C.28:4.0 - Public contract and support components
When a question or result must be referenced, recover its content and use the corresponding contract:
CausalUseQuestionRefidentifies the exact question content, normally a C.2.1 episteme.CausalEstimandRefidentifies the mathematical target or the episteme that describes it under its direct pattern.PotentialOutcomeContrastRefidentifies the exact contrast or its description.CausalUseSupportResultRefidentifies one C.2.1 result episteme defined below.
Support is composable. A real result may use several components:
CausalSupportComponentRefs:
evidencePathRefs?
empiricalDataRegimeRefs?
identificationResultRef?
estimateResultRef?
counterfactualSamplingRealizabilityResultRef?
simulationResultRef?
targetTrialMappingResultRef?
offPolicyCausalEvaluationResultRef?
causalVariableRepresentationRecordRef?
transportabilityResultRef?
These fields answer different questions. Do not compress them into one exclusive value. Each optional specialist ref identifies the exact result or record actually used. Keep a component’s own assumptions, uncertainty or sensitivity, supported and unsupported uses, and reopen information with that component when its defining contract requires them; do not copy fields merely to complete a standard list. The common CausalUseSupportResult still states its own supportedUse, unsupportedUse, limits, optional evidence window, and reopenCondition. Naming a specialist subject in prose does not make its result available to a consumer.
| Component | Question it answers | Does not establish |
|---|---|---|
| evidence path and data regime | What observations, assignments, or samples are available, and where did they come from? | identification or a valid estimate |
| identification result | Can the estimand be expressed or bounded from those data and assumptions? | a numerical estimate or direct sampling |
| estimate result | What value and uncertainty were obtained under an identification or design basis? | identification by the number alone |
| counterfactual-sampling realizability result | Can samples from the declared target distribution be obtained under the stated constraints, and how was that decided? | a WorkPlan, performed sampling, resulting data, or identification of every target |
| simulation result | What did the model produce under its assumptions and validation? | realized evidence or an intervention effect |
| target-trial mapping result | How does the declared trial protocol map to one observational data source, and which gaps, residual-confounding risks, and sensitivity checks remain? | identification, low risk of bias, or a valid estimate by reporting completeness alone |
| off-policy evaluation result | What does logged behaviour support about one evaluation policy under the stated history, confounding, overlap, endpoint, estimator, and uncertainty conditions? | authority to deploy or unqualified policy optimality |
| causal-variable representation record | Which learned, selected, or abstracted variables preserve the interventions, invariances, and queries needed for this use? | causal validity for every query, shift, or domain |
| transportability result | Which support transfers between exact endpoints, under which assumptions? | transfer by a shared label or population name |
CausalEmpiricalDataRegime is a local classification used only when it helps distinguish evidence:
CausalEmpiricalDataRegime =
observationalOrNaturalBehaviorData |
randomizedInterventionData |
governedInterventionData |
realizedCounterfactualSamplingData
realizedCounterfactualSamplingData is used only when an A.10 evidence path cites dated sampling Work and the resulting sample or data. A realizability result or WorkPlan alone establishes no empirical regime. Model output is recorded separately through simulationResultRef, not as an empirical regime.