No practical measurement determines a quantity with unlimited certainty. Instruments have finite resolution, experimental conditions fluctuate, calibration references have uncertainties and measurement models may be imperfect. Measurement uncertainty provides a quantitative description of the doubt associated with a measurement result.
Uncertainty is not identical to error. Measurement error is the difference between a measured value and a reference or true value, although the exact true value is generally unknown. Uncertainty describes the range of values that may reasonably be attributed to the measurand based on the available information.
The Guide to the Expression of Uncertainty in Measurement (GUM) distinguishes two broad methods for evaluating uncertainty: Type A and Type B. Type A evaluation uses statistical analysis of repeated observations. Type B evaluation uses other information, such as calibration certificates, instrument specifications, previous measurements, reference data, resolution limits or expert knowledge.
Both types are expressed as standard uncertainties, usually interpreted as standard deviations. The classification refers to the method of evaluation, not to whether an uncertainty source is random or systematic. For example, a systematic influence can be evaluated statistically, and a random influence may sometimes be evaluated using non-statistical information.
For repeated independent observations x₁, x₂, ..., xₙ, the arithmetic mean is used as an estimate of the measured quantity. The sample standard deviation describes the observed spread, while the standard uncertainty of the mean is the sample standard deviation divided by the square root of the number of observations.
When several independent uncertainty contributions affect a result, they are often combined using the root-sum-of-squares method. If the contributions are correlated, covariance terms must also be included. The resulting combined standard uncertainty is denoted uc.
Expanded uncertainty U is obtained by multiplying combined standard uncertainty by a coverage factor k. It is commonly used to report an interval around a measurement result. A coverage factor of approximately 2 often corresponds to a coverage level near 95% under suitable distributional assumptions, but this interpretation is not universal.
An uncertainty budget is a structured record of the individual uncertainty sources, their evaluations, sensitivity coefficients, probability distributions where relevant, correlations and contributions to the final result. It makes the measurement result more transparent and easier to reproduce.
Uncertainty should be reported with the measured value using appropriate significant digits. Excessive digits can imply a level of knowledge that the measurement does not support, while excessive rounding can discard useful information.
Measurement Uncertainty and Standard Uncertainty
Measurement uncertainty describes the dispersion of values that could reasonably be attributed to a measured quantity based on available measurement information. Standard uncertainty expresses this uncertainty in the form of a standard deviation and provides the foundation for combining, reporting and comparing measurement results.
For n repeated independent observations:
Mean: x̄ = (1/n) Σxᵢ
Sample standard deviation:
s = √[Σ(xᵢ − x̄)²/(n − 1)]
Standard uncertainty of the mean:
uA(x̄) = s/√n
For independent uncertainty contributions:
uc(y) = √[u₁²(y) + u₂²(y) + ... + uₙ²(y)]
Expanded uncertainty:
U = kuc(y)
For correlated input quantities, covariance terms must be included in the combined uncertainty.
Explanation
Variables
| Symbol | Meaning | Si Unit |
|---|---|---|
xᵢ | Individual measured observation | Same as the measurand |
x̄ | Arithmetic mean of repeated observations | Same as the measurand |
s | Sample standard deviation | Same as the measurand |
n | Number of observations | Dimensionless |
uA | Type A standard uncertainty | Same as the measurand |
uB | Type B standard uncertainty | Same as the measurand |
uc | Combined standard uncertainty | Same as the measured result |
U | Expanded uncertainty | Same as the measured result |
k | Coverage factor | Dimensionless |
Example or application
A worked example has not yet been added.
Definition
Measurement uncertainty is a non-negative parameter characterizing the dispersion of values attributed to a measurand based on the information used. Standard uncertainty is the uncertainty expressed as a standard deviation.
Purpose
To quantify the reliability and limitations of measurement results and provide a mathematically consistent method for combining and reporting uncertainty.
Importance
Uncertainty evaluation is essential for scientific reproducibility, precision measurements, instrument calibration, quality assurance and comparison of experimental results. Without uncertainty information, the significance of a numerical measurement cannot be properly assessed.
Types Of Uncertainty Evaluation
| Type | Description | Typical Sources | Typical Method |
|---|---|---|---|
Type A evaluation | Evaluates uncertainty using statistical analysis of repeated observations or other statistical data. |
| Calculate an appropriate statistical estimate, such as a sample standard deviation or standard uncertainty of the mean. |
Type B evaluation | Evaluates uncertainty using information other than the statistical analysis of the current repeated observations. |
| Assign a probability distribution justified by the available information and convert the assigned uncertainty into a standard uncertainty. |
Common Probability Distributions
| Distribution | Use | Standard Uncertainty |
|---|---|---|
Normal distribution | Often used when uncertainty information is given as a standard deviation or when many small independent effects contribute. | |
Rectangular distribution | Used when a value is considered equally likely within specified limits ±a. | u = a/√3 |
Triangular distribution | Used when values near the center of an interval are considered more likely than values near the limits. | u = a/√6 for a symmetric triangular distribution with limits ±a. |
Uncertainty Budget
An uncertainty budget lists the relevant input quantities and uncertainty contributions that influence a measurement result. It normally records the source, evaluation method, standard uncertainty, probability distribution, sensitivity coefficient, degrees of freedom where relevant and correlations. The budget helps identify which sources dominate the final uncertainty and where improvements will be most effective.
Combined Uncertainty With Correlations
For a result y = f(x₁, ..., xₙ), the first-order combined variance is calculated from sensitivity coefficients and the covariance matrix. In general, uc²(y) = ΣᵢΣⱼ (∂f/∂xᵢ)(∂f/∂xⱼ)Cov(xᵢ,xⱼ). If all input quantities are independent, the covariance terms for distinct inputs are zero and the expression reduces to a sum of squared contributions.
Expanded Uncertainty Interpretation
Expanded uncertainty U = kuc provides an interval around the measurement result. The coverage factor k must be reported or justified. A value of k = 2 is commonly used as an approximate 95% coverage interval when the output distribution is sufficiently close to normal and the uncertainty estimate has adequate degrees of freedom. For small datasets or non-normal distributions, a different coverage factor or method may be required.
Worked Example
Problem
Five repeated measurements of a length are 10.1 cm, 10.2 cm, 10.0 cm, 10.1 cm and 10.1 cm. Calculate the mean and Type A standard uncertainty of the mean.
Solution
The mean is x̄ = (10.1 + 10.2 + 10.0 + 10.1 + 10.1)/5 = 10.1 cm.
The deviations from the mean are 0, 0.1, −0.1, 0 and 0 cm.
The sum of squared deviations is 0.02 cm².
The sample standard deviation is s = √(0.02/4) ≈ 0.0707 cm.
The standard uncertainty of the mean is uA = 0.0707/√5 ≈ 0.0316 cm.
Result
Mean length = 10.100 cm; Type A standard uncertainty of the mean ≈ 0.032 cm. This calculation includes only the uncertainty estimated from the repeated observations, not any additional Type B contributions.
Reporting Guidelines
State the measured quantity and its unit.
Report the measurement estimate and its standard or expanded uncertainty.
Identify the coverage factor when expanded uncertainty is reported.
Explain the important uncertainty sources and evaluation methods.
Avoid reporting more digits than the uncertainty supports.
Use consistent units for the result and its uncertainty.
Include correlations when they materially affect the result.
Common Misconceptions
Uncertainty is not the same as measurement error.
Type A and Type B describe evaluation methods, not two mutually exclusive physical kinds of error.
Taking more measurements reduces some random uncertainty but does not necessarily reduce systematic uncertainty.
Expanded uncertainty is not automatically a guaranteed error limit.
A measurement with a small uncertainty is not necessarily accurate if important systematic effects have been overlooked.
Limitations And Considerations
The uncertainty model depends on the quality of the available measurement information.
Incorrect assumptions about probability distributions can produce misleading uncertainty estimates.
Correlations may significantly affect combined uncertainty.
A first-order uncertainty propagation model may be inadequate for strongly nonlinear relationships or non-Gaussian distributions.
Uncertainty evaluation does not correct a measurement; known significant biases should be corrected separately where possible.
Applications
Reporting experimental physical constants.
Calibration of measuring instruments.
Precision chemical analysis.
Quality control in industrial manufacturing.
Medical and environmental measurements.
Comparing results from independent laboratories.
Evaluating scientific databases and reference values.
Related Topics
SCI-REF-002-087
SCI-REF-002-088
SCI-REF-002-089
SCI-REF-002-091
SCI-REF-002-092
SCI-REF-002-095
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- JCGM 100 — Evaluation of Measurement Data: Guide to the Expression of Uncertainty in Measurement (GUM).
- JCGM 101 — Evaluation of Measurement Data: Supplement 1 to the GUM, Propagation of Distributions Using a Monte Carlo Method.
- Taylor, J. R. — An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements.