When you measure something you are conducting a
comparison of some sort. For instance, when using a ruler you compare the item being measured to the marks on the ruler.
There is always some
error in a
measurement. Perhaps the item being measured falls between the marks on the ruler, and you have to estimate (or
interpolate) the measurement. In cases where the error is known (perhaps you know that the inch mark on your ruler is actually an inch and a quarter), one can compensate for the error.
"When we make a measurement, we are prone to errors of unknown sign and magnitude. All errors that are unknowable to us and are uncorrected comprise the 'Measurement Error'. Since we do not know the sign and magnitude of the measurement error, we must attempt to quantify the extent of our uncertainty regarding this error."
(From "What is Measurement Uncertainty Analysis?" By James D. Jenkins)
Uncertainty components include temperature, resolution of the standard and/or the unit under test (UUT), and the uncertainty of the
calibration of each
traceable measurement standard used. The manufacturer's specifications for standards or instrument being tested may provide other parameters to be considered such as repeatability or stability versus time, etc. In addition, the following may apply:
· Thermal emf
· Contact resistance
· Local value of gravity
· Cosine error
· Humidity / density of air
· Parallax error
Measurement uncertainty can be expressed in different ways - depending on the needs or intent of the user. In general, the word "uncertainty" is used to mean the
expanded uncertainty at a 95%
confidence interval. This means 95% surety that the measurement error is within the stated uncertainty.