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.

We remarked in the results that support for the exponential and Cauchy model was found in post hoc analyses of the data. Model comparison from early snapshots supported in general spatially structured models with external infection, but could not select a dispersal kernel (most of the kernels tried, see e.g. Text S1, performed equally well). The choice of an exponential kernel for early estimations (Figure 2 ) would then be motivated by a strong prior belief on disease dispersal (for example, from results in the absence of the leaf miner in [53] ). Here, we also note that, in our case, the absence of such a prior belief would be of little importance, as the exact form of the kernel would not affect the main results of Figure 2 . Of the several kernels tried for the first few snapshots, most (e.g. the Gaussian, Text S1) produced estimates of dispersal scale and infection rates with patterns in time qualitatively very similar to those in Figure 2 (results not shown here).