Protocol for uncertainty assessment of half-lives

The apparent tendency to underestimate the uncertainty of experimentally determined half-life values of radionuclides is discussed. It is argued that the uncertainty derived from a least-squares analysis of a decay curve is prone to error. As it is quite common for a series of activity measurement results to be autocorrelated, the prerequisite of randomness of data for common statistical tests to apply is not fulfilled. In this work, an alternative data analysis method is applied that leads to a more realistic uncertainty budget. The uncertainty components are being subdivided in three categories according to the relative frequency at which they occur, an appropriate uncertainty propagation formula applied and then the total uncertainty obtained from an independent sum. An attempt is made to apply the protocol to problematic cases in literature, yet it is clear that the reporting is usually incomplete for a full uncertainty analysis. Suggestions are made for a concise but more complete reporting style, for the sake of traceability.