Entropy to Measure Intra-Subject Test-Retest Reliability

Operational definitions of reliability contradict its conceptual definition. Correlational techniques and internal consistency measures emphasize inter-subject reliability but omit intra-subject reliability. The proposed approach asserts that randomness is the error rather than assuming that the error is random. Using the entropy concept borrowed from information theory an index has been defined to quantify intra-subject reliability. The proposed index of intra-subject test-retest reliability was computed for 823 subjects in a 5 point Likert scale self-report personality inventory. The analysis of findings demonstrates that the proposed quantifier is appropriate to describe all of the possible response configurations. Independent intra-subject reliability indices will replace raw scores in computing inter-subject reliability coefficients. In addition it will be possible to report the reliability of the measurements in single-subject research.