EFFECTS OF AGGREGATION ON THE PROBABILITY OF DETECTING INFESTATIONS IN FRESH PRODUCE CONSIGNMENTS

Border inspections of imported fresh produce aim to detect pest or disease infestations of 0.005 prevalence with 95 % confidence and, based on the binomial distribution, a sample of 600 produce units is typically inspected from each consignment. Depending on the type of produce, units are selected at either the individual or the carton level to make up the sample size. However, sampling cartons (cluster sampling) violates the assumption of the binomial distribution that the sampling units are independent of one another. This may be of no consequence if the pests or diseases are distributed randomly throughout the consignment, but aggregation of invertebrates and disease within a consignment is likely to occur. Simulation models are used to demonstrate how cluster sampling leads to a decreased probability of detecting infestations and an underestimate of the true level of infestation with increasing levels of aggregation