Workspace · Interpretation & reporting
996102e0
Results Interpretation & Reporting Center
Use this center after you have the evidence object or audit pattern in hand and before converting it into manuscript wording.
The contracts below are generated from one governed catalog. Each card identifies:
- the evidence layer;
- whether the signal is a runtime status, interpretation pattern, or protocol-bound label;
- the denominator that must remain visible;
- what the evidence can support;
- what it cannot support;
- bounded Methods, Results, and limitation wording;
- the guide/example/API route that provides the surrounding context.
Showing all 11 reporting contracts.
pass
None of the implemented structural-QC conditions was detected in the canonical representation.
Evidence can support
Structural preflight detected none of the implemented structural conditions.
Do not claim
The dataset is validated, high quality, calibrated accurately, or scientifically valid.
Methods wording
Structural preflight evaluated non-finite coordinates and timestamps, missing canonical identifiers, duplicate participant × trial timestamps, and decreasing within-unit time.
Results wording
Structural preflight returned pass; none of the implemented structural conditions was detected in the canonical representation.
Limitation wording
Structural pass status does not establish calibration, detector, AOI, missingness, measurement-model, or inferential validity.
review
At least one implemented structural condition was detected and requires contextual inspection.
Evidence can support
Structural preflight detected specified issue families that were inspected against source and preprocessing provenance.
Do not claim
Every flagged row, trial, or participant is invalid or should be excluded.
Methods wording
Structural preflight was followed by row/group diagnostic review against source and preprocessing provenance; repair, retention, and exclusion decisions were recorded explicitly.
Results wording
Structural preflight returned review because [issue families] were observed across [row/group denominator]; the affected observations or units were inspected before any decision was applied.
Limitation wording
A review state identifies representation-level conditions, not their cause or an automatic scientific exclusion rule.
unassessed
No active ReadinessThresholds rule participated in the declared policy.
Evidence can support
No active readiness rule was declared for this policy.
Do not claim
All observations, trials, or participants passed readiness.
Methods wording
No structural-readiness threshold was active in the declared policy, so readiness was not assessed.
Results wording
The readiness report returned unassessed because the policy contained no active criteria.
Limitation wording
Unassessed readiness provides no evidence that structural thresholds would be satisfied under another declared policy.
ready_under_policy
All evaluated trial and participant units satisfy the active researcher-declared readiness rules.
Evidence can support
The canonical study satisfies the declared active structural policy.
Do not claim
The data are universally high quality, scientifically valid, or safe to publish.
Methods wording
Analysis readiness was evaluated under the declared [policy name] policy after structural-QC review; cohort impact was previewed before any filtering step.
Results wording
Readiness status was ready_under_policy under [policy name]; [row/trial/participant retention] was observed for the declared preview/application scope.
Limitation wording
Readiness is policy-relative and does not establish calibration, detector, AOI, missingness, or inferential validity.
review_under_policy
At least one evaluated trial or participant unit fails one or more active researcher-declared readiness rules.
Evidence can support
Some units did not satisfy the declared policy and the counterfactual cohort impact was quantified.
Do not claim
Failing units must be excluded or that another threshold would be less scientifically valid.
Methods wording
The declared readiness policy was evaluated at trial and participant level, and cohort consequences were inspected before the predeclared application scope was applied.
Results wording
Readiness status was review_under_policy under [policy name]; trial-scope preview retained [denominator], whereas participant-scope preview retained [denominator].
Limitation wording
Policy failure is not an automatic exclusion and remains conditional on the declared criteria and scope.
complete space; direction and magnitude relatively stable
All declared valid branches are accounted for and the realised estimates vary little in both direction and substantive magnitude within the declared space.
Evidence can support
The observed endpoint changed little across the declared specification space, within the uncertainty dimensions actually represented.
Do not claim
Every reasonable analysis would agree, sampling uncertainty is small, or the result is universally robust.
Methods wording
The complete declared specification space was executed against one common endpoint and summarised descriptively across all valid represented branches.
Results wording
Across [N] valid represented specifications, estimates remained [direction] and ranged from [min] to [max], with [median/quantiles] indicating limited variation within the declared space.
Limitation wording
This robustness description is conditional on the declared decision space and does not cover untested analytical, measurement, or sampling uncertainty.
direction stable; magnitude materially variable
Estimates retain one direction but vary enough in magnitude to change quantitative or substantive interpretation.
Evidence can support
Direction was less sensitive than magnitude within the declared specification space.
Do not claim
The effect size is robust merely because the sign is stable.
Methods wording
Direction and magnitude stability were evaluated separately across the complete declared specification set.
Results wording
All represented estimates were [direction], but magnitude ranged from [min] to [max], indicating that quantitative interpretation was specification-sensitive.
Limitation wording
Materiality of the magnitude spread requires a study-specific substantive scale fixed independently of the observed robustness pattern.
estimates cross the scientific null
Positive and negative (or exact-null) estimates occur across defensible represented specifications.
Evidence can support
The directional conclusion depends on analytical choices represented in the declared audit.
Do not claim
The most favourable branch is the correct result, or sign fractions are posterior probabilities.
Methods wording
Direction stability was assessed across the complete declared specification space rather than by selecting one preferred branch.
Results wording
Estimates crossed the scientific null across [N] represented specifications, with [counts/fractions] positive, negative, and exact-null.
Limitation wording
Sign fractions are descriptive properties of the declared specification set, not probabilities about the true effect direction.
declared valid branches are missing, failed, or not run
The original audit denominator is not fully represented by successful endpoint estimates.
Evidence can support
Available executed branches show [observed pattern], but the audit remains incomplete relative to the declared valid denominator.
Do not claim
The missing branches can be removed from the denominator or treated as null/zero estimates.
Methods wording
Every declared branch was retained in the execution ledger, including predeclared-invalid combinations, technical failures, and branches not run.
Results wording
[successful] of [valid] valid specifications produced endpoint estimates; [failure/not-run accounting]. The available estimates showed [pattern], but the declared audit remained incomplete.
Limitation wording
Unobserved branch estimates are unknown; robustness cannot be generalised to branches that did not successfully yield the common endpoint.
factor-level or pairwise sensitivity diagnostic
Endpoint variation aligns descriptively with declared factor levels or factor combinations in the realised specification set.
Evidence can support
A declared factor or pair showed the largest descriptive alignment with observed specification-level variation.
Do not claim
The factor caused a percentage of uncertainty or the pairwise diagnostic is an inferential interaction effect.
Methods wording
Marginal and pairwise sensitivity diagnostics were used descriptively to localise variation within the declared specification set.
Results wording
The largest descriptive marginal sensitivity involved [factor], and the strongest non-additive pairwise pattern involved [factor A × factor B].
Limitation wording
These diagnostics are descriptive, can overlap, need not sum to one, and do not provide causal variance decomposition or inferential interaction tests.
incomplete / robust_negative / materially_fragile
These labels are canonical outcomes of specific frozen GazeAudit validation programmes.
Evidence can support
The named frozen case had its documented canonical outcome under its fixed protocol.
Do not claim
The label applies to a new dataset, synthetic example, different endpoint, or altered protocol.
Methods wording
Validation evidence was interpreted only within the frozen case-specific protocol and source identity.
Results wording
Under the frozen [case] protocol, the canonical outcome was [label]; this statement is restricted to that validation record.
Limitation wording
Frozen labels are not generic runtime statuses, quality grades, or transferable classifications for new studies.
No reporting contract matches these filters
Clear one or more filters, use site search with Ctrl/Cmd + K, or open the evidence-vocabulary reference.
Use the evidence hierarchy in the right order
Before choosing wording, answer these questions:
- What is the evidence object? Runtime status, branch-level estimates, sensitivity summary, or frozen protocol outcome?
- What denominator produced it? Rows, trial units, participants, valid specifications, successful specifications, or a frozen case-specific execution set?
- Is the signal generated by the package or interpreted by the researcher?
- Does the wording stay inside the uncertainty dimensions actually tested?
- What limitation must remain next to the claim?
If any of those cannot be reconstructed, resolve the evidence record before polishing the sentence.
Runtime status ≠ interpretation pattern ≠ frozen label
These three categories are intentionally different.
Runtime status
Examples:
StudyQCReport.status == "pass";StudyQCReport.status == "review";AnalysisReadinessReport.status == "unassessed";AnalysisReadinessReport.status == "ready_under_policy";AnalysisReadinessReport.status == "review_under_policy".
These are software-returned states with bounded meanings.
Interpretation pattern
Examples:
- complete specification execution with relatively stable direction/magnitude;
- stable direction but materially variable magnitude;
- sign-changing estimates;
- incomplete execution;
- descriptive marginal/pairwise sensitivity.
These are not package enum values. They require researcher interpretation of the complete evidence record.
Protocol-bound evidence label
incomplete, robust_negative, and materially_fragile belong only to the frozen validation cases that define them.
Do not transfer those labels to a new dataset.
Preserve the denominator before the adjective
A sentence such as:
The result was robust.
is not reconstructable.
A bounded statement starts with the evidence denominator:
Across all [N] valid represented specifications in the declared analysis space, estimates remained [direction] and ranged from [min] to [max].
Likewise, readiness reporting should say whether retention refers to rows, trial units, participant units, trial-scope filtering, or participant-scope filtering.
Avoid inferential upgrades
Descriptive GazeAudit outputs should not be silently rewritten as:
- confidence intervals;
- Bayesian credible intervals;
- posterior probabilities;
- statistical significance;
- causal variance decomposition;
- inferential interaction effects;
- universal data-quality grades.
If the study includes a separate inferential model, report that model under its own assumptions and denominator.
Reporting workflow
Locate the runtime status, complete branch ledger, sensitivity summary, or frozen evidence record.
Write down exactly which rows, units, participants, valid branches, or protocol runs belong in the denominator.
Separate observation from researcher judgement, and direction from magnitude where relevant.
Add the limitation that prevents the descriptive evidence from becoming a stronger inferential claim.
Bind the final wording to the evidence object, decision record, software identity, and manuscript location.
Reporting examples are templates, not automatic prose
Every copy button on this page returns bounded template language. Bracketed placeholders must be replaced with the actual evidence record. Remove wording that does not apply.
Do not paste a template into a manuscript without checking:
- endpoint identity;
- analysis timing;
- denominator;
- execution completeness;
- filter scope;
- magnitude scale;
- unresolved uncertainty;
- temporal evidence layer.
Companion guidance
- Claim-boundary reporting guide — systematic workflow for status → denominator → claim → limitation → archive.
- Reporting-language rewrite exercise — rewrite over-strong synthetic manuscript sentences into evidence-bounded language.
- Evidence and denominator vocabulary — exact runtime vs audit terminology.
- Interpret an audit result — interpretation sequence for robustness evidence.
- Reporting robustness — descriptive specification/sensitivity reporting.
- Publication audits — preserve the final evidence/wording record.
Machine-readable reference
The same governed contracts are available as reporting-contract-reference.json.
That JSON is documentation metadata. It does not classify an analysis automatically.