G.9:4.3b — Conditional causal method rung parity

Use this extension only when a parity report compares causal methods or causal-use claims. Start with a cheap screen and stop at degraded parity or abstention when the methods answer different questions.

CausalRungParityScreen:
  comparedMethodsRef
  targetCausalityLadderRungSet
  causalSupportComponentTypeSet
  sameEstimand: yes | no | unclear
  sameOutcomeWindow: yes | no | unclear
  sameTransportEndpoints: yes | no | unclear
  cheapParityStop:
    comparableEnoughForFullRecord |
    crossRungDegrade |
    crossSupportComponentsDegrade |
    differentEstimandAbstain |
    differentOutcomeWindowAbstain |
    differentEndpointsAbstain |
    returnToC28

causalSupportComponentTypeSet records which methods rely on evidence paths/data regimes, identification, estimates, direct counterfactual sampling, simulation, or transport. Difference is not an automatic ban, but it must be exposed and bridged; one label cannot make unlike components equivalent.

Open the full record only when comparison remains meaningful:

CausalMethodRungParityRecord:
  comparedMethodsRef
  causalUseQuestionRef?: CausalUseQuestionRef
  targetCausalUseClaimKind: CausalUseClaimKind
  targetCausalityLadderRung: CausalityLadderRung
  causalEstimandRef: CausalEstimandRef
  declaredCausalityLadderBridgeOrLossRef?
  interventionBudgetOrActionSetRef?
  causalSupportComponentRefs: CausalSupportComponentRefs
  declaredCausalSupportLossRef?
  causalUseSupportResultRef?: CausalUseSupportResultRef
  causalFollowUpWindowRef
  outcomeMeasureRef
  sourcePopulationRef?
  targetPopulationRef?
  sourceDomainRef?
  targetDomainRef?
  sourceEnvironmentRef?
  targetEnvironmentRef?
  sourceDataGeneratingRegimeRef?
  targetDataGeneratingRegimeRef?
  transportabilityResultRef?
  estimateResultRef?
  parityVerdict: parityEstablished | degraded | abstain
  supportedParityUse
  unsupportedParityUse

The record names every changed transport endpoint that matters; population and semantic scheme do not substitute for domain, environment, or data-generating regime. Different rungs, estimands, windows, endpoints, or support components require a bridge/loss, degraded parity, or abstention. G.9 makes the parity conclusion; C.28 supplies the cited causal-support result and does not authorize the benchmark conclusion.