Ecological fallacy

The error of inferring individual characteristics or relationships from patterns observed in groups. Group-level associations do not by themselves establish individual-level patterns.

Ecological correlation: A correlation calculated from aggregate measurements across groups rather than individual observations. It is the group-level association that the fallacy mistakenly transfers to individuals.
Robinson Crusoe fallacy: A mistaken inference about individuals drawn from aggregate data, named for a critique of ecological reasoning in economics. It illustrates the danger of treating a group-level relationship as evidence about individual behavior.
Atomistic fallacy: The error of inferring group-level patterns from relationships observed among individuals. It reverses the ecological fallacy's direction of mistaken inference.
Unit of analysis: The entity—such as a person, household, or region—about which a study makes claims. Matching claims to the measured unit prevents group observations from silently standing in for individuals.
Aggregation bias: Distortion that occurs when data are combined into groups and analyzed as aggregates. Aggregation changes the relationships visible in data and can invite individual-level misinterpretation.
Durkheim's study of suicide: Émile Durkheim's 1897 sociological study relating suicide rates to social integration and religious affiliation. Its group-level comparisons are often used to explain why aggregate results do not settle individual motives.
Individualistic fallacy: The error of explaining group-level outcomes solely through individual characteristics or behavior. It resembles atomistic reasoning but emphasizes inadequate explanations of group phenomena.
Cross-level bias: Bias arising when relationships at one level of analysis are used to estimate relationships at another. It describes a central risk when aggregate evidence is applied to individual-level questions.
Simpson's paradox: A statistical reversal in which an association within subgroups differs from or reverses the aggregate association. It demonstrates how an aggregate pattern can differ sharply from patterns inside groups.
Voting behavior: The study of how individuals and groups make electoral choices. A region's vote share cannot establish how particular residents voted or why.

Linked from 28 pages