General idea
Every measurement result has an associated doubt. The uncertainty budget is the document in which each cause of that doubt is identified, its effect is estimated and it is systematically combined with the others. It follows the approach of the international guide to the expression of uncertainty in measurement, known as the GUM.
Steps to build it
- Define the measurand: exactly what is being measured and under what conditions.
- Write the measurement model: the relationship between the result and all the quantities on which it depends. If the result is the quotient of two readings, both appear in the model.
- Identify the sources of uncertainty: instrument resolution, calibration, repeatability, drift, temperature, setup effects, sample stability, applied correction and so on.
- Estimate each standard uncertainty: Type A uncertainties are obtained statistically from repeated measurements; Type B uncertainties come from certificates, specifications, experience or prior information, assigning a probability distribution (rectangular, normal, triangular) according to what is known.
- Determine the sensitivity coefficients: how much the result changes when each input quantity changes. They convert each contribution to the units of the result.
- Combine: if the sources are independent, the contributions are combined as the root sum of squares. If there is correlation between them, the corresponding term is added.
- Expand: multiply the combined uncertainty by a coverage factor, chosen for a stated confidence level, and report the result with its unit and that level.
How it is presented
A table with columns for the source, the estimated value, the distribution, the divisor, the sensitivity coefficient and the uncertainty contribution. It shows at once which sources dominate, and therefore where to invest effort to improve.
Common errors
- Forgetting important sources, such as the effect of temperature or the contribution of the reference instrument.
- Counting the same source twice.
- Confusing resolution, accuracy and uncertainty.
- Adding arithmetically instead of combining independent sources in quadrature.
- Omitting correlations when the same standard is used for several quantities.
How to interpret the result
The expanded uncertainty indicates the interval in which the true value is expected to lie at the stated confidence level. When it is compared with a tolerance, it is important to know how much of that tolerance the uncertainty uses up before declaring conformity.