The paper says “data are available from the corresponding author on reasonable request.” A doctoral student writes, waits three weeks, and receives a spreadsheet called data_final_2.xlsx with columns V1 to V87 and no documentation. Formally, the data were shared. In practice, nobody can reuse them.
Open data has value when others can find it, access it, understand it and reuse it. Those four verbs are the four letters of the FAIR principles.
What FAIR means in practice
| Principle | Meaning | Concrete step |
|---|---|---|
| Findable | A persistent identifier and a description | Deposit in a DOI-issuing repository with a clear title and keywords |
| Accessible | A clear download route or access procedure | Public link, or stated access conditions |
| Interoperable | Open formats and standard vocabularies | CSV rather than proprietary formats, standard country codes, explicit units |
| Reusable | A licence and documentation | README, codebook, Creative Commons licence |
FAIR does not mean “everything public”. Sensitive data can be FAIR if the metadata is open and the access process is transparent.
Where to deposit
- A discipline-specific repository if your field has a trusted one (genomic, social survey, climate data). Specialists look there first.
- A generalist repository such as Zenodo, OSF, Dryad, Figshare or Harvard Dataverse: DOIs, low or no cost, easy to use.
- Your institutional repository if it exists and commits to long-term preservation.
Avoid cloud folders or personal websites as the official home: links die when you change jobs or run out of storage. A DOI outlives your employment.
What goes in the package
- Data files in open formats (CSV, TXT). Include .sav or .dta for convenience if you like, but always an open version.
- A codebook: variable names, original question wording, value codes (1 = strongly disagree…), units, and how missing values are marked (-99 or blank).
- A README: who collected the data, when, where and how; folder structure; how to cite the dataset.
- Instruments: questionnaires and interview guides, where copyright allows.
- Processing code that takes raw data to the analysis file.
The codebook is the part most often skipped, and it is cheapest to write while designing the questionnaire. Three years later, even you will not remember what V47 was.
What must not be shared
- Direct identifiers: names, phone numbers, emails, addresses, ID numbers.
- Indirect identifier combinations: date of birth plus postcode plus gender, or a rare occupation in a small area.
- Audio recordings and photos showing faces, unless separately consented.
- Data whose provider agreement forbids redistribution.
- Data participants were not told would be shared.
For these, you can share coarsened versions (age bands instead of birth dates), aggregated tables, or controlled access through a repository with a review process.
Choosing a licence
Without a licence, reusers have no clear idea what they may do. Two common choices:
- CC0: waives rights entirely. Many data repositories recommend it because it makes combining datasets easy.
- CC BY 4.0: free reuse with attribution.
Think twice before a non-commercial licence for data. The commercial boundary is blurry and deters reuse, including by researchers at private institutions.
Writing the data availability statement
- Weak: “Data are available from the authors on reasonable request.”
- Strong: “De-identified data, codebook and analysis scripts are deposited at Zenodo (DOI: …) under a CC BY 4.0 licence.”
- Strong for sensitive data: “Because the data contain health information, individual-level records are not public. Researchers may apply for access through the [name] ethics committee as described at …; aggregated data are available at …”
A good statement says what, where and under which conditions.
Versioning after publication
Data sometimes need correcting after deposit: a miscoded variable, a participant who withdraws. Do not overwrite the original silently. Repositories such as Zenodo let you publish a new version under the same concept DOI while each version keeps its own DOI, so readers can see exactly which file a paper used. Record what changed in a short changelog inside the README.
Citing data and getting credit
A dataset with a DOI is citable like any other output. Cite your own dataset in the article and list it on your CV and ORCID record. Reusers will cite it too, and funders increasingly look at this kind of contribution.
A checklist for the start of every project
- State in the consent form that de-identified data may be shared for research.
- Build the codebook alongside the questionnaire.
- Use a naming convention such as project_survey_2026-03_raw.csv, never “final” and “final2”.
- Clean data with code rather than editing by hand in a spreadsheet.
- Choose the repository before writing, so the DOI goes straight into the manuscript.
A useful test for this week: hand your latest dataset to a colleague with only the README and codebook, and ask them to reproduce your first descriptive table. Wherever they get stuck is where your package needs work.
Câu hỏi thường gặp
Is data sharing mandatory for publication?
It depends on the journal and funder. Many require a data availability statement, and some require deposit in a public repository unless there are ethical or legal reasons not to.
Should I use Zenodo or OSF?
Both issue DOIs and are free for typical use. OSF is strong for project management and preregistration, while Zenodo is simple for archiving a fixed version of data or code. Prefer a discipline repository if one exists.
Can qualitative interview data be shared?
Yes, if participants consented at the outset and transcripts are carefully de-identified. In many cases controlled access is more appropriate than open publication.
What should a codebook include?
Variable names, original question wording, value codes, units, how missing values are marked, and notes on any derived or recoded variables.
Will sharing my data let others scoop me?
The risk is often overstated. You can release data alongside the article, and a dataset DOI means you are credited whenever others reuse it.
