Repeatability is the ability of a measurement method to give consistent results for a particular subject, i.e. a measurement will agree with prior and subsequent measurements of the same subject. Barnhart, Haber, and Lin (2007) emphasizes the importance of repeatability as part of an overall method comparison study, a view endorsed by . Before there can be good agreement between two methods, a method must have good agreement with itself. If one method has poor repeatability in the sense of considerable variability, then agreement between two methods is bound to be poor (Roy 2009). Barnhart, Haber, and Lin (2007) remarks that it is important to report repeatability when assessing measurement, because it measures the purest form of random error not influenced by other factors, while further remarking `.
strongly recommends the simultaneous estimation of repeatability and agreement be collecting replicated data. However Roy (2009) notes the lack of convenience in such calculations. Repeatability is defined by the “IUPAC Compendium of Chemical Terminology - the Gold Book” (2009) as “the closeness of agreement between independent results obtained with the same method on identical test material, under the same conditions (same operator, same apparatus, same laboratory and after short intervals of time)” and is determined by taking multiple measurements on a series of subjects.
Test-retest variability is practically used, for example, in medical monitoring of conditions.
A measurement is said to be repeatable when this variation is smaller than some pre-specified limit. In these situations, there is often a predetermined “critical difference”, and for differences in monitored values that are smaller than this critical difference, the possibility of pre-test variability as a sole cause of the difference may be considered in addition to, for examples, changes in diseases or treatments.
The British Standards Institute (1979) defines a coefficient of repeatability as the value below which the difference between two single test results may be expected to lie within a specified probability. In the absence of other indications, the probability is 95%.
The coefficient of repeatability may provide the basis of formulation a formal definition of a `gold standard’. For example, by determining the ratio of \(CR\) to the sample mean \(\bar{X}\).
Advisably the sample size should specified in advance. A gold standard may be defined as the method with the lowest value of \(\lambda = CR /\bar{X}\) with \(\lambda < 0.1\%\). Similarly, a silver standard may be defined as the method with the lowest value of $$ with \(0.1\% \leq \lambda < 1\%\). Such thresholds are solely for expository purposes.
http://goldbook.iupac.org/R05293.html; International Union
of Pure; Applied Chemistry.