Choose between citing a published statistic and a dataset you analysed
Decide whether an APA citation should identify a statistic you read in a publication or the exact dataset version you used for your own analysis.
About 4 min read · Allow extra time for the practical checks
In APA 7, cite the publication when you report a statistic you read there. Cite the dataset when it is the source of a secondary analysis you actually conducted. Begin with what supports the sentence, rather than choosing whichever reference looks more authoritative.
APA’s dataset reference guidance makes this distinction. Citing data alone can misrepresent your access if you only read a report about it.
Trace the sentence to your actual work
Identify whether you read a reported result or calculated it from data. Resolve that distinction before formatting the reference.
The following situations are wholly fictional. The institution, report and dataset are invented; no real statistics or identifiers are supplied.
| Your actual work | What the sentence represents | Source to identify |
|---|---|---|
| You read a median waiting time in the Harbour Service Review | A result reported by its authors | The review and relevant table or page |
| You downloaded the Harbour Requests dataset and calculated a median for a defined subset | Your own calculation from archived records | Exact dataset version and retained analysis record |
| You compared your calculation with a number in the review | Two results from different work | Both sources, with their roles distinguished |
| You found a chart in a news article but have not located its stated source | A number encountered indirectly | An unresolved access trail; check the underlying publication before claiming direct use |
This is a decision aid, not a completed reference list. It omits invented authors, years and DOIs. Those fields must come from the real material you used.
Write a working note such as “reported in the review, Table 2” or “calculated from downloaded version 2, retained records only.” Merely writing “source: dataset” leaves the number’s origin unclear.
Keep the dataset version attached to the calculation
Collect the creators, title, version, publication year, repository and persistent identifier from the dataset record. APA uses the publication year of the version used. Its example italicises the title and separates version details, a bracketed description and the repository.
DataFirst’s guidance also emphasises the exact version and checking the recommended citation. Its displayed format is a repository convention, not a universal APA template.
Keep four separate fields in your working record:
- Dataset identity and version supplied by the repository.
- File actually downloaded and access date.
- Variables and subset used in the calculation.
- Saved analysis output supporting the sentence.
The first supports source identification. The others document your workflow; they do not all automatically belong in the reference entry. Keep detailed coding decisions in your data dictionary.
Suppose your file is named “requests-final.csv” but the repository advertises a newer release. Do not update the reference version merely to make it look current. Establish which release produced your result. If you cannot, mark the record unresolved and inspect download history or retained metadata.
Make the attribution visible
In the fictional example, “The Harbour Service Review reports…” identifies a published finding. “We calculated… using…” identifies work undertaken by the current writers. Complete either sentence with checked details before using it in a manuscript.
When both a report and a dataset informed your work, explain their separate roles. A data citation should not silently replace the report supplying an interpretation. Equally, citing a report does not document your own calculation from a different version or subset.
If results disagree, compare the population, period, inclusion criteria, missing-data treatment and version in your records. This guide does not select a statistical method or decide which estimate is correct. Changing a citation cannot repair an unsupported explanation.
Audit the source trail
Ask whether a reader could distinguish the material you consulted from the calculation you conducted. Check reference details against that material, then match citations with the reference list.
Do not invent an identifier when none is supplied. Citation formatting does not establish permission to reuse restricted data; check access conditions separately. Where your access was indirect, resolve that trail before presenting the original as something you read.
A reported number should stay attached to its publication, and your own analysis to the version of the data that produced it.
