1 Fundamentals
1.1 Definition
Measurement uncertainty is a quantitative estimate of the doubt associated with a measurement result. It does not represent a mistake, but rather the range of values that could reasonably be attributed to the measurand, the quantity intended to be measured. In practice, uncertainty helps describe how reliable a result is and how much confidence can be placed in it.
In metrology, uncertainty is treated as an essential part of the result, alongside the measured value itself. It is usually expressed as a standard uncertainty, an expanded uncertainty, or another form tied to a stated confidence level.
1.2 Measurement error vs measurement uncertainty
Measurement error is the difference between a measured value and the true value of the quantity. Because the true value is usually unknown, the exact error is rarely available. Measurement uncertainty, by contrast, is an estimate of the possible size of that error.
Error is a property of a single result in relation to a true value, while uncertainty summarizes the quality of the measurement process and the likely spread of possible values. A result can have low uncertainty and still be biased, or it can have higher uncertainty while remaining centered near the true value.
1.3 True value and measured value
The true value is the value that would be obtained by a perfect measurement. In real situations, it is generally not known exactly and may only be approximated. For this reason, metrology often uses the term conventional true value or reference value when a recognized benchmark is available.
The measured value is the outcome reported by an instrument, method, or procedure. It may differ from the true value because of random variation, calibration effects, reading limitations, or environmental conditions. Uncertainty describes the extent to which such differences are expected.
1.4 Sources of uncertainty
Measurement uncertainty can arise from many parts of the measurement process. Some sources are intrinsic to the instrument, while others come from the object being measured, the environment, or the observer. A complete uncertainty evaluation considers all significant contributors.
1.4.1 Random effects
Random effects cause unpredictable variation from one observation to another. Examples include electronic noise, small fluctuations in operator response time, and natural variation in repeated tests. These effects typically broaden the spread of results and are often studied through repeated measurements.
1.4.2 Systematic effects
Systematic effects shift results in a consistent direction. Common causes include calibration offsets, zero drift, and method-dependent bias. Such effects may not be obvious from repeat measurements alone, which is why external references and calibration data are often needed.
1.4.3 Resolution limits
Every measuring device has finite resolution, meaning it can only display or detect values in discrete steps. If the smallest readable increment is large relative to the quantity being measured, the result carries a corresponding limitation. This can be a major source of uncertainty in simple instruments and digital displays.
1.4.4 Environmental influences
Temperature, humidity, pressure, vibration, electromagnetic interference, and lighting can all affect measurement outcomes. Even when these factors do not directly alter the measurand, they may influence the instrument or the procedure. Their impact is especially important when measurements are made under changing conditions.
2 Classification of uncertainty
2.1 Type A evaluation
Type A evaluation uses statistical analysis of repeated observations. The uncertainty is estimated from the spread of the data, commonly through the sample standard deviation or the standard error of the mean. This approach is useful when enough repeated measurements are available to support a statistical treatment.
2.2 Type B evaluation
Type B evaluation relies on information other than repeated sampling. Sources may include calibration certificates, manufacturer specifications, prior measurements, expert judgment, or published data. The estimate is then converted into a standard uncertainty using an appropriate probability model.
2.3 Combined uncertainty
Combined uncertainty brings together all relevant standard uncertainty components into a single value. It is usually calculated by combining the individual components according to the measurement model. The result expresses the overall expected dispersion of the measurement outcome.
2.4 Expanded uncertainty
Expanded uncertainty is obtained by multiplying the combined standard uncertainty by a coverage factor. This gives an interval intended to contain the true value with a stated level of confidence or coverage probability. It is often the most practical form for reporting results in technical documents and certificates.
3 Estimation methods
3.1 Statistical analysis
Statistical analysis is used to characterize variability in observed data. Means, standard deviations, confidence limits, and regression methods can all contribute to uncertainty estimation. The choice of method depends on the measurement design and the amount of data available.
3.2 Repeated measurements
Repeated measurements help reveal short-term variation and provide a basis for estimating random uncertainty. When the measurements are independent, their scatter can be summarized numerically and used in further calculations. Repetition is especially valuable when the instrument response is stable but noisy.
3.3 Instrument specifications
Manufacturers often provide accuracy limits, resolution, linearity, and drift information. These specifications are useful when direct experimental evidence is limited. However, they must be interpreted carefully, since stated limits may describe maximum permissible error rather than a full probability distribution.
3.4 Calibration data
Calibration data compare an instrument’s readings with values from a recognized reference. Such data can reveal offset, scale error, and nonlinearity. They are often among the most important inputs in an uncertainty budget because they connect the measurement system to traceable standards.
3.5 Modeling and propagation
Many measurement results are derived from several input quantities through a mathematical model. Each input contributes its own uncertainty, and the model determines how these contributions affect the final result. This process is known as uncertainty propagation.
3.5.1 Law of propagation of uncertainty
The law of propagation of uncertainty provides a formal way to estimate the uncertainty of a result calculated from multiple inputs. It uses the measurement function and the uncertainties of the inputs to approximate the uncertainty of the output. For many routine cases, this is done with a first-order approximation.
3.5.2 Sensitivity coefficients
Sensitivity coefficients describe how strongly the output responds to changes in each input quantity. A large coefficient means that even a small input uncertainty may have a substantial effect on the result. They are central to determining which sources dominate the final uncertainty budget.
4 Reporting measurement uncertainty
4.1 Significant figures and rounding
Uncertainty should be reported with a level of precision that matches its reliability. Excessive digits can imply a false degree of accuracy, while too little detail can obscure important information. As a general practice, the measured value is rounded to the same decimal place as the uncertainty.
4.2 Confidence intervals and coverage factors
Confidence intervals provide a range that is expected to contain the true value under specified statistical assumptions. Coverage factors are numerical multipliers used to convert standard uncertainty into expanded uncertainty. Together, these concepts make it possible to state both magnitude and confidence in a compact form.
4.3 Uncertainty statements
A clear uncertainty statement identifies the value, the uncertainty, and the basis for the estimate. It may also specify whether the uncertainty is standard or expanded, the coverage factor used, and the confidence level associated with it. Such statements improve comparability between results from different laboratories or methods.
4.4 Traceability and documentation
Traceability links a measurement result to recognized standards through an unbroken chain of calibrations and comparisons. Documentation records the methods, assumptions, data sources, and calculations used in the uncertainty evaluation. Together, these practices support reproducibility, auditability, and trust in the reported result.
5 Uncertainty propagation
5.1 Addition and subtraction
When quantities are added or subtracted, their uncertainties combine according to their variances, not simply by direct addition in most cases. Independent components usually contribute in quadrature. This means that the overall uncertainty is often smaller than the sum of the individual limits.
5.2 Multiplication and division
For multiplication and division, relative uncertainties are often more useful than absolute ones. The combined effect depends on the relative contributions of each factor in the expression. This approach is common in derived quantities such as density, concentration, and efficiency.
5.3 Nonlinear functions
Nonlinear functions can magnify uncertainty in uneven ways. In such cases, linear approximations may be adequate only over a limited range. If the relationship is strongly curved or the uncertainties are large, numerical methods or simulations may be more appropriate.
5.4 Correlated quantities
Correlated quantities are not statistically independent, so their uncertainties do not combine as if they were unrelated. Shared calibration sources, common environmental conditions, or repeated use of the same instrument can introduce correlation. Proper propagation requires accounting for these relationships, often through covariance terms.
6 Standards and guidelines
6.1 Guide to the Expression of Uncertainty in Measurement
The Guide to the Expression of Uncertainty in Measurement, widely known as the GUM, is a foundational document in modern metrology. It sets out principles for evaluating and reporting uncertainty in a consistent manner. Its framework is used in laboratories, calibration services, and technical standards around the world.
6.2 International metrology practices
International metrology practices aim to ensure that measurements are comparable across instruments, laboratories, and countries. These practices emphasize traceability, standardized methods, and clear uncertainty evaluation. They are especially important when results are used in trade, research, and quality assurance.
6.3 Laboratory accreditation requirements
Accredited laboratories are generally required to demonstrate competence in estimating and reporting uncertainty. Accreditation bodies typically expect documented procedures, trained personnel, and evidence that measurement claims are technically justified. Uncertainty evaluation is therefore part of laboratory quality systems as well as technical practice.
7 Applications
7.1 Laboratory measurements
In laboratory settings, uncertainty affects analytical chemistry, physics experiments, and routine calibration work. It helps determine whether a result is fit for its intended use and whether differences between samples are meaningful. Accurate uncertainty estimates also support method validation and quality control.
7.2 Engineering and manufacturing
Engineering relies on uncertainty to define tolerances, verify performance, and compare parts against specifications. In manufacturing, it can influence inspection decisions, process control, and acceptance testing. A well-characterized measurement process reduces the risk of rejecting acceptable products or accepting defective ones.
7.3 Environmental monitoring
Environmental monitoring often involves measurements that vary with location, time, and sampling conditions. Uncertainty is needed to interpret readings of temperature, pollutant concentration, rainfall, or radiation levels. It also helps distinguish real trends from noise in observational data.
7.4 Medical and clinical testing
In medical and clinical testing, uncertainty affects the interpretation of laboratory values and diagnostic thresholds. Measurements such as blood analytes or imaging-derived quantities may be close to decision limits, making uncertainty especially important. Clear reporting supports more informed clinical judgment.
7.5 Scientific research
Scientific research depends on uncertainty to evaluate reproducibility, compare results, and test hypotheses. It provides context for whether observed differences are likely to reflect real effects or measurement noise. In this way, uncertainty is closely tied to experimental rigor and scientific credibility.
8 Common challenges
8.1 Underestimating uncertainty
A frequent problem is assigning too small an uncertainty because not all sources have been considered. This may happen when only repeatability is evaluated, while calibration, drift, or environmental effects are ignored. Underestimation can lead to overconfident conclusions and poor decisions.
8.2 Distinguishing bias from variability
Bias and variability are related but distinct. Bias shifts the average result, while variability describes the spread of results around that average. Confusing the two can produce an incomplete uncertainty analysis, especially when a method has both random noise and systematic offset.
8.3 Limited sample sizes
Small datasets make it difficult to estimate uncertainty reliably. With few observations, random fluctuations can dominate the result, and statistical estimates may be unstable. In such cases, supplementary information from prior knowledge or instrument data becomes more important.
8.4 Complex measurement systems
Modern measurement systems may involve multiple sensors, software algorithms, and correction steps. Each stage can introduce its own uncertainty and correlations. Complexity makes it harder to identify dominant contributions, so structured modeling and careful documentation are essential.