Analysis bias from restricting a dataset using the very outcome being studied.
This occurs when only cases meeting a criterion defined by the outcome itself (e.g. only the comparisons that reached statistical significance) are carried forward into further analysis. It can manufacture spurious patterns—such as apparent moderator effects—that reflect which studies were adequately powered rather than a genuine underlying relationship.
This guide was auto-drafted and is pending editorial review.