Modern research runs on shared datasets and shared code. A single paper may rely on a public cohort, three R packages and an analysis pipeline someone else spent two years building — and then credit none of them, beyond a line in the acknowledgements. That is not just discourteous: it breaks reproducibility, because a reader cannot tell which version of what you actually used, and it denies credit to people whose careers depend on demonstrating impact.
Why formal citation matters
- Credit that counts. Acknowledgements are invisible to indexers; references are tracked and counted.
- Reproducibility. "Analysed in R" is not enough — results can differ between versions of a package.
- Provenance. Readers can find and verify the exact data behind your figures.
Citing a dataset
Include the same elements you would for any source, plus version:
Creator(s). (Year). Title of dataset (Version 2.1) [Data set]. Repository. https://doi.org/10.xxxx/xxxxx
Cite the dataset itself, not only the paper that described it — they are separate objects, and many datasets are reused far beyond their original study. Most repositories display a recommended citation; use it.
Citing software and code
Author(s). (Year). Name of software (Version 1.4.2) [Computer software]. Publisher/Repository. https://doi.org/10.xxxx/xxxxx
Two practical rules. First, always give the version — this is the single most useful element for reproducibility. Second, check whether the developers ask you to cite a specific paper instead; many projects publish one, and their preference is usually stated in a README or CITATION file.
Where these citations go
Name software and versions in the methods so a reader can follow the analysis, and place the full citation in the reference list so it is machine-readable and countable. Your data availability statement then points to the datasets you produced or used.
Making your own outputs citable
- Deposit data in a repository that issues DOIs, following the FAIR principles.
- Archive a tagged release of your code — linking a code repository to an archiving service produces a DOI for each version.
- Add a CITATION file stating exactly how you want to be cited.
- Choose an explicit licence so others know what they may do.
- Reference your own dataset and code in your paper, so the link between them is discoverable in both directions.
Do this once and your outputs stop being invisible by-products of a paper and become citable contributions in their own right.