The available surveillance datawas used as a representative base for
fitting a distribution to infer the variation for the whole population
(all meat available for sale). Here we chose to use the MLE for estimation
of pest, μ and σ by fitting a zero-inflated lognormal distribution.
Zero-inflation is needed because many batches of broiler meat are not
found to be contaminated. The lognormal distribution is frequently
used to describe the variation of concentrations of foodborne pathogens
(Kilsby and Pugh, 1981; Crépet et al., 2007; Busschaert et al., 2010).