Estimation of upper limits using a Poisson statistic

Bayesian, classical, and extended maximum-likelihood approaches to estimation of upper limits in experiments with small numbers of signal events are surveyed. The discussion covers only experiments whose outcomes are well described by a Poisson statistic. A new approach is proposed, based on the statistical significance of a signal rather than on the number of events in the signal region. Application of these techniques is illustrated using a toy model. (C) 2000 Elsevier Science B.V. All rights reserved.