Preregistration

Preregistration is the act of publicly committing to a research plan — hypothesis, primary outcome, sample size, and analysis — before data are collected. It is a powerful anti-dashboarding measure (the "fishing" / "p-hacking" problem) and is now mandated or strongly recommended by many funders (US Office of Science and Technology Policy 2023 memo; the UK Reproducibility Network; the NIH analytic guide). This skill covers OSF Registries, AsPredicted, ClinicalTrials.gov (required for US clinical trials), and Registered Reports — the four preregistration pathways.

When to use

• Confirming a study design before data collection starts — to commit to analysis.
• Responding to reviewers or funder concerns about p-hacking / analytic flexibility.
• Required ClinicalTrials.gov registration for any interventional trial with a US site, NIH-funded, or FDA-regulated.
• Submitting a Registered Report (peer review of the protocol before results).
• Applying for a grant that asks for a preregistration plan.
• Planning a replication study where the design must match the original.

When NOT to use

• For preprints (posting a completed manuscript) — see ors-open-science-preprints.
• For data deposits without a study plan — see ors-open-science-fair-data.
• For code release (no hypothesis) — see ors-open-science-code-release.
• For exploratory / pilot / feasibility work that is intended to generate hypotheses. Preregistration is for confirmatory research.
• After data are already collected — preregistration is a pre-data commitment. You cannot preregister after the fact.
• For purely observational / database / secondary-data analyses — these often use analytic pre-registration (specifying the analysis plan, not the data collection) on OSF.

Prerequisites

• A complete study protocol (hypothesis, primary outcome, sample size, analysis plan).
• ORCID iD for all authors (required for OSF and AsPredicted).
• IRB/IACUC approval number (or proof of exemption) — most registries ask for it.
• For ClinicalTrials.gov: a Unique Trial Identifier (NCT number, if an amendment) or a sponsor-designated protocol.

Core workflow

1. Write the study protocol and analysis plan. Use a template (see "Document patterns" below). Be specific: sample size N, primary outcome variable, primary analysis model, alpha, power, correction for multiple comparisons.
2. Choose the platform. Use the decision tree:
   • Basic: AsPredicted — simplest (one page, 10 questions).
   • Standard: OSF Registries — most comprehensive, searchable, indexed in Crossref.
   • Regulated: ClinicalTrials.gov — required by law for US interventional trials.
   • Publishing: Registered Report format via a participating journal.
3. Upload to the platform. Fill in the structured fields; attach the protocol as a PDF.
4. Timestamp. The platform mints a DOI and date stamps the record. This is your proof of priority.
5. Collect the data. Do not look at the primary outcome until data collection is complete. No peeking (or disclose that you peeked).
6. Run the analysis as specified. If the analysis must change, disclose the change and explain why. Post-hoc changes are allowed but must be disclosed in the paper.
7. Link the preregistration to the paper. Include the DOI in the final manuscript; cite the preregistration in the methods. Check the box in the journal submission system.
8. Update the record for any changes (protocol amendments, sample-size re-estimation, analytic modifications).

Document patterns

Pattern 1: OSF vs AsPredicted vs ClinicalTrials.gov decision tree

Is this a US-regulated interventional clinical trial?
├── YES → ClinicalTrials.gov (required by law)
│         • Interventional (drug, device, gene therapy, etc.)
│         • US site OR NIH-funded OR FDA-regulated
│         • Must register BEFORE first patient enrolled
│         • Required fields: NCT number, recruitment status,
│           primary/secondary outcomes, analysis plan
└── NO  → Is this for a Registered Report (journal publication)?
         ├── YES → Use the Registered Report format from the target journal
         │        (e.g., Cortex, Open Science Framework, BMJ Open Science)
         │        • Journal sends protocol out for peer review
         │        • In-principle acceptance conditional on methods
         │        • Results paper gets fast-track if methods are sound
         └── NO  → Does the field expect a one-page confirmation?
                  ├── YES → AsPredicted (simplest, 10 questions)
                  │       • Common in psychology, social sciences, behavioral bio
                  │       • Lightweight, fast (~15 min to complete)
                  │       • No DOI (but has a unique URL and timestamp)
                  └── NO  → OSF Registries (full-featured)
                        • DOI ( searchable in Crossref)
                        • 50+ fields, all searchable
                        • Templates for different designs (RCT, survey, etc.)
                        • The default for biology, neuroscience, most fields

Pattern 2: The minimum preregistration fields (any platform)

Every preregistration must include:

1. Hypothesis (what you are testing)
2. Primary outcome (exactly what you will measure)
3. Sample size (N, or stopping rule)
4. Inclusion/exclusion criteria (who is in the study)
5. Analysis plan (the statistical model)

Optional but recommended:

• Secondary outcomes (exploratory outcomes)
• Power calculation (alpha, power, effect-size assumption)
• Correction for multiple comparisons (correction method: Bonferroni, FDR, etc.)
• Exclusion criteria (how you handle missing data, outliers)
• Data exclusions (any QC criteria decided before seeing the data)
• Analysis software (R, Python, JASP, etc.)

Pattern 3: OSF Registries preregistration template (abbreviated)

Most OSF fields map to this skeleton:

# Title
[Hypothesis in one sentence.]

# Hypothesis
[State the directional hypothesis: "We predict that X will increase Y..."]

# Primary outcome
[Name the variable; how it will be measured; units.]

# Secondary outcomes
[Optional: list up to 3 secondary exploratory outcomes.]

# Inclusion criteria
- Age range: [...]
- Population: [...]
- Other: [...]

# Exclusion criteria
- [...]
# Sample size
- Target N: [...]
- Power calculation: 80% power to detect d = [...] at α = 0.05

# Analysis plan
- Primary test: [e.g., linear mixed model with fixed effect of GROUP, random effect of SUBJECT]
- Covariates: [age, sex, baseline score]
- Alpha: 0.05 (two-tailed)

# Exclusion rules
- Any samples with < 80% missing data will be excluded.
- Outliers identified as > 3 SD from the mean will be flagged
  and sensitivity analyses run with and without them.

# Known deviations from the analysis plan
[None to date / Will be added if the protocol changes]

# IRB approval
- Protocol #[...] obtained from [Institutional Review Board name]
- Approval date: [...]

Pattern 4: AsPredicted (10 questions)

AsPredicted simplifies to 10 questions:

1. What is the hypothesis? (1 sentence)
2. What are the variables?
   • Independent variable(s):
   • Dependent variable(s):
3. How many participants will you collect data from? (N, or description of stopping rule)
4. When are you going to collect the data? (dates or "continuous")
5. Is this a between- or within-subjects design?
6. What analysis will you run to test the hypothesis? (specify the test, e.g., "t-test", "mixed-effects model")
7. What does that analysis tell you?
8. Any other analysis? (list any secondary/ exploratory)
9. How can we contact you? (email, ORCID)
10. Anything else?

Pattern 5: ClinicalTrials.gov required fields

For US interventional trials, you must register:

Register before first patient enrolled. Updates required within 21 days of a change (recruitment status, primary outcome, sample size). Results must be posted within 12 months of the primary completion date.

Pattern 6: Registered Report formats

Registered Reports come in three "formats" (not versions, but stages):

Most journals offer Stage 1 as the default. The workflow is:

1. Submit protocol (preregistration + analysis plan)
2. Stage 1 peer review → in-principle acceptance (IPA) or revise and resubmit
3. Collect data (do not look at results yet)
4. Submit results paper with protocol in supplement
5. Stage 2 peer review → publication OR major concerns (rare if IPA was given)

Journals that offer Registered Reports include:

• Cortex (and other Center for Open Science partner journals)
• eLife (exceptions apply)
• BMJ Open Science
• F1000Research
• Royal Society Open Science
• Open Science Framework journals

Common pitfalls

Validation

A preregistration is complete when:

• The hypothesis is stated as a directional prediction (not a question).
• The primary outcome is defined with exact variable name, measurement method, and time point.
• The sample size is justified with a power calculation (or an explicit Bayesian stopping rule).
• The analysis model is named (e.g., "linear mixed model", "two-sample t-test", "Cox proportional hazards").
• The platform has minted a timestamp and (for OSF) a DOI.
• All co-authors are listed and have approved the submission.
• The IRB/IACUC number is included.
• The preregistration is linked in the final manuscript (DOI in the methods).
• Any changes from the plan are disclosed in the final paper.

Open alternatives

References

• OSF Registries help: https://osf.io/registries
• OSF Preregistration templates: https://osf.io/prereg
• AsPredicted: https://aspredicted.org/
• AsPredicted FAQ: https://aspredicted.org/faq
• ClinicalTrials.gov: https://clinicaltrials.gov/
• ClinicalTrials.gov FAQs: https://clinicaltrials.gov/ctk/about-trials
• Registered Reports (cRedi): https://cos.io/rr/
• Registered Reports overview: https://cos.io/rr/
• SPIRIT 2013 statement: https://www.spirit-statement.org/
• CONSORT 2010: https://www.consort-statement.org/
• NIH Analytic Guide: supports preregistration for secondary analysis.
• US OSTP 2023 memo: requires federally funded trials to register.
• UK Reproducibility Network recommendations: https://ukrn.org/
• G*Power: https://gpower.software.informer.com/

Related skills

• ors-open-science-preprints — to post the results after data collection.
• ors-open-science-fair-data — to deposit the data after the paper is published.
• ors-open-science-code-release — to release the code used in the analysis.
• ors-omics-statistics- (multiple skills) — for statistical analysis methods.
• ors-bioinformatics-clinical- (multiple) — for clinical data handling.

Changelog

• 1.0.0 (2026-06-10): Initial adaptation by Pradyumna Jayaram. Synthesised OSF Registries help; AsPredicted FAQ; ClinicalTrials.gov fields; SPIRIT 2013; Registered Reports info (cRedi); NIH analytic guide; US OSTP memo. OSF vs AsPredicted vs ClinicalTrials decision tree; OSF template skeleton; AsPredicted 10 questions; Registered Report format comparison are original compositions.