GCRI · Labs
Research protocols
A research protocol fixes, before data are collected, what a study asks, how it will answer and what would count as an answer: the hypotheses, the design, the controls, the outcomes, the sample and the analysis. Many failures of credibility trace back to a protocol that was vague, unwritten or changed after the data were seen. This page sets out how protocols are built, registered, reviewed and reported, and what ethics committees, editors and funders rely on when they accept a study’s claims.
How the work runs.
Protocol and replication board
Research protocols · Labs · Quarter 3
Protocols
| PR-378 sensor drift | Field experiment | Yes | Approved | Running |
| PR-705 AI triage | Randomised | Yes | Conditions | Setup |
| PR-694 message recall | Survey experiment | Yes | Approved | Complete |
| PR-360 heat exposure | Cohort | No | Pending | Draft |
Replication R-25
- Original effect
- 0.4 SD
- Replication effect
- 0.2 SD
- Verdict
- Smaller, same sign
- Published
- with data and code
Testbed availability
Fig.Protocol and replication board: example names and illustrative values, not a record or a live system.
Research protocols: how the risk develops.
Hypotheses, controls and scientific acceptance. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Flexibility becomes bias
When outcomes, exclusions, covariates or stopping rules are chosen after the data are seen, many analyses become possible and the most striking one tends to be reported. Each choice looks reasonable alone; together they inflate false positives and effect sizes.
Weak controls confound results
A missing control, unblinded assessment or non-random allocation lets something other than the intervention explain the difference. Design flaws cannot be repaired by analysis; they travel into every result and review built on the study.
Underpowered designs mislead twice
Small samples miss real effects, and the effects they do detect tend to be overestimated. Underpowered studies that reach significance seed literatures that later, larger replications cannot match.
Deviations go unreported
Protocols change for good reasons, such as slow recruitment or a safety signal. When a change is not dated, explained and reported, readers cannot tell a justified amendment from outcome switching, and reviews pool what should not be pooled.
How research protocols 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.
Model, equations and assumptions
The model structure used for research protocols, the assumptions it makes and how each is tested.
Ri = f(Hi, Ei, Vi, Ci)hazard, exposure, vulnerability and coping capacityEAL = ∑k pk · Lkover events k with annual probability pkxt+1 = A xt + B ut + εtA carries the couplings, u the shock, ε the error| Assumption | Why it is made | How it is tested |
|---|---|---|
| Hazard and exposure are independent within an event | Keeps the loss integral tractable | Compare with a copula fit on joint records |
| Couplings are linear near the operating point | Lets shocks propagate through A | Perturb beyond the linear range and compare |
| Vulnerability curves transfer between places | Local curves are rarely available | Validate against observed losses where they exist |
- Inputs
- versioned datasets with provenance
- Outputs
- distributions, never single numbers
- Code
- released with the model card and tests
- Reproduction
- same data, code, environment and seed
Protocols are registered before results in every domain.
- HealthIn practiceTrials register protocols following SPIRIT.What a reader can checkRegistration before enrolment
- WaterIn practiceField experiments state hypotheses and methods first.What a reader can checkPre-registration
- FoodIn practiceAgricultural trials state sites, seasons and treatments.What a reader can checkTrial registry
Protocols feed the Labs Evidence series. Registered Reports and EQUATOR guidelines set practice; studies with people follow the Declaration of Helsinki (revised 2024).
- Protocols for review.
- Replication proposals.
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 |
|---|---|---|---|---|
| Registration before data collection | percent of studies with a time-stamped registration | annual | Trial and study registries | Published statistics |
| Primary outcome consistency | percent of reports matching the registered primary outcome | annual | Registry and publication comparison | OBSERVED |
| Sample size justification | percent of protocols with a power or precision rationale | annual | Protocol audit | OBSERVED |
| Reporting guideline adherence | percent of applicable checklist items reported | per study | Assessment against CONSORT, PRISMA, ARRIVE or STROBE | ESTIMATED |
| Protocol amendments | count per study, each with a stated reason | per study | Registry history and ethics committee records | Published statistics |
| Registered report availability | number of journals offering the format | annual | Center for Open Science journal list | REPORTED |
| Results posting timeliness | months from primary completion to posted results | per study | Trial registries | Published statistics |
| Randomisation and blinding reported | percent of eligible studies | annual | Risk-of-bias assessments | ESTIMATED |
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 |
|---|---|---|
| OSF Registries (Center for Open Science) | Time-stamped preregistrations and registered report protocols across disciplines | Open, licence per registration |
| ClinicalTrials.gov (US National Library of Medicine) | Registered trials, protocol change history and posted results | Public domain |
| WHO International Clinical Trials Registry Platform | Search across primary national trial registries | Free, WHO terms of use |
| PROSPERO (NIHR) | Registered protocols for systematic reviews | Free access, terms of use |
| protocols.io | Versioned, citable step-by-step laboratory and field methods | Open, CC BY for public protocols |
| EQUATOR Network library | Reporting guidelines by study design, with extensions | Free access, licence per guideline |
| Crossref metadata | Links between registrations, articles, corrections and funders | Open metadata, no restrictions |
Who decides, and what the evidence serves.
Labs 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 |
|---|---|---|
| Research ethics committees and institutional review boards | Whether a protocol’s risks, consent and design justify approval, and whether an amendment needs fresh review | Complete protocol, sample size rationale, amendment history with reasons |
| Journal editors | Whether to accept a registered report in principle, or whether a submitted analysis matches its plan | Time-stamped registration compared line by line with the manuscript |
| Research funders | Whether a proposal’s design can answer its question before money is committed | Power or precision rationale, control conditions, data management plan |
| Medicines and device regulators | Whether trial evidence was generated under the protocol submitted to them | Registry history, deviation logs, results posting record |
Outputs this topic produces
- Protocol templates stating hypotheses, controls, outcomes and analysis plans before data collection
- Registration records with time stamps and version history
- Deviation logs that date and explain every amendment
- Reporting checklists completed against the relevant EQUATOR guideline
What experts ask of any protocol.
- What exactly is the hypothesis, and what result would count against it?
- What is the comparison: a control group, a baseline period, a counterfactual model or a randomised arm?
- Which outcomes are primary, which are secondary, and how will each be measured?
- What sample, duration or number of runs is needed to detect an effect of the size that matters?
- What would make the team stop early, for safety or for futility, and who decides?
- What will be published if the result is null or the experiment fails?
Why experiments drift without a protocol.
Every experiment involves choices: which observations to exclude, which covariates to adjust for, which time window to use. Each choice may be reasonable alone, but when they are made after seeing the data, the number of possible analyses grows quickly and one of them will usually show something. This is the problem methodologists call analytic flexibility, and it is a central reason why published effects often shrink or vanish when repeated.
Interdisciplinary work adds its own drivers. Disciplines differ in what they treat as a control, how they handle missing data and what level of evidence they expect. A hydrologist, an economist and an epidemiologist may each consider a design rigorous by their own standard while disagreeing about what it shows. A protocol forces those differences into the open before work starts.
Incentives matter too. Positive, novel findings are easier to publish and fund, which rewards flexibility and penalises null results. Pre-registration and registered reports are the main structural responses.
Tools protocols draw on, and where they stop.
- 01
Pre-registration
A time-stamped record of hypotheses and analysis plans. It limits post hoc changes but cannot stop a poorly chosen question.
- 02
Power and design analysis
Estimates how much data is needed to detect a meaningful effect, using probability and statistics. It depends on assumed effect sizes that are often uncertain.
- 03
Causal designs
Randomisation, difference-in-differences, regression discontinuity and instrumental variables from causal inference. Each rests on assumptions that must be stated and, where possible, tested.
- 04
Acceptance criteria
Pre-agreed thresholds for when a method, model or prototype is considered to work. They make success testable but can be gamed if set too loosely.
A protocol cannot make a bad measurement good or fix a biased sample. Deviations from protocol are permitted when justified, but are reported with their reasons.
How it is studied, and what the methods cannot say.
- 01Probability and statisticsSample size and power planning · Meta-analysis with bias assessmentApplied here
- 02Causal inference and impact evaluationRandomised pilots and stepped rollouts · Quasi-experimental programme evaluationApplied here
- 03Uncertainty quantification and model validationMeasurement uncertainty budgets · Sensitivity analysis of simulation modelsApplied here
- 04Systems engineeringRequirements to verification traceability · Stage-gate readiness reviewsApplied 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.
What is the difference between a preregistration and a registered report?
Does preregistration forbid exploratory analysis?
Which reporting guideline applies to my study?
Can a protocol be changed after registration?
Is pre-registration required for every Labs study?
Can a protocol change during a study?
Who writes the protocol?
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.