What is research data?
There is as yet no universally accepted definition of what constitutes research data. Here we provide the OECD definition that is most useful from a practical point of view:
Research data are factual records (numerical results, textual records, pictorial and audio material) that are used as primary sources for the purpose of scientific research and are generally accepted in the scientific community as necessary to validate research findings. A research dataset is a systematic, partial representation of a research topic.
This term does not include: laboratory notebooks, preliminary analyses, drafts of scientific papers, plans for future research, peer reviews, communication with colleagues and objects (e.g. laboratory samples, bacterial strains and test animals).
Source: OECD
Research data management (RDM) is an important part of the research process. It covers all activities related to the research data lifecycle: collecting, describing, storing, processing, analysing, archiving and making data accessible.
Research data life cycle
Gaelen Pinnock, CC BY-SA 4.0, via Wikimedia Commons
.svg){kind=link}
The life cycle of research data starts even before the data are generated, with the appropriate designing the research process. The scientific or methodological aspect of data design is described in registration reports registered reports), and administrative in a research data management plan.
The second step of the research data lifecycle is getting or creating. We need to keep very detailed records of all the circumstances in which the data were generated, which is also known as provenance or origin. In step 4, we include provenance information in the metadata that must accompany the data if it is to be reusable by other researchers. You can read more about provenance on DiRROS Data.
The third step of the research data lifecycle is Analysis, possibly in collaboration with other members of the project team. In this step, the data is statistically processed and visualised, and meaning is extracted. Useful guidance on how to visualise your data can be found in educational material from the University of Queensland.
The fourth step of the research data lifecycle is design, storage and long-term preservation. For more information on formatting research data and recommended file formats, see DiRROS Data. It is best to store your data in trusted repositoriesthat have the appropriate infrastructure to ensure the safe, permanent storage and retrievability of research data.
The fifth step of the research data lifecycle is sharing/publication. Data stored in trusted repositories can be found using web browsers such as Google Dataset Search. But it makes even more sense to publish the data in the form of data article, a peer-reviewed scientific publication in its own right. Some journals offer data articles as one of the possible publication formats, but there are also journals that are dedicated to research data only (e.g. Data in Brief published by Elsevier and Scientific Data Springer Nature). For more information on the data papers, please visit DiRROS Data.
The sixth and final step of the research data lifecycle is discovery, re-use and citation research data In addition to data quality, a prerequisite for data retrievability and usability is that they are accompanied by persistent identifiers. The assignment of persistent identifiers is provided by trusted repositories, which also offer default citation formats for datasets in one or more citation styles. If this option is not available, you can use the online tool to help you cite your research data Cite This for Me.
Why is handling research data important?
Good data management is good for us and for others. It helps make our data findable, accessible and reusable. And at the end of a project, it makes it easier to archive and share our datasets.
More specifically, with properly managed research data, we can:
- Ensure reproducibility (provide understandable and reusable data),
- contribute to transparency, quality and reproducibility of research,
- maintain research integrity,
- prevent data loss,
- save time for future research,
- achieve a greater research impact (data can also be cited, reaching a wider professional audience),
- meet the demands of funders and publishers.
Sources:
OECD (2007), OECD Principles and Guidelines for Access to Research Data from Public Funding, https://doi.org/10.1787/9789264034020-en-fr.
Social Science Data Archive (2020), Why open data?. http://adp.fdv.uni-lj.si/usposobi/odprti_podatki/.
Description of the research data
Metadata
Metadata is structured "data about data" that contains minimal information about the provenance or origin of the data (e.g.: persistent identifier (DOI), date of publication, title, author, description, keywords, licence, etc.) and is both human-readable and machine-readable. Data with rich metadata are more likely to be found and reused in the future.
Generic metadata standards (e.g. Dublin Core) or domain-specific metadata standards (e.g. CESSDA for social sciences and humanities data) can be used to produce metadata. Metadata standards and domain-specific policies are recorded by:
If you decide to deposit your research data only in a repository and not publish it as a data article, you will need to write it in the form of a ReadMe text file. An example of a README.txt template can be found on the Cornell University website: https://cornell.app.box.com/v/ReadmeTemplate
More detailed information on metadata can be found at DiRROS Data.
Documentation
Additional documentation helps users of our data to understand and re-use the data. It should be deposited with the research data in a trusted repository.
Examples of supporting documentation:
- laboratory logs and protocol descriptions
- Questionnaires
- code lists, variable definitions, data dictionaries
- Units of measurement
- ontologies, controlled vocabularies
- programming language syntax and software output files
- information on equipment settings and instrument calibration
- database schema, description of file directory structure, naming structure
- reports on methodology
- information on analysis and procedures
- information on the origin of the data extracted or digitised
- software code - does it need to be kept for data reproducibility?