Improved statistical methods for estimating infestation rates in quarantine research when hosts are naturally infested

Abstract

Trading partners often require phytosanitary or quarantine treatments for fresh horticultural produce to ensure no economically important pest species are moved with the imported product. When developing such treatments, it is essential that the level of treatment efficacy can be determined. This is often based on the mortality of the total number of target pests exposed to treatment, but in naturally infested products this number is not always known. In such cases, the infestation rate and subsequently an estimate of the number of pests are obtained directly from a set of untreated control samples of the host product. The International Plant Protection Convention (IPPC) Secretariat has provided 2 formulas for these situations that place an interval around the point estimate obtained from the control samples to obtain an estimate of the infestation rate. However, these formulas do not allow a confidence level to be assigned to the estimate, and there are concerns with the assumptions regarding the distribution and the measure of variability used in the formulas. In this article, we propose 2 alternative formulas. We propose that the lower one-sided confidence limit should be applied to all infestation datasets that are approximately normally distributed. As infestation data are sometimes skewed, it is proposed the lower one-sided modified Cox confidence limit is applied to data approximately log-normal distributed. These well-recognized formulas are compared to the formulas recommended by the IPPC and applied to 3 datasets involving natural infestation.