Relative uncertainty-based Bayesian interlaboratory consensus building

Abstract

An interlaboratory comparison is typically conducted among the laboratories for the purpose of providing quality assurance and control. To solve the interlaboratory agreement problem, a distinct type of metrological challenge, a new uncertainty-based Bayesian strategy was developed and tested among environmental laboratories. A holistic algorithm with the key phases of sampling, outlier analysis, recognition, and simulation-based structure identification was developed and is being addressed in place of conventional indices and plots. Computer simulations showed that the proposed hybrid approach has no discernible sensitivity to outliers and that the agreement structure is transparent and robust. Some meta-data is also generated by the analysis based on relative uncertainty. To measure the performance and capability of Bayesian consensus building algorithm, the uncertainty intervals were established and comparative evaluations have been carried out using the conventional techniques. As a result, the suggested algorithm can explore both the laboratory performances (harmony) and the conformity between two independent samples. The algorithmic procedure features a generalizable framework that may be adapted in other fields to obtain a consensus among the laboratories.