How to cite a dataset, software and a preprint, with worked examples

Why cite data, software and preprints at all?

Citing them gives credit to the people who produced them and lets readers find the exact material your results depend on. A finding can only be checked if the data and code behind it can be located, and a citation with a persistent identifier is the most reliable pointer there is.

Formal principles exist for both: the Joint Declaration of Data Citation Principles (2014) and the Software Citation Principles (2016) both ask that data and software be cited in the reference list, like papers, rather than mentioned only in the text or the acknowledgements. Many journals now require a data availability statement that cites the underlying datasets.

What does a complete citation need?

Whatever the style, the same facts make a citation complete. Styles only change their order and punctuation.

You will find these details on the record page of the repository that holds the item. On Life of Research every record page shows a ready-made APA citation and a BibTeX entry with copy buttons.

How do you cite a dataset?

In APA 7th edition, a dataset reference takes the form: Author, A. A. (Year). Title of dataset (Version) [Data set]. Publisher. https://doi.org/…

A record on Life of Research that has no DOI is cited the same way, with Life of Research as the publisher and the record's permanent link, such as https://lifeofresearch.com/datasets/ followed by its number, in place of the DOI.

Points that are easy to get wrong:

• Cite the version you used. Datasets are corrected and extended; version 2 may not reproduce results from version 1.
• Cite the dataset, not only the paper. If you used the data, cite the dataset; if you also relied on the analysis, cite the paper too.
• Use the dataset's own identifier. A data DOI or record URL, not the DOI of the article that describes it.
• Subsets: name the files or the subset you used in the text, and cite the whole dataset in the reference list.

DataCite, which registers most data DOIs, recommends citations that include at least the creator, publication year, title, version, publisher, resource type and identifier, the same elements as above.

How do you cite research software?

In APA 7th edition, cite software you used directly as: Author, A. A. (Year). Title of software (Version) [Computer software]. Publisher. URL or DOI. Common tools such as word processors or general statistics packages used without special dependence on them are not usually cited; research software that produced or analysed your results should be.

• Cite the release, not the repository. A GitHub link points at code that keeps changing and can disappear. Cite an archived release with a DOI, such as one archived on Zenodo or deposited here, and give its version number.
• Look for the authors' preference. A CITATION.cff file in the project's repository (GitHub turns it into a "Cite this repository" button), a "How to cite" section in the README, or a built-in command such as R's citation("packagename") tells you exactly what the authors want cited.
• Cite the paper as well if the authors ask for it, but not instead of the software.

How do you cite a preprint?

In APA 7th edition, a preprint is cited as an informally published work: Author, A. A. (Year). Title of preprint [Preprint]. Server or repository name. https://doi.org/…

• Check for a published version. Search the title or the authors; many preprint pages link to the journal version once it exists. If you relied on the peer-reviewed findings, cite the journal article; if you are discussing the preprint itself, cite the preprint.
• Keep the label. The [Preprint] description tells readers the work has not been through peer review at that version.
• arXiv: papers there have identifiers such as arXiv:2401.01234 and DOIs beginning 10.48550/arXiv.; either identifies the work, and the DOI form suits most styles.

How do you write these in BibTeX or a reference manager?

Classic BibTeX has no dataset or software entry type, so these are usually written as @misc with the details in howpublished or note. The biblatex package adds @dataset and @software types; with classic BibTeX styles an unknown type produces a warning and is formatted generically.

@misc{okafor2025river,
  author       = {Okafor, Chidinma N. and Lindqvist, Maja},
  title        = {Hourly river temperature measurements, Upper Tana basin, 2019--2024},
  year         = {2025},
  version      = {2},
  publisher    = {Example Data Repository},
  howpublished = {Data set},
  doi          = {10.5555/hydro.2025.017},
  url          = {https://doi.org/10.5555/hydro.2025.017}
}

Reference managers such as Zotero and Mendeley can save a record directly from its web page by reading the page's citation metadata. Check the item type after saving: a dataset saved as a web page or journal article will be formatted wrongly in your bibliography.

What are the most common citation mistakes?

• Citing a repository's home page instead of the specific record.
• Leaving out the version, so nobody can tell which data or code you used.
• Writing the DOI as doi:10… or as a bare number instead of a full https://doi.org/ link.
• Mentioning data or software only in the acknowledgements or the methods text, where citation counts never see it.
• Citing a preprint for a claim that was changed or removed in the peer-reviewed version.

Related guides