exclusionsbias

Exclusions bias is a bias that occurs when relevant information, cases, or perspectives are systematically excluded from consideration in analysis, reporting, or decision making. This selective exclusion can distort findings and lead to overconfident or biased conclusions.

The bias can arise through methods that consciously or unconsciously omit data or viewpoints. Common mechanisms include strict inclusion or exclusion criteria in studies, discarding missing or outlier data without justification, selective reporting of favorable results, and censoring or truncating information in policy or media analyses. Exclusions bias overlaps with, but is not identical to, selection bias, as it primarily focuses on the influence of excluded elements rather than the process of choosing among included ones.

In practice, exclusions bias can appear in research (for example, excluding participants who drop out or fail to complete follow-up assessments in a way that favors a particular outcome), in journalism (omitting adverse events or counterexamples), and in data science (removing data from underrepresented groups before model training). Its consequences include biased effect estimates, reduced generalizability, and misguided decisions or policies.

Mitigation involves transparent documentation of exclusion criteria, preregistration of study design, and comprehensive reporting of all data and reasons for exclusion. Analytical approaches such as intention-to-treat analysis, sensitivity analyses, and robust data handling practices help reduce its impact. In data-driven fields, thorough data audits, justification for exclusions, and reporting of the impact of exclusions on results are recommended.

See also: selection bias, survivorship bias, publication bias, data dredging.