Research Reproducibility Checklist

Reproducibility is the cornerstone of scientific integrity. When methods, data, and code are transparent, peer verification accelerates, funding efficiency improves, and public trust strengthens.

This checklist synthesizes current best practices from the NIH, Wellcome Trust, and major open-science initiatives. Use it during project planning, manuscript preparation, or lab-wide protocol standardization.

📦

Data Management & Storage

Adopt a FAIR Data Strategy

Ensure datasets are Findable, Accessible, Interoperable, and Reusable. Register in a trusted repository (Zenodo, Dryad, OSF) with persistent identifiers.

FAIR Principles

Document Metadata Rigorously

Include collection dates, instrumentation parameters, calibration records, units, and data dictionary mappings. Avoid ambiguous column headers.

Metadata

Version Control Raw & Processed Data

Maintain immutable raw copies. Log every transformation step with timestamps and script references. Use DVC or Git-LFS for large datasets.

Versioning

💻

Code & Computational Workflow

Containerize Analysis Environments

Use Docker or Singularity to capture OS dependencies, library versions, and runtime configurations. Publish container images with DOIs.

Docker/Singularity

Document Dependency Trees

Freeze package versions (requirements.txt, environment.yml, renv). Note hardware specs (CPU/GPU, RAM) for computationally intensive pipelines.

Dependencies

Implement Automated Testing

Include unit tests for critical functions and snapshot tests for output validation. CI pipelines should run on push to catch regressions early.

CI/CD

Seed Randomness Explicitly

Hardcode or log random seeds for ML models, simulations, and statistical sampling. Verify stochastic outputs remain stable across runs.

Reproducibility

🔬

Methodology & Experimental Design

Pre-register Study Protocols

Submit hypotheses, sample sizes, exclusion criteria, and analysis plans to OSF, AsPredicted, or clinical trial registries before data collection.

Pre-registration

Detail Inclusion/Exclusion Criteria

Specify participant demographics, material sourcing standards, and quality thresholds. Note deviations from the original protocol with justification.

Protocol

Report Effect Sizes & Confidence Intervals

Move beyond p-values. Include Cohen’s d, η², or odds ratios with 95% CIs. Report Bayesian factors or equivalence testing where applicable.

Statistics

📝

Reporting & Documentation

Adopt Field-Specific Reporting Guidelines

Follow CONSORT, STROBE, PRISMA, ARRIVE, or MIAME depending on your discipline. Journals increasingly mandate these for peer review.

Reporting Standards

Publish Negative & Null Results

Reduce publication bias by sharing failed experiments or non-significant findings in preprint servers or dedicated repositories.

Open Science

Provide Clear Access Instructions

Include README files, license types (CC-BY, MIT, Apache 2.0), and contact information. Specify any embargo periods or IRB restrictions.

Licensing

🤝

Peer Review & Collaboration

Enable Replication Packages

Bundle raw data, analysis scripts, output logs, and environment files into a single downloadable archive for reviewers and readers.

Review Process

Document Ethical & Compliance Approvals

List IRB/IACUC approval numbers, consent forms, data use agreements, and export licenses (e.g., dual-use research guidelines).

Compliance

Maintain a Lab-Wide Protocol Wiki

Centralize SOPs, troubleshooting guides, and equipment calibration logs. Assign version owners and schedule annual audits.

Lab Management

Implementation Best Practices

🔄 Iterative Adoption

Start with 3–5 high-impact items. Reproducibility is a practice, not a one-time audit. Revisit the checklist at project milestones.

🏛️ Institutional Alignment

Sync with your university’s research data management policy. Many funders now require explicit reproducibility statements in grant proposals.

🌍 Community Standards

Join discipline-specific reproducibility initiatives (e.g., Many Labs, Center for Open Science). Peer validation accelerates field-wide improvements.