Reliability (measurement)

Reliability is the consistency of measurements, assessments, or research results across repeated occasions, items, or evaluators. It concerns reproducibility, not whether a measure captures the intended construct.

Test–retest reliability: The consistency of scores from the same measure administered to the same people on separate occasions. It estimates stability over time while assuming the measured trait has not changed.
Measurement error: The difference between an observed measurement and the value it is intended to represent. Unsystematic error makes repeated measurements less consistent.
Psychometrics: The field concerned with the theory and practice of measuring psychological attributes. Psychometric methods use reliability evidence to evaluate tests and scales.
Agreement (measurement): The closeness of repeated measurements or ratings to one another on their original measurement scale. High correlation can coexist with poor absolute agreement between measurements.
Internal consistency: The extent to which items in a measure produce related responses as indicators of a common construct. It captures consistency among items within one administration.
True score: In classical test theory, the expected score for a person across hypothetical equivalent measurements. Reliability estimates how much observed score variation reflects stable differences rather than error.
Measurement invariance: The condition that a measurement model relates items to a construct comparably across groups or occasions. Consistency alone does not ensure scores are comparable across populations or time.
Pearson correlation coefficient: A statistic measuring the strength and direction of a linear association between two variables. Correlation measures association, not necessarily agreement or interchangeable scores.
Inter-rater reliability: The degree to which different observers or evaluators assign consistent ratings to the same subjects or events. It isolates disagreement among evaluators as a source of measurement inconsistency.
Validity (measurement): The degree to which evidence and theory support interpretations of scores for a specified purpose. A measure can yield consistent scores without measuring the intended construct.

Linked from 7 pages