Case Study of Genetic Algorithms in Metrology: Assessment of Inter-laboratory Comparisons

This study reviews conventional consensus estimation methods, including mean-based, median-based, and pooling-based approaches, and evaluates their performance under challenging scenarios involving outliers and deviations from normality. While traditional methods such as the weighted mean and weighted median often fail to handle extreme values and non-Gaussian distributions, advanced techniques like the Monte Carlo Median (MCM) and Power Moderated Mean (PMM) offer improved robustness. The Genetic Algorithm (GA), a novel optimization-based approach, demonstrates exceptional resilience to outliers. To facilitate its application, the GA is made available through the Python package consensusGen, accessible via the Python Package Index and GitHub. This ensures that practitioners and researchers can easily implement the GA in their consensus estimation tasks, benefiting from its superior robustness and precision.