GCRI · Registry
Evidence classes
Not all evidence carries the same weight. A peer-reviewed trial, a published statistic, a model output, a press statement and an owner’s own declaration can all support the same sentence, yet each warrants a different degree of reliance. Evidence classes make that difference visible. This page sets out how records are classed by evidence type, how confidence is stated, how reliance limits are set, and why claims drift beyond what their evidence supports.
Evidence classes on the record.
Evidence classes: how the risk develops.
Evidence type, reliance limits and confidence. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Class collapse
When observed, modelled, reported and declared evidence share one field with no label, a reader cannot tell a measurement from a projection or a self-assessment. The weakest source then borrows the authority of the strongest.
Claim creep
A finding moves from a cautious report to an abstract, a summary and a headline, losing qualifiers at each step. The claim widens while its evidence stays the same, until it exceeds what any source supports.
Single-source dependence
Many records rest on one upstream source cited by several intermediaries. Counting the intermediaries makes the evidence look plural and independent when it is one source repeated.
Confidence without basis
A confidence word such as likely or high is attached without saying what it rests on. Readers in different fields interpret the word differently, so one label carries different meanings to an editor, a funder and a court.
Record ledger.
The screen a team would use to follow this subject: its measures, states and decisions. Names are examples and values are illustrative.
Record ledger
Evidence classes · Registry · Ledger head 3142470
- Records registered5701
- Complete provenance, % of records92Normal
- Open corrections10Watch
- Rights holds4Info
Latest records
| NX-R-12221 | Programme status | v2 | 4472615 | Registered |
| NX-P-88287 | Publication | v3 | 1324692 | Registered |
| NX-E-72567 | Evidence record | v4 | 4410118 | Correction open |
| NX-C-27773 | Participation | v1 | 7438493 | Rights hold |
Version history
- v1NEWRegistered with sources
- v2FIXUnit corrected, reason recorded
- v3OPENCorrection requested
Evidence classes mean the same thing in every domain.
- WaterIn practiceA flood extent from radar is estimated, not observed.What a reader can checkLabel with each map
- EnergyIn practiceA capacity projection is a scenario.What a reader can checkScenario label
- HealthIn practiceCase counts are published statistics with reporting lags.What a reader can checkLag stated
Labels follow the IPCC uncertainty guidance (2010) in spirit and apply to every series in Reports. The proposed Nexus Standards would carry the label in each record’s metadata.
- Cases where a label was misapplied.
- Proposals to refine a class.
How evidence classes is modelled, measured and tested.
The model, its assumptions, its data and its uncertainty, in the form a scientist or engineer would check. Structures follow published methods. The values drawn are illustrative, not findings.
Datasets and provenance
The data classes behind evidence classes and what is recorded about each before it is used.
| Data class | Provenance recorded | Resolution stated | Licence checked |
|---|---|---|---|
| Satellite imagery, optical and radar | sensor, processing level, acquisition date | pixel size and revisit | per scene |
| Reanalysis and climate model output | model, experiment, member, version | grid and time step | per release |
| In-situ sensors and gauges | station, instrument, calibration record | point and sampling rate | per network |
| Population and asset exposure | method, vintage, input sources | cell size and unit | per release |
| Boundaries and reference geography | producer, edition, date | scale | per edition |
- Lineage
- W3C PROV record from source to figure
- Principles
- FAIR: findable, accessible, interoperable, reusable
- Gaps
- shown as missing, never filled silently
What is measured, and how well.
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Assertions with an evidence class | share of assertions | monthly | Schema validation | OBSERVED |
| Assertions by evidence class | count by class | quarterly | Registry records | OBSERVED |
| Assertions resting on a single source | share of assertions | quarterly | Provenance graph | OBSERVED |
| Confidence statements with a stated basis | share of confidence statements | quarterly | Curation review | OBSERVED |
| Class reassignments after review | count | quarterly | Review records | OBSERVED |
| Agreement between independent classifiers | agreement coefficient | per audit | Double-coding audit | ESTIMATED |
| Claims exceeding their evidence class | count | monthly | Claim-to-evidence audit | OBSERVED |
Where the evidence comes from.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| GRADE handbook | A method for rating the certainty of evidence and the strength of recommendations | Free to read, GRADE Working Group terms |
| Cochrane Handbook for Systematic Reviews of Interventions | Methods for assessing risk of bias and synthesising evidence | Free to read online, Cochrane copyright |
| IPCC guidance note on consistent treatment of uncertainties | Calibrated language for confidence and likelihood in assessments | Free to read, IPCC terms |
| Evidence and Conclusion Ontology (ECO) | A controlled vocabulary of evidence types used in scientific curation | Open ontology, free to reuse |
| Oxford CEBM levels of evidence | A hierarchy of evidence types for clinical questions | Free to read, attribution |
| Campbell Collaboration methods guidance | Methods for evidence synthesis in social and economic policy | Free to read, Campbell terms |
Who decides, and what the evidence serves.
Registry prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Guideline panels and systematic reviewers | How much weight an assertion can bear in a recommendation | Evidence class, certainty rating and independence of the sources behind it |
| Funders and assessment panels | Whether reported outcomes are evidenced or only declared | Class labels separating observed outcomes from self-reported and planned ones |
| Journal editors | Whether a manuscript’s claims match the class of evidence it cites | Evidence class of each cited record and any single-source dependence |
| Regulators and public bodies | Whether evidence is strong enough to justify a rule or an intervention | Class, confidence basis and reliance limits for each supporting record |
Outputs this topic produces
- An evidence class assigned to each assertion and carried into every view and export
- A glossary of classes stating what each supports and what it does not
- Reliance notes marking assertions that rest on a single source
- Confidence statements with their stated basis and the reviewer who set them
What experts ask about evidence types.
- How should a classification separate the source of a claim from its reliability?
- How can a label survive aggregation, when estimates are mixed with observations?
- How should confidence be communicated to non-specialists without misleading them?
- When is a scenario mistaken for a forecast, and how is that prevented?
- What should a record say when the evidence type is unknown?
How evidence loses its label.
Evidence is most often misused not because it is wrong but because it is read as something it is not. A modelled estimate printed in a table beside published statistics acquires their apparent authority. A scenario designed to explore a possibility is quoted as a prediction. Assessment bodies in climate science, public health and statistics have each developed calibrated language for confidence and likelihood for this reason.
The Registry uses the evidence labels shared across the programme: Observed, Published statistics, Institution-reported, Estimated, Modelled, Forecast, Scenario, Expert judgement and Unknown. A label names a source type. “Published statistics” means published by the responsible agency; it is not a claim of endorsement. Estimates stay labelled as estimates and never replace published figures.
Classifying and qualifying evidence.
- 01
Source type
The label records how the evidence was produced, separately from any judgement of quality.
- 02
Reliance limits
Each record states the uses it can and cannot support, such as the geographic scale or period for which it holds.
- 03
Confidence
Where a method yields uncertainty, the record carries it, drawing on statistical practice from the STEM Commons. See data engineering, interoperability and stewardship for the data side.
- 04
Limits
Any classification simplifies. Mixed evidence needs the most cautious label of its parts, which can understate strong findings, and labels depend on honest declaration by issuers.
How it is studied, and what the methods cannot say.
- 01Data engineering, interoperability and stewardshipSchema-validated ingestion · Persistent identifier reconciliation · Lineage capture in pipelinesApplied here
- 02Cybersecurity and cryptographyTamper-evident record history · Signed exports and releases · Threat modelling of edit pathsApplied here
- 03Digital identity and public digital systemsIssuer authentication and authority · Contributor disambiguation by the personApplied here
Where this page’s references lead.
The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.
Questions and answers.
Is a declared assertion worthless?
Why can confidence words mislead?
Can an assertion’s class change?
Why are many citations not many sources?
Is “published statistics” a stronger label than Modelled?
Why keep an Unknown label?
Can a record carry more than one label?
Contribute evidence from this page.
Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
- EmailGateway GW-GCRISend it from this pageEvidence, a correction, a case, a method, a review, a translation or profile input. It arrives tagged with this page, so the editor who owns it receives it.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsGCRISee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Never send personal data, sensitive infrastructure detail or anything you have no right to share. A contribution is reviewed before use; sending it confers no authorship or editorial rights, and sponsorship never buys a finding.
Bring a method, a dataset or a question.
Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Legal documents.
Maps, data and figures on this site are shown for context and general information. They are not advice, forecasts or warnings; the limits of the information are set out in the Disclaimers and Data Notices.