Group Testing for Estimating Infection Rates and Probabilities of Disease Transmission

Publisher Website
Google Scholar
Abstract
Group-testing or multiple-vector-transfer designs are shown to be usually more efficient than testing individuals for estimating p, which is an infection rate or a probability of pathogen transmission by a single vector. The bias, variance, and mean-squared-error properties of these designs are explored, as some understanding of them is essential to choosing experimental designs that are efficient, convenient, and safe. For the case in which N, the number of tests (or test plants), is limiting, a method is illustrated for selecting k, the number of individuals per test (group size, vectors per test plan), to obtain a near-optimal experimental design. For the case in which N .times. k (the total number of individuals [vectors]) is limiting, alternative choices of N and k are compared. Making an appropriate choice for a particular experiment requires considering relative costs and convenience. It is important that treatment differences be judged by comparing estimates of ps, and not by comparing observed fractions of positive tests, since the latter are functions of the ks that were used as well as of the treatments; this applies even when the same value of k is used throughout.