Feilkvantifisering

Feilkvantifisering (Norwegian for "error quantification") is a framework in measurement science and data analysis for characterizing and communicating the size and impact of errors in measurements and model outputs. It encompasses both systematic deviations and random fluctuations.

The approach involves constructing an error model, identifying error sources (instrumental, environmental, methodological), quantifying their magnitudes, and combining them to express total uncertainty. It draws on principles from metrology, statistics, and uncertainty quantification to provide a transparent assessment of confidence in results.

Common methods include uncertainty propagation using analytical formulas such as first-order Taylor expansions, Monte Carlo simulations to sample error distributions, Bayesian inference for probabilistic modeling of uncertain quantities, bootstrapping to estimate sampling variability, and calibration against reference standards. These techniques help quantify how errors in inputs affect final measurements or predictions and how different sources interact.

Applications span experimental science, engineering, quality control, climate and environmental modelling, and data analysis in software and machine learning workflows where assessments of predictive risk and reliability are important.

Challenges include the need for reliable knowledge of error sources and their distributions, potential correlations between errors, model misspecification, data limitations, and the requirement for transparent reporting of assumptions and methods. Computational cost can be significant in complex models, and adherence to standards may be necessary in regulated contexts.

Related topics include uncertainties in measurements, uncertainty quantification, error analysis, metrology, and calibration.