Repeatability or test-retest reliability[1] is the variation in measurements if they would have been taken by a single person or instrument on the same item and under the same conditions. A less-than-perfect test-retest reliability causes test-retest variability. Such variability can be caused by, for example, intra-individual variability and intra-observer variability. A measurement may be said to be repeatable when this variation is smaller than some agreed limit.
Test-retest variability is practically used, for example, in medical monitoring of conditions. 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.[2]
Contents |
According to the Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, the following conditions needs to be fulfilled in the establishment of repeatability:
Repeatability methods were developed by Bland and Altman (1986).[3]
If the correlation between separate administrations of the test is high (e.g. 0.7 or higher as in this Cronbach's alpha-internal consistency-table[4]), then it has good test-retest reliability.
The repeatability coefficient is a precision measure which represents the value below which the absolute difference between two repeated test results may be expected to lie with a probability of 95%.
The standard deviation under repeatability conditions is part of precision and accuracy.
Test-retest reliability is desirable in measures of constructs that are not expected to change over time. For example, if you use a certain method to measure an adult's height, and then do the same again two years later, you would expect a very high correlation; if the results differed by a great deal, you would suspect that the measure was inaccurate. The same is true for personality traits such as extraversion, which are believed to change only very slowly. In contrast, if you were trying to measure mood, you would expect only moderate test-retest reliability, since people's moods are expected to change from day to day. Very high test-retest reliability would be bad, since it would suggest that you were not picking up on these changes.