Probability of Attempted Armed Theft of Nuclear Weapons: Application of the Method of Gott

The Bayesian methodology pioneered by Gott is applied to an event that could have occurred but has not (e.g., attempted armed theft of a nuclear weapon). Several different possible classes of choices for the required prior distribution are discussed, and details of the resulting posterior distribution are presented for the case of a Jeffreys prior (as was used by Gott). Sample results include, for example, the prediction that there is a 95% probability such an initial attempt will occur after mid-2014, and likewise before 1280 years from the present time. The posterior pdf is monotonically decreasing, so that the most likely time of occurrence is now, but “heavy-tailed” in the sense that the associated expected time of occurrence is infinite. These results are intuitively somewhat dissonant, but the latter accords well with the zero frequency of occurrence that results from applying the maximum-likelihood frequentist approac