Estimating Consignment-Level Infestation Rates from the Proportion of Consignment that Failed Border Inspection: Possibilities and Limitations in the Presence of Overdispersed Data.

Introduction of pests and diseases through trade is one of the main socioecological challenges worldwide. Targeted sampling at border security can efficiently provide information about biosecurity threats and also reduce pest entry risk. Prioritizing sampling effort requires knowing which pathways are most infested. However, border security inspection data are often right-censored, as inspection agencies often only report that a consignment has failed inspection (i.e., there was at least one unit infested), not how many infested units were found. A method has been proposed to estimate the mean infestation rate of a pathway from such right-censored data (Chen et al.). Using simulations and case study data from imported germplasm consignments inspected at the border, we show that the proposed method results in negatively biased estimates of the pathway infestation rate when the inspection data exhibit overdispersion (i.e., varying infestation rates among different consignments of the same pathway). The case study data also show that overdispersion is prevalent in real data sets. We demonstrate that the method proposed by Chen et al. recovers the median infestation rate of the pathway, rather than its mean. Therefore, it underpredicts the infestation rate when the data are overdispersed (in right-skewed distributions, the mean is above the median). To allow better monitoring and optimizing sampling effort at the border, we recommend that border protection agencies report all the data (the number of infested units found together with the sample size of the inspection) instead of only that the consignment failed inspection.