Glossary // Glossary term Technical
Selection bias
A distortion that arises when the people who end up in a study, or in its analysis, differ systematically from the population of interest in ways related to both exposure and outcome. Examples include studying only people who sought help, only participants who completed follow-up, or only crying episodes that respondents chose to remember. In causal terms it results from conditioning on a common effect of two variables, which distinguishes it from confounding by a common cause. How one could reason about recording a catharsis-like change of state: signals, features, inference and context, baselines, self-report, error costs and ethics, and why every step stays uncertain. Nothing on this site measures anyone. Every public claim on this site with its type, evidence level, confidence, sources and review date, and an appraisal of each reference. How claims are graded, how art is kept apart from evidence, and how to report a problem. The works the definition rests on, with the level of evidence each was appraised at. (2004). A structural approach to selection bias. doi:10.1097/01.ede.0000135174.63482.43 Supports the structural distinction between selection bias, which results from conditioning on a common effect, and confounding, which results from a common cause. Conceptual and methodological; examples come from epidemiology. (2018). Thinking clearly about correlations and causation: Graphical causal models for observational data. doi:10.1177/2515245917745629 Supports statements that correlation does not establish causation, that statistical control depends on assumptions, and that controlling for colliders or mediators can bias estimates. Methods guidance.Where the site uses it
Studies behind it