testretest

Test-retest reliability, or test-retest stability, is the degree to which scores on a measurement instrument remain consistent across repeated administrations when the target construct is assumed to be stable. It assesses the temporal stability of the instrument.

To evaluate it, the same participants complete the same test on two (or more) occasions under similar conditions. The time interval should be long enough to reduce recall of previous responses but short enough to avoid real change in the construct. Reliability coefficients, such as Pearson or Spearman correlations or the intraclass correlation coefficient (ICC), quantify agreement between administrations.

Interpretation depends on the context and data type. Higher coefficients indicate greater stability. However, interval length, practice effects, memory, changes in the construct, and external factors can inflate or deflate estimates. Generally, coefficients above .70 are considered acceptable for research purposes, with higher values preferred for clinical decision making.

Applications include psychological testing, educational assessment, and survey research, where a stable trait or state is measured repeatedly. Limitations include that it only assesses stability over the chosen interval, assumes no real change in the construct, and can be biased by learning, fatigue, or regression to the mean.

Related concepts include parallel-forms reliability, which uses different but equivalent items; and internal consistency, which assesses consistency among items within a single administration. Test-retest is complementary to these approaches in evaluating the reliability of measurement instruments.