How to Cite a Dataset Correctly

Data is a first-class research output

For a long time, the dataset behind a paper was invisible — mentioned in passing, if at all. That is changing fast. Datasets are now recognised as citable research outputs in their own right, and citing them properly is both good scholarship and a way to give data creators the credit they have earned.

The standard elements of a data citation

A data citation contains the same core building blocks as any reference: 1. Creator(s) — the individuals or organisation responsible for the dataset. 2. Publication year — when the dataset was released or last versioned. 3. Title — the name of the dataset. 4. Publisher / repository — where it is archived. 5. Version — datasets change; cite the exact version you used. 6. Persistent identifier — the DOI, the single most important element. A typical format looks like:

Creator (Year): Title. Repository. Version. https://doi.org/10.xxxx/xxxxx

Why the DOI is non-negotiable

The DOI is what makes a data citation work. It resolves to the exact dataset, survives moved files and renamed servers, and lets indexing services link the citation back to the data. A data citation without a persistent identifier is a citation that will eventually break.

Cite the version you actually used

Because datasets are often updated, a reference to "the dataset" is ambiguous. Good repositories mint a version-specific identifier, so cite the precise version your analysis relied on. This is essential for reproducibility — someone repeating your work needs the same data, not a later revision.

Why it matters beyond good manners

• Credit. Citations are how research contributions are counted; citing data ensures data creators are recognised.
• Reproducibility. A precise data citation lets others retrace your steps.
• Discovery. Citation links help others find the data through the papers that use it.

Whenever your work rests on someone else's data — or your own — cite it fully, with its DOI and version. It costs one line and it moves the whole field toward taking data seriously.