Preregistration
Make planned tests distinguishable from exploration
Preregistration records research questions, design and analysis decisions before the relevant outcomes are known. It can make confirmatory tests distinguishable from exploratory work and expose deviations transparently. Registration is useful only when sufficiently specific and timed appropriately. Secondary-data studies should document prior data access, and justified changes remain possible when disclosed.
Use for confirmatory research where flexible decisions could be influenced by observed results. Register hypotheses, exclusions, outcomes, models and stopping rules at a point that credibly precedes the relevant analytic knowledge.
Strengths
- Makes prospective decisions reviewable
- Reduces ambiguity about planned and exploratory analyses
Limitations
- Vague plans leave substantial undisclosed flexibility
- Registration alone does not guarantee good design or faithful reporting
Know the boundary
Preregistration is not peer review, and exploratory analyses remain valuable when labelled accurately.