Required inputs
Hypothesis family
Primary comparison
Adjustment method
Fixed or sequential stopping rule
Maximum sample
Ordered execution
Work the protocol
- 1
Enumerate every planned test.
- 2
Designate the primary test.
- 3
Freeze multiplicity adjustment.
- 4
Freeze when data collection stops.
- 5
Report all tests and looks.
Outputs
- Hypothesis-family manifest
- Adjusted results
- Stopping audit
Done only when
- Every planned test is accounted for.
- Adjustment is reproducible.
- Stopping follows the locked rule.
Fail closed
Refusal conditions
- Only the best frame is reported.
- Data collection stops after a favorable look.
- Exploratory tests are called confirmatory.