← Back to the library
Usability & InteractionMember ingredient
Simpson’s Paradox
An association in aggregated data can differ from, or reverse, the associations within relevant groups because the groups have different sizes or exposure patterns.
Public evidence overview
One checked source, with its limits.
This preview is not the complete evidence base. Use it as a question to investigate in your own product, not as a prediction of what an individual will do.
Checked source previewAcademic paper
Comment: Understanding Simpson’s Paradox
Research-assisted source review
- What it supports
- No purely statistical property identifies the correct adjustment in every case; the causal relation between treatment, outcome and the third variable determines the relevant comparison.
- Where it may not transfer
- Causal diagrams encode assumptions rather than prove them. The commentary is conceptual and does not validate a specific product analytics workflow.
See how sources and limitations are checked in our editorial method.
Member depth
Continue into the complete evidence and application.
Practitioner includes the complete checked source set, study context and practical takeaways for this ingredient. The public overview above remains available without upgrading.