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.

Two additional models were fitted to the entire dataset. Rather than representing scenarios where observation is initiated at different times, as for the sliding-window estimates, these models, like model M 0 , are post facto analyses of the epidemic. In a four parameter model, henceforth denoted with M V (cf. Table 1) , a and e are constant over time (as in model M 0 ), while the secondary transmission rate is a continuous, linearly decreasing Table 1 ) has heterogeneous time scales for the parameters, with a constant for the whole dataset and rates b and e changing by time intervals. Essentially, this approach implies: choosing a time resolution (e.g., DT = six months) for the rates b t and e t ; partitioning the whole epidemic time span into regular intervals (e.g., for DT = 6 months, four intervals: 0-6, 6-12, 12-18, and 18-24 months); fitting different b t and e t to each time interval (in the same example, four secondary rates b t and four primary rates e t ), but a single a to all the intervals.