CDISC-Gloss – The degree to which a psychometric ‘instrument’ is free from random error either by testing the homogeneity of content on multi-item tests with internal consistency evaluation or testing the degree to which the instrument yields stable scores over time.
NOTE: Reliability pertains to questions concerning whether an instrument is accurate, repeatable, and sensitive. Reliability is distinguished from validation, which answers whether the instrument (e.g., questionnaire) actually measures the selected “construct” (latent variable). For example, a balance (scale) is easily understood as a possibly valid instrument to measure body weight. Its reliability would be assessed by measuring the sensitivity, repeatability, and accuracy of the balance. The validity of using the balance for a particular purpose could then be established by comparing the measured reliability to the reliability required for that purpose. [After Patrick, D.L., 2003]. Compare to psychometric validation. See also validation, instrument.