Short answer
The combined standard uncertainty expresses the dispersion of the result as a standard deviation. To report an interval within which the value of the measurand can reasonably be expected to lie, it is multiplied by a coverage factor, k. The result is the expanded uncertainty: U equals k times the combined standard uncertainty.
Why it is needed
One standard deviation on its own corresponds to a modest level of confidence. In practice you want an interval with a higher probability of containing the value. Multiplying by k widens the interval; the larger k is, the higher the associated level of confidence.
Relationship to the distribution
The level of confidence depends not only on k but on the distribution of the results. If the distribution is approximately normal, k equal to two corresponds approximately to a level of confidence of about ninety-five percent. It is the most common choice, which is why many certificates state it.
When k equal to two is not enough
If the uncertainty is dominated by a few repeated measurements, the appropriate distribution is not the normal but Student’s t, and the value of k needed for the same level of confidence is larger. It is determined from the effective degrees of freedom, estimated with the Welch-Satterthwaite formula. Likewise, if a dominant component has a rectangular distribution, such as the resolution of an indicator, the normal assumption may be inappropriate.
What should appear on a certificate
- The expanded uncertainty.
- The coverage factor used.
- The associated approximate level of confidence.
With those three data you can recover the standard uncertainty by dividing U by k. This is useful when you need to combine the uncertainty of a standard with that of other sources in your own budget.
How to interpret the result
An expanded uncertainty of a certain size with k equal to two cannot be compared with another stated with a different k without first converting to the same basis. To decide whether an instrument meets a tolerance, you need to know what level of confidence backs the reported interval.