Biased reduced sampling: Detectability of an attribute and estimation of prevalence

In surveilling a population, detection of systems with all attribute of interest and estimation of the prevalence of the attribute in the population are two main goals. Due to cost constraints, only a subset of all components of sampled systems may be fully tested. Biasing the sampling to increase the probability of choosing a component with all attribute of interest ameliorates the impact of reduced sampling. In this paper, we consider the impact of biased reduced sampling oil detection and propose an approach for estimating the prevalence of the attribute in the population which properly accounts for the biasing. The proposed method is illustrated with a simulated example. Copyright (C) 2005 John Wiley & Sons, Ltd.