for the parameters (cf. model Table 1 and Methods), where a was kept constant for the whole epidemic history, while the rates b t and e t changed with frequency DT. All the analyses from now on concern model M DT a , and focus on two different time intervals for the infection rates, obeying two different purposes. The first, DT = 6 months, is intended to capture the main temporal trend in rates; the second, DT = 1 month (corresponding to the highest possible resolution given data censoring), is used to analyse short-time fluctuations.