Key ideas
- NIST states that the measurement model should relate the measurand to the input quantities and cover the significant sources of uncertainty.
- Contributions can be evaluated as Type A through statistical analysis of observations or as Type B by other means; they are expressed as standard uncertainties.
- The combined standard uncertainty incorporates the individual contributions and, where applicable, covariances; the interpretation of intervals depends on specific conditions.
Uncertainty is part of the result of a measurement
A measurement does not always consist of directly reading the quantity of interest. Sometimes the measured quantity—the measurand—is calculated from other quantities using a functional relationship. NIST presents this approach as a process that must represent not only a physical law, but also the measurement procedure and the relevant sources of variability. Its summary adapts guidelines from NIST technical note NIST TN 1297, itself based on the ISO international guide for expressing measurement uncertainty.
The distinction matters because the calculated value depends on the input values and the relationship used to combine them. NIST illustrates the approach with the power dissipated by a temperature-dependent resistor:
P = f(V, R0, b, t) = V2/R0[1 + b(t - t0)].
In this example, the estimate of power depends on the potential difference V, the resistance R0, the thermal coefficient b, and the temperature t. Therefore, uncertainties in those estimates contribute to the uncertainty of the result. The excerpt presents the formula as it appears in the source; it does not provide a derivation here or specify additional conventions for interpreting it.
To construct a useful evaluation, the model must include quantities that may contribute significantly to the uncertainty. NIST notes that, in addition to physical quantities, corrections and factors reflecting differences between observers, instruments, samples, laboratories, or times of observation may also be involved. Uncertainty is therefore not an isolated property of the instrument: it is evaluated in relation to the measurement process described.
Two ways to evaluate contributions
The NIST summary classifies components according to the method used to evaluate them. In a Type A evaluation, uncertainty is estimated through statistical analysis of series of observations. For that component, the standard uncertainty is represented by the estimated standard deviation, denoted as sᵢ in the explanation, and is associated with a number of degrees of freedom νᵢ.
A Type B evaluation uses means other than statistical analysis of series of observations. The contribution is represented by a quantity uⱼ, which can be regarded as an approximation of the corresponding standard deviation. The assumed probability distribution is based on the available information. In both cases, NIST expresses the components as standard uncertainties, that is, by quantities equivalent to estimated standard deviations.
| Aspect | Type A evaluation | Type B evaluation |
|---|---|---|
| Basis of evaluation | Statistical analysis of series of observations | Other means besides that statistical analysis |
| Representation described by NIST | Estimated standard deviation sᵢ | Quantity uⱼ, an approximation of the standard deviation based on the available information |
| Relationship indicated | uᵢ = sᵢ | The standard uncertainty is uⱼ |
The classification describes how a contribution is evaluated, not whether it is large or small. The material provided gives no numerical values for the resistor example and does not allow its uncertainty to be calculated: input estimates and their uncertainties are missing.
Combining contributions and handling corrections
NIST technical note TN 1297 explains that the combined standard uncertainty, for which the suggested symbol is u꜀, represents the estimated standard deviation of the result. It is obtained by combining the individual standard uncertainties and, where applicable, the covariances. This procedure is known as the law of propagation of uncertainty; in common usage it is also called the root-sum-of-squares method, or RSS, for combining components estimated as standard deviations.
Combining them does not eliminate the need to describe corrections. NIST indicates that a correction—or correction factor—is assumed to have been applied for each recognized systematic effect that significantly influences the result, and that efforts have been made to identify such effects. The uncertainty associated with that effect is the standard uncertainty of the applied correction.
This does not mean that error and uncertainty are synonyms. According to the technical note, the uncertainty of a correction is not the systematic error that the effect would have produced in the result; it represents the uncertainty due to incomplete knowledge of the required correction value. The source recommends applying corrections for recognized significant systematic effects, while acknowledging that doing so may not be practical in every case because of resource limitations.
What it means to express an interval
When the probability distribution of the result and its combined standard uncertainty is approximately normal, and u꜀(y) itself has negligible uncertainty, NIST states that the interval from y − u꜀(y) to y + u꜀(y) corresponds to an approximate confidence level of 68 %. The note describes conditions for this interpretation, including that the relevant conditions for approximating the distribution by a normal distribution are met and that the linear approximation used in propagation is adequate.
Expressing an interval does not authorize automatically calling any interval constructed from u꜀ a “confidence interval.” NIST reserves that statistical term for intervals that meet specific conditions, including Type A evaluation of all components contributing to u꜀. In its guidelines, an interval based on u꜀ is interpreted as an interval that covers a fraction p of the probability distribution characterized by the result and its combined uncertainty; p is the coverage probability or confidence level of the interval.
Scope of the method
The approach summarized by NIST provides a framework for modeling a measurement, evaluating its contributions, and communicating its uncertainty. It does not replace the specific identification of input quantities, justification of the assumed distributions, or evaluation of relevant corrections. The NIST page is a summary of guidelines; TN 1297 provides additional technical context on combining and interpreting uncertainty. Neither reference on its own provides data for assigning an uncertainty to a particular measurement.
As methodological context, CERN describes experiments that depend on accelerators and detectors, and explains that accelerator characteristics include, among other things, energy and luminosity. This description illustrates the variety of experimental processes in physics, but does not constitute independent confirmation of NIST's uncertainty rules or a quantitative example of their application.
What we know and its limits
The explanation summarizes the guidelines described by NIST and does not calculate uncertainties for a specific measurement. The material provided offers no input values, uncertainties, or distributions for applying the method to a real case. The approximate 68 % interpretation requires the conditions specified by NIST; not all intervals based on combined standard uncertainty are confidence intervals in the statistical sense. CERN is used only as context about experiments, not as independent corroboration of the NIST method.
Original sources3
The original sources for exploring this explanation and its limits.
- Experiments and Collaborations at CERN CERN
-
Measurement uncertainty
NIST
Institutional reference for context. It does not constitute independent confirmation of the main source's findings. - How the different contributions to uncertainty are combined NIST
Image credits
Flat lay of laboratory glassware including beakers and flasks on a neutral surface. · Photo by Tara Winstead on Pexels · Original source · Pexels License · Resized and re-encoded as JPEG; full framing preserved.