Five steps to evaluate measurement uncertainty:
- Analyse measurement system. Error budget
- Correct for known measurement errors
- Evaluate (standard) measurement uncertainties with methods of type A alternatively type B
- Combine standard measurement uncertainties => uc
- Expand measurement uncertainty => U = k• uc
Measurement result = ŷ ± U
Where possible, an error budget should be later verified with empirical results from e.g. method validation, internal control and laboratory comparisons. A lack of agreement can depend amongst others on
- underestimates of variations and uncertainties from different error sources;
- as yet unidentified error sources.
If the empirical results differ significantly from the error budget, then it should of course be revised with new estimates, the introduction of further components in the error model and/or modified functions of input quantities in the error model. If further error sources cannot be identified, then it may be necessary to use an error model for the final measurement result (that is not decomposed in terms of various input quantities) as can be estimated from empirical results.