On Mon, 16 Jul 2001, Bill Potts wrote:
> ...
> One need go no further than Webster's New Universal Unabridged Dictionary.
> ...
Webster's definitions are insufficient for calculating the precision
and determining the accuracy of experimental data. Consider the following:
Calculation of *precision* results from the analysis of *random errors*
in a set of measurements of the same quantity under circumstances which
are held as nearly constant as possible. Standard deviation (root mean
square deviation from the mean value) is the usual measures of the precision
of the set of observations.
Accuracy is a determination or estimate of *systematic errors* (the
closeness (lack of deviation) of the mean value of a set of measurements
from a reference value established by better quality instruments, better
techniques, and usually by better qualified observers or laboratories.
Accuracy is ofter only an estimated value because a better reference value
is not readily available, or does not exist.
A set of measurements can be of good precision but poor accuracy.
(e.g. tightly clustered measured values with small standard deviation, but
having a mean value widely in error compared with a more correct value)
Another set of measurements can be of good accuracy but poor precision.
(e.g. widely scattered measured values but having a mean value with little
deviation from a more correct value)
Try making a scatter graph of twenty hypothetical observed values which
illustrate good precision but poor accuracy; and a scatter graph of
twenty values which illustrate poor precision but good accuracy.
This is an easy exercise for a person who understands the difference
between precision and accuracy!
Gene.
