Standard uncertainty expresses uncertainty in the form of a standard deviation. It is the common scale used to combine different uncertainty contributions. Type A evaluation is performed by statistically analyzing a set of observations obtained under specified conditions. For repeated independent measurements of the same quantity, the arithmetic mean estimates the measured value, while the experimental standard deviation describes the dispersion of individual observations. The standard uncertainty of the mean is the sample standard deviation divided by the square root of the number of observations. Type B evaluation is based on information that is not obtained by statistical analysis of the current measurement series. Examples include calibration certificates, manufacturer specifications, instrument resolution, published reference values and previous experimental knowledge. The information must be interpreted using a suitable probability distribution. A certificate may provide an expanded uncertainty and coverage factor, allowing standard uncertainty to be calculated by division. A manufacturer's stated maximum permissible error may require a rectangular or another justified distribution, depending on what the specification actually means. Both methods produce standard uncertainties that can be used in a combined uncertainty calculation. Type A does not automatically mean random uncertainty, and Type B does not automatically mean systematic uncertainty. The classification depends on how the uncertainty is evaluated, not on the physical nature of its source.
Type A and Type B Evaluation of Standard Uncertainty
Type A and Type B evaluations are the two general approaches used to determine standard uncertainty. Type A evaluation uses statistical analysis of repeated observations, whereas Type B evaluation uses other relevant information such as calibration certificates, specifications, reference data and scientific knowledge.
For n repeated observations, u_A(x̄) = s/√n, where s = √[Σ(xᵢ − x̄)²/(n − 1)]. For a rectangular distribution with half-width a, u_B = a/√3.
Explanation
Variables
| Symbol | Meaning | Si Unit |
|---|---|---|
xᵢ | The i-th observed measurement value | Unit of the measured quantity |
x̄ | Arithmetic mean of repeated observations | Same unit as xᵢ |
s | Experimental sample standard deviation | Same unit as xᵢ |
n | Number of observations | Dimensionless count |
u_A | Standard uncertainty evaluated using Type A methods | Same unit as the measured quantity |
u_B | Standard uncertainty evaluated using Type B methods | Same unit as the measured quantity |
a | Half-width of a symmetric rectangular interval | Same unit as the measured quantity |
Example or application
A worked example has not yet been added.
Definition
Type A evaluation is the evaluation of standard uncertainty by statistical analysis of a series of observations. Type B evaluation is the evaluation of standard uncertainty by means other than statistical analysis of such a series.
Purpose
To convert different kinds of experimental and reference information into standard uncertainty values that can be combined consistently.
Importance
These evaluations allow uncertainty to be estimated even when repeated measurements are available only for some sources. They also establish a common framework for uncertainty budgets and traceable measurement results.
Type AEvaluation
Principle
Statistical variation in repeated observations is used to estimate the uncertainty associated with the measured quantity.
Procedure
Perform repeated measurements under defined conditions.
Check that the observations correspond to the same measurand and measurement process.
Calculate the arithmetic mean.
Calculate the sample standard deviation.
Calculate the standard uncertainty of the mean when the reported result is the mean.
Investigate trends, drift, dependence and unusual observations.
Important Distinction
The sample standard deviation describes the spread of individual observations. The standard deviation of the mean describes the uncertainty associated with the estimated mean under the applicable assumptions.
Type BEvaluation
Principle
Available non-statistical information is interpreted using an appropriate probability distribution and converted into standard uncertainty.
Common Information Sources
Calibration certificates.
Manufacturer specifications.
Reference standards and published data.
Previous measurements and experiments.
Instrument resolution.
Known physical limits and scientific knowledge.
Typical Distribution Models
| Distribution | Condition | Standard Uncertainty |
|---|---|---|
Rectangular | All values within a stated interval are considered equally plausible. | u = a/√3 for a symmetric interval with half-width a. |
Triangular | Values near the center are considered more plausible than values near the interval limits. | u = a/√6 for a symmetric triangular distribution with half-width a. |
Normal | A normal distribution is supported by the available information. | If an expanded uncertainty U is supplied with coverage factor k, u = U/k. |
Other justified distributions | The available information supports a different probability model. | Calculate the standard deviation from the adopted probability distribution. |
Worked Example Type A
Problem
A length is measured five times, producing 10.1 cm, 10.2 cm, 10.0 cm, 10.2 cm and 10.1 cm. Calculate the standard uncertainty of the mean.
Mean
x̄ = 10.12 cm
Sample Standard Deviation
s ≈ 0.0837 cm
Standard Uncertainty
u_A = 0.0837/√5 ≈ 0.0374 cm
Result
The Type A standard uncertainty of the mean is approximately 0.037 cm.
Interpretation
This result describes the statistical uncertainty of the mean under the assumptions of independent observations and a stable measurement process.
Worked Example Type B
Problem
A specification states that a quantity is within ±0.06 mm, and a rectangular distribution is considered appropriate.
Calculation
u_B = 0.06/√3 ≈ 0.0346 mm
Result
The Type B standard uncertainty is approximately 0.035 mm.
Interpretation
The interval has been converted into a standard uncertainty using the assumed rectangular probability distribution.
Degrees Of Freedom
Description
Degrees of freedom describe the amount of statistical information available for estimating a variance or uncertainty contribution.
Type A
For the sample standard deviation calculated from n independent observations, the degrees of freedom are generally n − 1.
Type B
Degrees of freedom for Type B evaluations depend on the reliability of the information and the assumptions used. They should not automatically be treated as infinite.
Limitations And Precautions
Repeated observations may be correlated, making the simple s/√n formula inappropriate.
A changing measurement process may invalidate the assumption of a stable population.
A rectangular distribution should not be selected automatically for every instrument specification.
Type B evaluations require justified assumptions about available information.
Increasing the number of repetitions reduces the Type A uncertainty of the mean only under suitable statistical conditions.
Important systematic influences may remain even when the Type A uncertainty is small.
Applications
Calibration laboratories
Experimental physics
Engineering measurements
Instrument testing
Uncertainty budgets
Research data evaluation
Reference
JCGM 100:2008, Evaluation of Measurement Data — Guide to the Expression of Uncertainty in Measurement (GUM).
NIST Technical Note 1297, Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results.
Related Topic
SCI-REF-003-082
SCI-REF-003-084
SCI-REF-003-085
SCI-REF-003-086
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.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.
Suggest a correction or source · More in Units & Measurement