Data Citation on Crossref

Joint statement on research data

Hylke Koers — Tue, 28 Nov 2023 00:00:00 +0000

STM, DataCite, and Crossref are pleased to announce an updated joint statement on research data.

In 2012, DataCite and STM drafted an initial joint statement on the linkability and citability of research data. With nearly 10 million data citations tracked, thousands of repositories adopting data citation best practices, thousands of journals adopting data policies, data availability statements and establishing persistent links between articles and datasets, and the introduction of data policies by an increasing number of funders, there has been significant progress since. It now seems appropriate to focus on providing updated recommendations for the various stakeholders involved in research data sharing.

The premise of the original joint statement still stands: most stakeholders across the spectrum of researchers, funders, librarians and publishers agree about the benefits of making research data available and findable for reuse by others. This improves utility and rigor of the scholarly record. Still, research data sharing is not yet a self-evident step in the research lifecycle. We now have sufficient scholarly communication infrastructure in place to bring about widespread change and believe momentum is building for collective action.

It is in this context that DataCite, a global membership community working with over 2800 repositories around the world, and STM, whose membership consists of over 140 scientific, technical, and medical publishing organisations, are issuing this joint statement. Crossref, a nonprofit open infrastructure with over 18,000 institutional members from 150 countries, joins this call, recognising the need for an amplified focus on data citation. The aim of this statement is to accelerate adoption of best practices and policies, and encourage further development of critical policies in collaboration with a wide group of stakeholders.

Signatories of this statement recommend the following as best practice in research data sharing:

When publishing their results, researchers deposit related research data and outputs in a trustworthy data repository that assigns persistent identifiers (DOIs where available). Researchers link to research data using persistent identifiers.
When using research data created by others, researchers provide attribution by citing the datasets in the reference section using persistent identifiers.
Data repositories enable sharing of research outputs in a FAIR way, including support for metadata quality and completeness.
Publishers set appropriate journal data policies, describing the way in which data is to be shared alongside the published article.
Publishers set instructions for authors to include Data Citations with persistent identifiers in the references section of articles.
Publishers include Data Citations and links to data in Data Availability Statements with persistent identifiers (DOIs where available) in the article metadata registered with Crossref.
In addition to Data Citations, Data Availability Statements (human- and machine-readable) are included in published articles where appropriate.
Repositories and publishers connect articles and datasets through persistent identifier connections in the metadata and reference lists.
Funders and research organisations provide researchers with guidance on open science practices, track compliance with open science policies where possible, and promote and incentivize researchers to openly share, cite and link research data.
Funders, policymaking institutions, publishers and research organisations collaborate towards aligning FAIR research data policies and guidelines.
All stakeholders collaborate in the development of tools, processes, and incentives throughout the research cycle to enable sharing of high-quality research data, making all steps in the process clear, easy and efficient for researchers by providing support and guidance.
Stakeholders responsible for research assessment take into account data sharing and data citation in their reward and recognition system structures.

We, the following signatories shall adopt and promote the relevant best practices laid out above. We hope that our action inspires the community, including researchers, research funders, research institutions, data repositories and publishers, to join us in making it easy for researchers to share, link and cite research data.

Endorse the statement here.

Start citing data now. Not later

Geoffrey Bilder — Thu, 23 Mar 2023 00:00:00 +0000

Recording data citations supports data reuse and aids research integrity and reproducibility. Crossref makes it easy for our members to submit data citations to support the scholarly record.

TL;DR

Citations are essential/core metadata that all members should submit for all articles, conference proceedings, preprints, and books. Submitting data citations to Crossref has long been possible. And it’s easy, you just need to:

Include data citations in the references section as you would for any other citation
Include a DOI or other persistent identifier for the data if it is available - just as you would for any other citation
Submit the references to Crossref through the content registration process as you would for any other record

And your data citations will flow through all the normal processes that Crossref applies to citations. And it will be distributed openly to the community (including DataCite!) via Crossref’s services and APIs. All data citations deposited with Crossref will be exposed in the (soon-to-be launched) Data Citation Corpus.

And then, you can sit back and congratulate yourself for making your publication more useful to researchers who want to be able to reuse the data underlying your publications.

Background

You might ask, “So if submitting Data Citations to Crossref has long been possible, why do you have to write this?”

Historically, authors did not cite data in the way they cited publications. Instead, they would often refer to the data in the main text of the article. This has made it hard to determine what data lay behind the research and/or access the data.

But the research community has increasingly recognized that data is a first-class research output and that we should treat it as such. In short, we should formally cite data.

But because citing data is a comparatively new practice, it has been subject to a lot of new analysis. And unsurprisingly, people analyzing data citation have discovered that there is a lot of nuance to citation of any kind.

There are lots of reasons for citing something. There are lots of internalized conventions for citing things. And there are different conventions for citation for different research objects. And SSH citation practice differs from STEM. And legal citation practices are different from scholarly citation practices. And citation practices even vary by subdiscipline and by journal.

Those who have been looking at what it means to “cite data” have naturally stumbled into a thicket of divergent practices - some of which are historical holdovers, some of which are stylistic preferences, and some of which are clearly adaptations to deal with the specific needs of certain research objects/containers or different disciplines.

The temptation has been to try and rationalize this before extending the practice of citation to data.

“Maybe because data is a distinct record type, we should include the fact that it is a data citation in the citation itself?”

“Maybe because people cite data for different reasons, we should include a typology of citation types in all data citations?”

And so you may hear some people say, “hold off on data citation - we don’t have an optimal way to do it yet, and it can be very complicated.”

But guess what?

We currently don’t label citations to monographs as “citation to monograph.”

And we don’t currently include the reason for citation when we are citing a journal article.

It would be very cool if we did. And it would likely make citations even more useful if we did.

But citations are already useful even without these features. And so, to delay citing data indefinitely because we have an opportunity to improve the act of citation is just perverse. Our community has always opted for progress over perfection.

For one thing - the efforts are not mutually exclusive. We can start citing data with the current limitations of citation practices and simultaneously propose mechanisms for making citation more useful in the future, including new guidelines to deal with the unique issues that citing data poses.

But in the meantime, we will be doing researchers a giant favour if we at least include our imperfect and ambiguous, and unconventional references to data in the references section of an article so that they can be accessed and processed along with all the other imperfect, ambiguous and variant citations that we find so useful.

Some of our members are already doing this. They have been for a long time. And they haven’t found it any more complicated than managing non-data references in the past.

Join them and make your metadata more useful.

Cite data now. Don’t put it off.

And Crossref will continue to work with DataCite and the rest of the community to make the distribution even easier and more useful.

So who is already citing data?

Top 10 members depositing data citations from November-May 2022

(broken down by DOI prefix, which is why you see some publishers listed twice):

Prefix	Member name	Data citations deposited
10.1038	Springer Science and Business Media LLC	7174
10.1016	Elsevier BV	6527
10.1007	Springer Science and Business Media LLC	4748
10.5194	Copernicus GmbH	3017
10.1080	Informa UK Limited	2346
10.1177	SAGE Publications	2082
10.1002	Wiley	2048
10.1111	Wiley	1888
10.1108	Emerald	1876
10.3390	MDPI AG	1827

Top 10 data citations per deposited work

(again, broken down by prefix)

Member name	Prefix	Data citations deposited	Data citations per work
Consortium Erudit	10.7202	580	1.149
SLACK, Inc.	10.3928	462	0.646
S. Karger AG	10.1159	1653	0.532
Proceedings of the National Academy of Sciences	10.1073	973	0.502
American Academy of Pediatrics (AAP)	10.1542	486	0.397
F1000 Research Ltd	10.12688	552	0.341
American Association for the Advancement of Science (AAAS)	10.1126	952	0.317
Springer Science and Business Media LLC	10.1038	7174	0.231
JMIR Publications Inc.	10.2196	864	0.187
American Geophysical Union (AGU)	10.1029	692	0.166

These are for the prefixes with the most data citations deposited (>500 in 6 months) so there might be smaller members doing better than this.

Summaries are great, but I want to see some actual examples!

Here are some examples showing how data is cited by our members:

This eLife article: https://doi.org/10.7554/eLife.26410 cites this dataset in Dryad https://doi.org/10.5061/dryad.854j2.
This Copernicus article: https://doi.org/10.5194/acp-22-7105-2022 cite to this dataset https://doi.org/10.24381/cds.bd0915c6
This Sciendo article: https://doi.org/10.2478/plc-2021-0008 cites this APA-hosted language competence test https://doi.org/10.1037/t15159-000
This De Gruyter article: https://doi.org/10.1515/opth-2020-0160 cites this bibliography at Oxford Bibliographies: https://doi.org/10.1093/OBO/9780195396584-0012

And here are some example API requests for discovering more metadata citations. You can use these API requests as examples and adapt to your own needs.

Doing more with relationships - via Event Data

Martyn Rittman — Fri, 14 May 2021 00:00:00 +0000

Crossref aims to link research together, making related items more findable, increasing transparency, and showing how ideas spread and develop. There are a number of moving parts in this effort: some related to capturing and storing linking information, others to making it available.

By including relationship metadata in Event Data, we are taking a big step to improve the visibility of a large number of links between metadata. We know this is long-promised and we’re pleased that making this valuable metadata available supports a number of important initiatives. We will also be backfilling, so all previously deposited relationships will eventually become available as events. The first step will be to add relationships between items that have DOIs, such as between a research article and a related review report or dataset.

What are relationships?

When members register metadata with us, they have the possibility to identify other works, items, and websites that they know are related. This might be supplementary material or previous versions of a work (especially for preprints and working papers). Equally, identifiers for a protein, gene, or organism used in the research can be included. These are recorded as ‘relationships’ and can be accessed in the same way as the rest of the metadata we hold about registered content.

Some examples

Relationships in the metadata show links to the published article from this bioRxiv preprint. In the Crossref Rest API:

"relation": {
 "is-preprint-of": [
 {
 "id-type": "doi",
 "id": "10.1038/s41467-020-17892-0",
 "asserted-by": "subject"
 }
 ],
 "cites": []
},

And now in Event Data:

"subj": {
 "pid": "https://doi.org/10.1101/2020.05.21.109546",
 "url": "https://doi.org/10.1101/2020.05.21.109546",
 "work_type_id": "posted-content"
},
"obj": {
 "pid": "https://doi.org/10.1038/s41467-020-17892-0",
 "url": "https://doi.org/10.1038/s41467-020-17892-0",
 "method": "doi-literal",
 "verification": "literal",
 "work-type-id": "journal-article"
},

Linking to a dataset in the Dryad Digital Repository by a recent eLife article. In the Crossref metadata:

"relation": {
 "is-supplemented-by": [
 {
 "id-type": "doi",
 "id": "10.5061/dryad.s58qh",
 "asserted-by": "subject"
 }
 ],
 "references": [
 {
 "id-type": "doi",
 "id": "10.5061/dryad.s58qh",
 "asserted-by": "subject"
 }
 ],
 "cites": []
},

And now in Event Data:

"subj": {
 "pid": "https://doi.org/10.7554/elife.19920",
 "url": "https://doi.org/10.7554/elife.19920",
 "work_type_id": "journal-article"
},
"obj": {
 "pid": "https://doi.org/10.5061/dryad.s58qh",
 "url": "https://doi.org/10.5061/dryad.s58qh",
 "method": "doi-literal",
 "verification": "literal",
 "work-type-id": "Dataset"
},

If you are interested in relationships for a single DOI, we still recommend checking the metadata of that record, however Event Data is a great option for looking across multiple records. For example, to check for relationships across a prefix, in a given time period, or for a specific type of relationship.

Data citation

Data citations can be included in data deposits in relationship metadata, usually using the ‘is-supplemented-by’ relationship. By creating an event from each relationship, the links between journal articles and books, and the data they rely on are more visible. This makes the data much easier to locate.

Many datasets have DOIs which are usually recorded with DataCite, meaning you are unlikely to find them via searches of Crossref metadata. Making data citation relationship metadata available in Event Data means it will be available in the same format as citations from datasets to articles (which DataCite sends to Event Data) and citations from articles to datasets from Crossref reference metadata (more to come on this later this year). It also means we will convert this information into Scholix format so that it can be harvested and combined with other sets of Scholix-compliant article/data links. Data citations will therefore be available for the community to identify, share, link and recognise research data. We’re working with initiatives like Make Data Count and STM’s research data program to support the growing uptake of good data citation practices. This is a big step forward in making data citation happen for the community; we have more to do, but Crossref is committed to completing this work as a strategic priority.

What’s next?

In this first stage we are adding relationships that link two objects with a DOI, and later this year we will bring in relationships using other identifiers such as accession numbers and URIs. That will make it more straightforward to ask questions of Event Data such as which organisms have relationships to which works with a DOI.

More info and staying in touch

Find out more about Event Data in our support documentation or check out tickets in the GitLab repo.
Keep informed and ask us anything via our community forum for Event Data discussion

Data Citation on Crossref

Joint statement on research data

Start citing data now. Not later

TL;DR

Background

So who is already citing data?

Top 10 members depositing data citations from November-May 2022

Top 10 data citations per deposited work

Summaries are great, but I want to see some actual examples!

Find all the DOIs that cite Dataset X (identified by DOI)

Find all data citations from Crossref member X (identified by member prefix)

Find papers with supplementary data

Find all data citations to Crossref member X

Find all data citations to DataCite member X

Doing more with relationships - via Event Data

What are relationships?

Some examples

Relationships in the metadata show links to the published article from this bioRxiv preprint. In the Crossref Rest API:

And now in Event Data:

Linking to a dataset in the Dryad Digital Repository by a recent eLife article. In the Crossref metadata:

And now in Event Data:

Data citation

What’s next?

More info and staying in touch