Consistency of a reaction dataset

The framework of Data Collaboration supports a rigorous numerical approach to dataset consistency, which provides a combined way to look at system uncertainties originating from either rate parameters or experimental observations [70]. The measure of dataset consistency 

The usefulness of this measure can be illustrated with a simpler, easier to visualize test. We consider pairs of dataset units (Ug, Uy) for e,f (not necessarily distinct) in the index set of the dataset D. For each pair we compute the minimum level of uncertainty, Ug/, that leaves the two-element dataset Dgf = Ug, Uj consistent, i.e., 

The analysis of dataset consistency is assisted by Lagrange multipliers, which gauge sensitivity of the consistency measure to the dataset uncertainties. The example is presented in Fig. 9. The two largest peaks are the sensitivities of the consistency measure with respect to the lower bound of the experimental uncertainty of dataset unit 57, and the upper bound Msg of the experimental uncertainty of dataset unit 58 (bottom left and right panels of Fig. 9, respectively). These same two dataset units surfaced in the pairwise test. Both facts gave us reason to suspect C/57 and/or f/sg as possible outliers. On reexamination of the original experimental data by the researchers involved (see Ref. [70]), errors in data recording were found and the corrections made were exactly in line with what the consistency analysis had indicated was probable. 

---