Experiment Preregistration¶
An experiment preregistration is a time-stamped, versioned commitment to a specific hypothesis, design, and analysis plan written before data are collected. Preregistration prevents HARKing (Hypothesizing After the Results are Known), curbs analytical flexibility, and creates a traceable audit trail that downstream FCC Critique personas can verify. Produce this artifact during the Find phase of any FCC cycle that touches empirical data — simulation, benchmarking, A/B testing, or model comparison.
Template¶
Section 1: Registration Metadata¶
Instructions: Give the preregistration a globally unique ID, link it to any existing Innovation / Ecosystem IDs, and record the author set and timestamp. Status must move monotonically:
Registered → In Progress → Completedwith anyAmendedtransitions explicitly documented below.
| Field | Value |
|---|---|
| Preregistration ID | [FILL — e.g. PRE-2026-001] |
| Innovation / Ecosystem ID | [FILL] |
| Date registered | [FILL — YYYY-MM-DD] |
| Authors | [FILL] |
| Version | [FILL — 1.0] |
| Status | [Registered / In Progress / Completed / Amended] |
Section 2: Hypothesis¶
Instructions: State the primary hypothesis in If → then → because form. Separate confirmatory secondary hypotheses from exploratory questions; the latter are never treated as confirmatory evidence later.
- Primary hypothesis:
[FILL] - Secondary hypotheses (confirmatory):
[FILL] - Exploratory questions (non-confirmatory):
[FILL]
Section 3: Study Design¶
Instructions: Declare the design family, enumerate all variables (independent, dependent, controlled, known-confounding), and commit to a data scope before any exploratory access.
- Design family:
[Observational / Randomised / Quasi-experimental / Simulation / A/B / Parameter sweep / Other] - Variables table:
[FILL] - Population / sample:
[FILL]
Section 4: Sample Size & Stopping Rule¶
Instructions: Commit to a planned sample size with a power justification (α, β, effect size) and a stopping rule (fixed-N, pre-specified sequential, or Bayesian). No post-hoc sample changes without amendment.
- Target N and rationale:
[FILL] - Power calculation inputs:
[FILL] - Stopping rule:
[FILL]
Section 5: Analysis Plan¶
Instructions: Lock the statistical method, multiple-comparison correction, effect-size measure, and success threshold. Sensitivity analyses and robustness checks belong here too.
- Primary analysis:
[FILL] - Correction method:
[FILL — Bonferroni / FDR / none] - Effect-size measure:
[FILL] - Success criteria:
[FILL] - Sensitivity analyses:
[FILL]
Section 6: Amendments¶
Instructions: Any change after data collection begins must be logged with date, original text, amended text, and rationale.
| Date | Section | Original | Amended | Rationale |
|---|---|---|---|---|
[FILL] |
[FILL] |
[FILL] |
[FILL] |
[FILL] |
Adoption Checklist¶
- All required sections completed
- Artifact peer-reviewed by at least one R.I.S.C.E.A.R. peer
- Stored in the project's designated docs location
- Linked from README or equivalent index
- Versioned + date-stamped before any data are touched
References¶
- PHOENIX v4.0.0 —
docs/resources/templates/open-science/preregistration.md - Center for Open Science (COS) — Simulation Studies Preregistration Template (2024)
- AAAI-26 — Reproducibility Checklist
- NeurIPS — Paper Checklist (16 categories)
- Nosek et al. (2018) — The preregistration revolution, PNAS 115(11)
FCC integration¶
This template is referenced from the Forensic Auditor persona
(src/fcc/data/personas/forensic_auditor.yaml) as part of the
Critique-phase evidence set. The auditor consumes preregistrations to
verify that downstream results match the pre-declared hypothesis and
analysis plan. See also the open-science quality gates in
src/fcc/data/governance/open_science_gates.yaml and the broader
gate catalogue in src/fcc/data/governance/quality_gates.yaml.