15. Reliability
This chapter is a work in progress.
Reliability is the consistency of a measure. This is a crude, rough definition. More exactly, reliability is the extent to which observed or measured scores are consistent with differences in their true scores. We can never get a true score for participants, and so reliability helps us determine how close we got to a true score with observed scores.
There are three types of consistency:
Test-retest reliability: consistency over time (measured with a correlation, usually)
Internal consistency: consistency across items (measured with Cronbach’s alpha usually, but there are other methods out there including Omega which may be more appropriate; more info provided soon)
Inter-rater reliability: consistency across different researchers (various methods, including percent agreement, interclass or intraclass correlation, Cohen’s Kappa, Fleiss’ Kappa, etc.)