A practical guide to understanding reliability in studies of within-person variability

This article concerns how to estimate reliability (defined as the internal consistency of responses to a scale) in designs that are commonly used in studies of within-person variability. I present relevant issues, describe common errors, make recommendations for best practice, and discuss unresolved issues and future directions. I describe how to estimate the reliability of scales administered in studies in which observations are nested within persons, such as daily diary and “beeper” studies and studies of social interaction. Multilevel modeling analyses that include a measurement level can estimate the occasionlevel (e.g., days or beeps or interactions) reliability of scales. In such models, items on a scale are nested within occasions of measurement and occasions of measurement are nested within persons.