C.28:4.3 - Common causal-validity screen

Run only the questions relevant to the current claim. A live threat either points to an existing specialist field/result or lowers the support result; it does not trigger a mandatory dossier.

CommonCausalThreatScreen:
  causalUseQuestionRef
  interventionWellDefinedOrConsistency?: clear | liveThreat | notApplicable
  temporalOrdering?: clear | liveThreat | notApplicable
  exchangeabilityOrConfounding?: clear | liveThreat | notApplicable
  positivityOrOverlap?: clear | liveThreat | notApplicable
  interferenceOrSpillover?: clear | liveThreat | notApplicable
  selectionCensoringOrMissingness?: clear | liveThreat | notApplicable
  measurementErrorOrConstructShift?: clear | liveThreat | notApplicable
  transportToTarget?: clear | liveThreat | notApplicable
  routedThreatRefs?
  resultingSupportBoundary

Ordinary effect case. A randomized treatment study records interventionWellDefinedOrConsistency=clear, temporalOrdering=clear, positivityOrOverlap=clear, interferenceOrSpillover=notApplicable, selectionCensoringOrMissingness=clear, and measurementErrorOrConstructShift=clear for its declared target and window. The screen points to the trial and estimate results; it does not repeat them.

Countercase. An observational cohort has the right rung label and a plausible estimand, but records exchangeabilityOrConfounding=liveThreat and positivityOrOverlap=liveThreat because severity is unmeasured and one treatment region has no comparator. The claimed effect remains unsupported until a suitable design or new evidence closes those threats; a suitable bound can instead support a correspondingly bounded claim. “Observational data” was classified correctly; that label does not establish validity.