How to Cite a Dataset in APA: Version and Source

To cite a dataset in APA, identify the data you actually analyzed and the version you used, then collect its creator, publication year, title, repository and persistent link. If you only repeat a statistic from a report or article, cite that publication instead. A repository homepage alone may not identify your data.

By plagiarism-checker-online.net Editorial Team | Published October 9, 2026

You downloaded a spreadsheet from a public repository, cleaned it and ran your analysis. Then the reference list needs an entry, and the quickest option is to cite the paper that introduced the data or paste the repository's homepage. Either choice breaks the link between your results and the file you actually used. A dataset reference exists to fix that link.

APA writes “data set” as two words in its own guidance, and the style treats it as a source type with its own format. This guide follows APA 7, shows how to read a repository record, and repairs three common mistakes. It's a citation guide, not a research-methods course.

Dataset, codebook or published statistic?

The APA Style data set reference page sets the boundary. Provide a data set citation when you conducted secondary analyses of publicly archived data, or when you're archiving your own data and presenting it for the first time. If you're citing existing data or statistics, cite the publication in which they appeared, such as a journal article, report or webpage, rather than the data set itself.

A three-case source decision table

If a codebook is a separate document with its own record, look at that record before deciding. The classic error runs the other way: someone analyzes a file and then cites only a summary article, so nobody can tell which release produced the numbers.

Find the exact data version

DataCite's data citation guidance explains why this matters. Citing data supports access, transparency and reproducibility, gives researchers credit and makes the repository visible. DataCite isn't an APA authority, so use it for the reason, not the format.

Repository landing page

Open the page for your specific study or file, not the repository's front page. Look for a citation box or “cite this” section, the creator names, the publication year, the title, the version label, the repository name and a persistent link. A homepage address such as a repository's main domain tells readers where to start looking, not what you used.

DOI, version and download record

Write down four things on the day you download: the DOI or persistent URL, the version label, the date and the file name. Repositories update data, and a later release may differ from the one you analyzed. This record goes in your methods notes. It isn't part of the reference itself.

Build an APA reference from verified metadata

APA's page gives the pattern. Start with the creator, then the year in parentheses, then the title in italics. After the title, put any numerical identifier and the version in parentheses, not italicized and separated by a semicolon. Then comes a bracketed description such as [Data set], then the publisher, then the DOI or URL. Here's where each element comes from.

Worked example: metadata to generated reference

One verifiable archived dataset

APA's own example is a real ICPSR study. We checked its public DOI record on DataCite on October 9, 2026. ICPSR's landing page itself wouldn't load for our automated tools, so check the repository page before you copy anything for real work. The record lists the creator William O'Donohue, the title “Content Analysis of Undergraduate Psychology Textbooks,” a 2017 publication year, the publisher Inter-University Consortium for Political and Social Research, the DOI 10.3886/ICPSR36966.V1 and an ICPSR study number of 36966, with the version shown as Version 1.

APA's published reference for that record reads: O'Donohue, W. (2017). Content analysis of undergraduate psychology textbooks (ICPSR 36966; Version V1) [Data set]. ICPSR. https://doi.org/10.3886/ICPSR36966.v1

Notice what APA did. It shortened the publisher to ICPSR, wrote the title in sentence case, and joined the study number and version with a semicolon inside one set of parentheses. We ran the same metadata through the pinned citation engine behind our Citation Generator using its dataset type, and it printed: O’Donohue, W. (2017). Content analysis of undergraduate psychology textbooks (Version V1) [Dataset]. ICPSR. https://doi.org/10.3886/ICPSR36966.v1

That's close, but not identical. The generator's dataset type has a version field and no slot for a study number, and it prints [Dataset] where APA's page has [Data set]. So the study number, and the bracket wording, are yours to check against APA's page before you paste the entry into a paper. A generated reference is a draft.

One more difference turned up. A university library guide we read (USC Libraries, updated June 2026) shows a similar ICPSR example with a comma between the study number and the version, where APA's page uses a semicolon. When a campus guide and APA's own page disagree, follow APA's page and your instructor's preference.

A clearly fictional missing-field scenario

Fictional example: a city open-data portal offers a daily bike-count table. Everything below is invented, including the organization and the address.

From that record the generator's dataset type prints: Riverside Transportation Office. (2025). Riverside bike counts, daily totals [Dataset]. Riverside Open Data Portal. https://data.example.org/riverside/bike-counts

You'd change [Dataset] to [Data set] to match APA. For a portal table that changes over time, USC Libraries also calls for a retrieval date in the reference, in the form “Retrieved Month day, year, from URL.” The generator's dataset type doesn't print that element, so add it manually using your actual retrieval date. Keep the downloaded file and record its name and download date in your methods notes too. For an archived, versioned file, those methods notes identify your working copy without adding a retrieval date to the reference.

Explain your analysis separately from the citation

A citation says what you used. Your methods section says what you did with it. Keep the two apart, and don't try to squeeze file names, filters and cleaning steps into the reference. A sentence like this works, with every detail swapped for your own: “We analyzed the public file titled [title] ([creator], [year], version [label]), downloaded on [date], and kept the [number] records with complete [variables].”

That sentence ties the reference to your file without pretending the repository vouches for your choices. If you keep an annotated bibliography, our guide to annotated bibliographies covers what to write when your source is a dataset rather than a summary article.

Repair three references that lose reproducibility

Homepage only

Problem: the reference list gives the repository's main address and nothing else. A reader can't tell which study or release you used. Repair: go to the record for your study and rebuild the entry with its creator, year, title, version and DOI.

Wrong version

Problem: you analyzed an older release, but the citation copies the newest record. The year and version no longer match your numbers. Repair: find the version you actually downloaded, using your download notes, and cite that one. A fictional version-2 portal file would print as: Riverside Transportation Office. (2025). Riverside bike counts, daily totals (Version 2) [Dataset]. Riverside Open Data Portal. https://data.example.org/riverside/bike-counts

Article substituted for analyzed data

Problem: you ran your own analysis, but the reference list shows only the article that describes the data. Readers will think you reproduced the article's numbers. Repair: cite the data set you analyzed. Keep the article too, if you relied on its documentation or findings, so the credit is complete.

Final data-source checklist

1. Decide whether you analyzed the data or only repeated a published number.
2. Open the record for your specific study, not the repository's front page.
3. Copy the creator, year, title, version, publisher and DOI from the record.
4. Check that the year and version match the file you used.
5. Write [Data set] or another fitting description in brackets.
6. Compare any generated or recommended citation with APA's page.
7. Save the download date, file name and version in your methods notes.
8. Make sure the text of the paper says what you did with the data.

Once the data references are right, the rest of your references follow the same logic. Our guide to citing properly covers the APA basics, and the citation and referencing guide shows where datasets fit. A text comparison can help with the prose in your draft through the plagiarism checker, but it can't verify your data, calculations, permissions or reproducibility.

FAQ