Uncertainties in Nuclear Proliferation Modeling

There have been various efforts in the research community to understand the determinants of nuclear proliferation and develop quantitative tools to predict nuclear proliferation events. Such systematic approaches have shown the possibility to provide warning for the international community to prevent nuclear proliferation activities. However, there are still large debates for the robustness of the actual effect of determinants and projection results. Some studies have shown that several factors can cause uncertainties in previous quantitative nuclear proliferation modeling works. This paper analyzes the uncertainties in the past approaches and suggests future works in the view of proliferation history, analysis methods, and variable selection. The research community still lacks the knowledge for the source of uncertainty in current models. Fundamental problems in modeling will remain even other advanced modeling method is developed. Before starting to develop fancy model based on the time dependent proliferation determinants' hypothesis, using graph theory, etc., it is important to analyze the uncertainty of current model to solve the fundamental problems of nuclear proliferation modeling. The uncertainty from different proliferation history coding is small. Serious problems are from limited analysis methods and correlation among the variables. Problems in regression analysis and survival analysis cause huge uncertainties when using the same dataset, which decreases the robustness of the result. Inaccurate variables for nuclear proliferation also increase the uncertainty. To overcome these problems, further quantitative research should focus on analyzing the knowledge suggested on the qualitative nuclear proliferation studies