Computation of confidence intervals for Poisson processes (Q1581156)

Many experiments in physics have a Poisson distributed outcome variable, such as the number of distinct events in high energy collisions. The statistical analysis seems to become problematic when the experiment reaches its limits, especially when the number of background events exceeds the number of objective events. Mostly after finishing an experiment one has to decide whether to construct for the parameter of the Poisson distribution two-sided or a one-sided upper confidence limits. This decision may cause a bias. \textit{G. J. Feldman} and \textit{R. D. Cousins} [Phys. Rev. D 57, 3873 (1998)] proposed a way to construct confidence limits that are unbiased with regard to this decision aspect. The paper describes an efficient way to construct the correct confidence limits. After a general description of the algorithm in section two a detailed listing how to implement and to realize the algorithm is given. Finally an example shows how the procedure works and with which precision. In the appendix the corresponding FORTAN statements are given.