How to Create a Research Reproducibility Package
Use a checklist for a paper’s reproducibility package and download a complete data, code, README and expected output example.
What goes in a research reproducibility package?
A reproducibility package gathers the files and instructions needed to rerun a reported analysis. Include a README, data or access instructions, a data dictionary, the analysis code, the tested software and data versions, and an expected result for comparison.
Download the study-session reproducibility package (ZIP). Extract it, open README.md, and run python analysis/analyze.py from the study-sessions/ folder. Compare results/means.csv with expected/means.csv.
Check what you can share
Review the rights and access rules for data, code, documentation and figures separately. If an input is restricted, describe how authorized readers can request it. A synthetic example can demonstrate the workflow for everyone else. Label any differences between the public demonstration and the original analysis.
Look for secrets, participant identifiers and private notes in the files you plan to share. Treat repository history as part of the release if you share a Git repository.
Test the handoff from a clean copy
- Open the README in a fresh copy and follow only what it tells you.
- Find or request the input and check its version.
- Set up the documented environment and run the analysis.
- Compare the new output with the expected result and record the revision.
Ask a colleague to try the same steps if you can. Fix the first instruction they had to infer. When sharing a release, point readers to the exact version you checked.
Run a five-minute handoff
Give a colleague a fresh copy of the example or your own small project and one question: “Can you recreate this table using the README?” Ask them to record where they hesitated.
Reproducibility package checklist
- The README identifies the reported result and names a starting command.
- Readers can locate or request the input data and interpret the fields.
- The code revision, software versions and data version are recorded.
- The package includes an expected result for comparison.
- Applicable licenses and access conditions are visible.
Show the handoff as a workflow diagram
If your project has several steps, sketch how the source data and analysis produce the final table. In Miss Formula’s AI Diagram Generator, start with four boxes: data, analysis script, result table, and independent check.
References
Adapted from a chapter by The Turing Way Community (CC BY 4.0). Miss Formula reorganized and expanded the explanations and wrote the worked example and exercises.