Repository Citation

A repository citation is how a dataset, software release, or other non-article research output deposited in a repository gets attributed and located in scholarly text — built around a persistent identifier rather than the journal/volume/page structure used for a journal article. In practice, for most research data this means a citation anchored on a DataCite DOI, the identifier scheme minted by DataCite, the DOI registration agency built specifically for research data, software, and other non-textual outputs (as distinct from Crossref, which registers DOIs primarily for journal articles, books, and conference proceedings).

How it differs from a journal article citation

A journal article citation identifies a publication through bibliographic fields a citation style (APA, Vancouver, etc.) expects: authors, year, article title, journal name, volume, issue, pages. A repository citation identifies a deposited data object instead, and the fields that matter are different: who created the dataset, when it was published (deposited and released, not necessarily when any accompanying paper appeared), what the dataset is called, which repository holds and is responsible for it, what kind of object it is, and — critically — the persistent identifier that resolves directly to it. A dataset frequently has its own citation independent of, and in addition to, any journal article that discusses or analyzes it; citing only the article is not the same as citing the data.

The DataCite citation format

DataCite documents a recommended (encouraged, not mandated) display format built directly from its own metadata schema‘s mandatory properties:

Basic format: Creator (PublicationYear): Title. Publisher. Identifier

Enhanced format, adding optional properties when they matter to the reader (Version is particularly important for citing software, since two versions of the same software can behave differently): Creator (PublicationYear): Title. Version. Publisher. (ResourceTypeGeneral). Identifier

DataCite also recommends that the DOI be displayed as a resolvable, linkable URL (https://doi.org/10.xxxx/...) rather than the bare DOI string, since that is what actually takes a reader to the dataset.

The elements behind the format

The format above is not arbitrary — it maps directly onto the properties DataCite’s Metadata Schema requires a repository to supply before it can mint a DOI for a deposit at all. The six mandatory properties are Identifier, Creator, Title, Publisher, PublicationYear, and ResourceType. A repository cannot register a DataCite DOI without all six, which is why a citation built from them is always constructible for any DataCite-registered dataset. Recommended (not mandatory) properties — Subject, Contributor, Date, RelatedIdentifier, Description, GeoLocation, Version — add useful context but are not required for the citation itself to be complete.

Why the dataset DOI matters

The persistent identifier is what makes a repository citation actionable rather than merely descriptive: unlike a plain URL, a DOI is designed to keep resolving to the dataset (or a documented, versioned successor) even if the repository migrates platforms or reorganizes its file structure. This is also the mechanism behind data-citation tracking — counting how often a dataset itself, not just an associated paper, has been cited — and behind linking a dataset to related outputs (the paper that uses it, prior versions, associated software) via DataCite’s RelatedIdentifier property. See also CASRAI’s guide to persistent identifiers and citation for AI models and training datasets for how the same identifier logic extends to newer output types.

Relationship to the data citation principles

Repository citation is the practical, format-level answer to the broader commitments set out in the Joint Declaration of Data Citation Principles and the general data citation principle — that data should be a first-class, citable product of research in its own right. Those principles establish why data should be cited; DataCite’s schema and citation format establish how.

Frequently asked

Do I need a special citation style to cite a repository dataset?

No — most citation styles (APA 7th, Vancouver, Chicago) now have explicit guidance for citing datasets, and all of them expect the same underlying elements DataCite’s format uses: creator, year, title, repository/publisher, and identifier. The style just governs punctuation and ordering, not which facts you need.

What if the dataset has no DOI?

Cite it with whatever persistent identifier the repository does provide (a Handle, an ARK, or in weaker cases a stable accession number), and prefer repositories that mint DOIs or another recognized persistent identifier where a choice of deposit location exists — a plain URL is not considered persistent and is a weaker basis for a repository citation.

Is a repository citation the same as a data availability statement?

No. A data management plan or manuscript’s data availability statement typically states where and under what conditions the data can be accessed; a repository citation is the formal bibliographic reference to that dataset, and the two commonly appear together — the availability statement pointing to the data, the citation providing the reference a reader would use to cite it themselves.