Required inputs
Registered hypothesis
Locked analysis
All eligible outcomes
Protocol deviations
Replication status
Ordered execution
Work the protocol
- 1
Verify the registration and data digest.
- 2
Run the locked analysis once.
- 3
Classify positive, null, inconclusive, or adverse.
- 4
Publish deviations and missingness.
- 5
Keep replication distinct from discovery.
Outputs
- Complete result report
- Deviation log
- Replication requirement
Done only when
- All outcomes reconcile.
- Classification follows the locked policy.
- Claims do not exceed the observed protocol.
Fail closed
Refusal conditions
- A null result is omitted.
- An adverse result is relabelled exploratory.
- A same-dataset rerun is called replication.