Decision Theoretical Analysis of Random Sampling Procedures Past- Present- Future

Stimulated by the IAEA's renewed interest in advanced technologies and methodologies for verifying declared nuclear material and activities, and even more so by the results of the last Non-Proliferation Treaty Review Conference this contribution draws the attention to methodologies for evaluating random sampling procedures which have already been developed in the 80s of the last century, but which have been used in practice only scarcely since that time. In fact, the IAEA Safeguards Manual on Statistical Concepts and Techniques presents a few applications of these methodologies which demonstrate at least their potential. The so-called IAEA formula for Stratified Attribute Sampling gives an heuristic advice how to distribute a given inspection effort on the verification of items of nuclear material collected in different strata, as it occurs, e.g., in the course of the verification of an inventory of an nuclear plant. With the help of a decision theoretical analysis, which has been refined since the 80s it is shown under which assumptions the IAEA formula can be justified, which technical parameters have to be known for the optimization of strata sample sizes and, most importantly, what is the efficiency and effectiveness of the safeguards measures. Practical examples are presented which give numerical evidence of these findings. Furthermore, the question of the inspection effort necessary for inducing the plant operator to legal behavior – in the sense of the verification agreement – is discussed. It is shown that for this purpose at least conceptionally subjective parameters – in contrast to the technical ones used so far – have to be introduced which admittedly is a controversial issue.