Probability Bounds Analysis Solves the Problem of Incomplete Specification in Probabilistic Risk and Safety Assessments

The unavailability of empirical information that is simultaneously relevant and complete enough to fully specify the analysis is a ubiquitous problem in probabilistic risk and safety assessments (PRAs and PSAs). There is commonly uncertainty about the distributions of input variables and their mutual dependencies. In some cases, even details about the mathematical model that should be used to combine the input variables may be in dispute. However, to obtain any answer at all from a traditional probabilistic assessment such as a Monte Carlo simulation, the assessment model, all the distribution shapes, parameters and dependence structures, and correlation coefficients must be fully specified at the start. Therefore, an analyst generally needs to make many assumptions that go considerably beyond what can be justified empirically. Naturally, these assumptions of convenience diminish the credibility of the final conclusions derived from the assessments. Probability bounds analysis allows assessors to sidestep uncertainty about the precise specifications of input variables, imperfect information about the correlation and dependency structure among the variables, and even dispute about the form of the risk model itself. The package RAMAS Risk Calc implements probability bounds analysis under the Windows operating system. The convenient interface offers a powerful array of standard functions, which are extensible by means of user-written programs. The uses of probability bounds analysis and the features of the software are described.