Analytica Chimica Acta 455 (2002) 267–275 Estimating the uncertainty of binary test results to assess their compliance with regulatory limits A. Pulido ∗ , I. Ruisánchez, R. Boqué, F.X. Rius Department of Analytical and Organic Chemistry, Institute of Advanced Studies, Universitat Rovira i Virgili, Pça. Imperial Tàrraco 1, 43005 Tarragona, Spain Received 30 July 2001; received in revised form 21 November 2001; accepted 10 December 2001 Abstract A procedure to calculate the uncertainty associated to a binary response, of the type yes/no, originated from an instrumental screening system, is described. Useful equations have been developed based on statistical intervals, and their validation were performed with standard commercial solution. It was applied to a differential pulse anodic stripping voltammetry (DPASV) screening system to analyse tap water, specifically, for determining Pb 2+ and Cd 2+ above the UE legislation limits, 10 and 5 g/l, respectively. A cut-off value from the responses of the calibration samples is calculated and compared with the responses obtained for unknown samples. From this comparison, we will be able to take a compliance yes/no decision, with an alpha probability of committing a type I error. Another parameter, r screening , is calculated in order to compare different methods for the same screening application. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Uncertainty; Screening analysis; Binary response; Semi-quantitative analysis; Compliance values 1. Introduction Methods of analysis provide large amounts of quan- titative and qualitative information in measurement laboratories. Every day, the need for rapid and reli- able information increases. Although the commonest information refers to quantitative aspects, more and more analytical systems are being developed to pro- vide binary responses of the ‘yes/no’ [1] type, thus, indicating whether an analyte concentration is above or below a specified limit. These systems are related to the field of screening systems that produce qualitative or semi-quantitative information. The rapid instru- ∗ Corresponding author. Tel.: +34-977-558187; fax: +34-977-559563. E-mail address: pulido@quimica.urv.es (A. Pulido). mental responses provided by these systems are often used for taking immediate decisions: for instance, whether or not to submit a tested sample to a more elaborate quantitative analysis. As in quantitative analysis, instrumental responses from a measurement process involve random errors. These errors introduce uncertainty into the final binary response and the decision to be taken. The estimation of the uncertainty in quantitative analysis has been the goal of many studies [2–7]. However, the estimation of uncertainty for semi-quantitative results such as the ones in the present study has not been so widely studied. Moreover, uncertainty in semi-quantitative analysis cannot be expressed in the same way as in quantitative analysis [8], where it is a characteristic parameter of the dispersion of results. Uncertainty should be related to the process of taking right or 0003-2670/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII:S0003-2670(01)01604-X