Assessment of Data Fitness for Use RDA/WDS WG Case Statement
WDS/RDA Publishing Data Interest Group
WDS/RDA Certification of Digital Repositories Interest Group
Assessment of Data Fitness for Use
WG Charter
The increasing availability of research data and its evolving role as a first class scientific output in the scholarly communication requires a better understanding of and the possibility to assess data quality, which in turn can be described as conformance of data properties to data usability or fitness for use. These properties are multifaceted and cover various aspects related to data objects, access services, and data management processes such as the level of annotation, curation, peer review, and citability or machine readability of datasets. Moreover, the compliance of a data repository or data center providing datasets - for example with certification requirements - could serve as a useful proxy.
Currently, there is a fairly good understanding on how to certify the quality of a data center / repository as a whole, but there is no generally acknowledged concept for assessment of data usability (or fitness for use) of individual datasets. Some of the properties describing data usability are not available or not transparent to users and requirements for other properties cannot be matched with standards. Furthermore, current certifications and accreditations of data repositories only allow limited conclusions on the re-usability of individual datasets. Thus assessing the fitness for purpose and making a decision whether to reuse a dataset is not straightforward. This situation reduces the chances of shared data being reused and in case of reuse could decrease the reliability of research results.
Firstly, a concept of data fitness requires assessment of quality criteria to be included as well as the weighing of each of those criteria. The process should preferably lead to the development of a corresponding metric. Secondly, we want to find effective ways to expose and communicate this metric, for e.g. by using a labelling or tagging system whereby different usability levels are importantly made explicit.
The proposed working group would work towards the following deliverables:
-
The definition of criteria and procedures for assessment of fitness for use
-
The development of a system of badges/labels communicating fitness for use of individual datasets
Criteria would be used such as:
-
Trustworthiness of the data centers/repositories (such as assessed through existing certifications: DSA-WDS, DIN, ISO 16363 etc.)
-
Data accessibility in terms of discoverability, openness, interoperability etc.
-
Level of curation applied (citability, metadata completeness, data harmonization, machine readability etc.)
Value Proposition
The following stakeholders would benefit:
-
Researchers who deposit data can visibly improve and communicate the quality of their datasets, thereby increasing reuse and citation, which provides researcher with additional metrics showing their productivity.
-
Researchers who reuse data can more easily assess the quality of a dataset and in particular its fitness for their reuse. This makes reuse of data safer and more efficient.
-
Data centers/repositories can offer better quality data publication services - such as more transparent curation - thus increasing the overall usage of services which in turn might lead to improving the facility's financial base.
-
Science publishers can better integrate referenced data into the editorial process and improve the review of articles and related datasets as well as citations and cross-linking of datasets and literature as a result of more transparency about data usability.
-
Funders can make provisions for funded data archiving and publication services in accordance with their funding requirements and expectations in terms of data fitness for use (and reuse).
Overall impacts:
-
Improved and standardized data publication services
-
Improved communication of data fitness for use
-
Improved reliability and efficiency in the reuse of research data
Engagement with existing work in the area
Data fitness for use has been addressed in literature over the last 20 years. The topic received more attention with the general increase of data production. The following gives a brief overview of selected publications. It is by no means exhaustive. In 1998 Tayi and Ballou stated that the concept of data quality is relative with quality being dependant on users and applications. Some authors concentrated on special aspects as for example assessment of accuracy of geospatial data (de Bruin 2001) or de-duplication relevant for example to data mining approaches (Christen & Goiser 2007). A further aspect is preservation of usability of sensitive data (Bhumiratana & Bishop 2009). In 2007, the OECD underlined the importance of efficiency in reusing data (OECD 2007). For example efficient compilations of data from multiple providers require harmonized and machine readable data, in particular for data with high volumes. Correspondingly, the FAIR Data Publishing group supplies a set of principles for publishing data and emphasizes machine readability of data as one of the major challenges (Wilkinson et al. 2016). More recently authors also started to investigate data usability with respect to big data approaches (Jianzhong 2013). The effect of peer-review on data quality, respectively usability was stressed by Lawrence et al (Lawrence 2011) and an editorial in the Nature Scientific Data Journal (2016). Costello linked data fitness for use with the data publication concept (Costello 2013). Also worthy to note is the ISO/IEC 25012 data quality model (ISO/IEC 2008) and the ISO 8000 Requirements for Quality Data (ISO 2009). The W3C Data on the Web Best Practices Working Group elaborated vocabularies needed to describe data quality and highlights the importance of data provenance (W3C 2016), which – if applicable — should include also detailed information about physical samples, for example in the case of biocollections (Bishop 2016). Finally, fitness for use of datasets should be transparent and comprehensive to users. The effectiveness of using badges or labels for this purpose was shown by Kidwell et al (Kidwell 2016).
In addition to works published in the literature, the WG can build on a wide range of activities that are relevant to the aims and scope of the group. In particular:
-
The Working Group would operate under the umbrella of the RDA-WDS Data Publishing IG and RDA/WDS Certification of Digital Repositories IG
-
This Working Group will follow up on the work of the RDA/WDS Data Publishing Workflows WG and assess the impact of workflows on fitness for use (Austin et al. 2016)
-
This Working Group will follow up on the work of the Repository Audit and Certification DSA–WDS Partnership WG and develop a related certification system for individual datasets
-
The Working Group would incorporate the criteria defined by the FAIR working
Group (Wilkinson 2016) as a starting point. -
The Working Group will collaborate with the NIH Commons FAIR metrics group to elaborate on the FAIR criteria (NIH 2016)
-
This Working Group would incorporate the W3C data quality vocabulary to define quality processes (W3C 2016).
Work Plan
Work will be along four strands:
-
Descriptions and definitions of data fitness criteria. In a first step we will gather literature and initiatives having addressed the topic. To sort out ambiguities of term definitions relevant to this group, we will collaborate with the CODATA/CASRAI development of an International Research Data Management glossary (IRiDiuM) and maintain consistency with terms in the RDA Term Definition Tool (TeD-T). The selection of data fitness criteria will be set out to the wider community before finalizing the document.
-
Development of a fitness for use label at the level of datasets
-
Conceptual model
-
Selection and evaluation/weighing of criteria with respect to the different aspect of fitness for use such as curation or accessibility
-
Considerations for adoption by stakeholders (archives/repositories: for e.g built into workflows, science publishers)
-
-
Design of label/badge
-
-
Development of service components
-
Investigate how a fitness of use concept can be integrated into current certification procedures for data centers/repositories (WDS/DSA)
-
Investigate data centers/repositories service components
-
Setup of a testbed of several data centers/repositories
-
-
Governance and sustainability:
-
Concept for a long-term organizational structure to operate elaborated services successfully and in a way that meets the needs of all stakeholder groups. This stream will also deliver a process through which new organizations can connect to the service.
-
Deliverables
-
Addition or revision of relevant terms in the IRiDiuM glossary (CODATA/CASRAI)
-
Document defining fitness for use criteria
-
Description and design of fitness for use label (badge system)
-
Concept for a certification procedure including the fitness for use aspect
-
Concept for a data centers/repositories service components
-
Adoption plan including certifying organizations and governance
-
Manuscript for submission to a peer-reviewed journal.
Milestones
-
Fitness for use concept ready
-
Setup of a testbed with several data centers/repositories and science publishers
-
Prototype of fitness for use label available
Mode & frequency of operation
-
Telecons every 4 weeks
-
Face to face meetings during RDA plenaries and at least one additional workshop. RDA plenaries in particular will be used to engage the wider community and coordinate the work with related groups.
-
Additional meetings of subgroups working on particular deliverables including adoption
Timeline
Months |
Action |
Deliverable |
April - July 2017 |
Terminology & definition of criteria |
Overview of criteria, for discussion at 9th plenary meeting |
July - December 2017 |
Pilot assessment of criteria |
Report on outcomes of pilot, for discussion at 10th plenary meeting |
December - February 2017 |
Development/design of badge system and integration with current certification schemes |
Guide for repositories |
February - August 2018 |
Concept for integration of data repository service components. Piloting Integration of badge system. |
Governance structure and adoption plan |
May 2017 - October 2018 |
Draft article for peer review |
Submission of article to a peer-reviewed Journal. |
Adoption Plan
Members of the proposed working group are planning to carry out a pilot during the 12-18 month timeframe in which they incorporate the insights that come out of the working group. In this pilot, a first assessment of the fitness for use of individual datasets will be carried out. This simultaneous pilot will provide the working group with important information about both benefits of and challenges with adoption which will make it easier for additional organizations to adopt the outcomes of the working group. The goal is that at the end of the 18 month timeframe, a first network of adopters will exist.
Initial Membership
Claire Austin (Research Data Canada, Co-Chair, claire.austin@gmail.com )
Bradley Wade Bishop (Univ. Tennessee)
Helena Cousijn (Elsevier, Co-Chair, h.cousijn@elsevier.com )
Michael Diepenbroek (PANGAEA, Co-Chair, mdiepenbroek@pangaea.de )
Amy Nurnberger (Columbia University Libraries)
Ingrid Dillo (DANS)
Stephane Pesant (MARUM)
Mustapha Mokrane (ICSU-WDS)
Markus Stocker (PANGAEA)
Rob Hooft (DTL)
Peter Doorn (DANS)
Christina Lohr (Elsevier)
Robert R. Downs (CIESIN, Columbia University)
Daniel Fowler (Open Knowledge International)
Martina Stockhause (WDC Climate, DKRZ)
Ian Bruno (CCDC)
Tim Smith (CERN/Zenodo)
Donna Scott (NSIDC)
Jonathan Petters (Virginia Tech)
Kathleen Gregory (DANS)
References
Austin CC, *Bloom T , *Dallmeier-Tiessen S, Khodiyar V, Murphy F, Nurnberger A, Raymond L, Stockhause M, Tedds J, Vardigan M, & Whyte A (2016). Key components of data publishing: Using current best practices to develop a reference model for data publishing. International Journal on Digital Libraries (IJDL), Research Data Publishing Special Issue. Pages 1-16. DOI 10.1007/s00799-016-0178-2
Bhumiratana B & Bishop M (2009) Privacy aware data sharing: balancing the usability and privacy of datasets, in: Proceedings of the 2nd International Conference on PErvasive Technologies Related to Assistive Environments, https://doi.org/10.1145/1579114.1579187
Bishop, B. W. & Hank, C. F. (2016) Fitness for Use in Data Curation Profiles for Biocollections [Presentation] American Society for Information Science and Technology Annual Meeting, October 2016, Copenhagen, Denmark
de Bruin S, Bregt A, van de Ven M (2001) Assessing fitness for use: the expected value of spatial data sets, International Journal of Geographical Information Science, v15, no5, p457-471
Christen P & Goiser K (2007) Quality and Complexity Measures for Data Linkage and Deduplication, in: Guillot FC & Hamilton HJ (eds) Quality Measures in Data Mining, Studies in Computational Intelligence pp 127-151
Costello M et al (2013) Biodiversity data should be published, cited, and peer reviewed, Trends in Ecology & Evolution, p1-8
International Renewable Energy Agency (2013) Data quality for the Global Renewable Energy Atlas – Solar and Wind, https://goo.gl/a8xr1Q
ISO (2009ff) Data quality, https://en.wikipedia.org/wiki/ISO_8000
ISO/IEC (2008) Data quality model, http://www.iso.org/iso/catalogue_detail.htm?csnumber=35736
Kidwell MC, Lazarević LB, Baranski E, Hardwicke TE, Piechowski S, Falkenberg L-S, et al. (2016) Badges to Acknowledge Open Practices: A Simple, Low-Cost, Effective Method for Increasing Transparency. PLoS Biol 14(5): e1002456. http://doi.org/10.1371/journal.pbio.1002456
Lawrence, B., Jones, C., Matthews, B., Pepler, S. & Callaghan, S. (2011). Citation and Peer Review of Data: Moving Towards Formal Data Publication. International Journal of Digital Curation 6, 4–37
Li Jianzhong & Liu Xianmin (2013) An important aspect of big data: data usability, Journal of Computer Research and Development, v6
NIH Commons FAIR metrics group (2016) WG interim report, https://goo.gl/n4PpWv
OECD (2007) OECD Principles and Guidelines for Access to Research Data from Public Funding, http://www.oecd.org/sti/sci-tech/38500813.pdf
Scientific Data Journal (2016) Let referees see the data, editorial, Nature Scientific Data Journal, 3, 160033. http://doi.org/10.1038/sdata.2016.33
Tayi GK & Ballou DP (1998) Examining data quality, Communications of the ACM, v41, no2, p54-57
W3C (2016) Data on the Web Best Practices: Data Quality Vocabulary, W3C Working Group Note, https://www.w3.org/TR/vocab-dqv/#mapping-ISOZaveri
Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., Appleton, G., Axton, M., and Baak, A. et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Sci. Data 3:160018, http://doi.org/10.1038/sdata.2016.18
- Log in to post comments
- 15897 reads
Author: Siri Jodha Khalsa
Date: 08 Dec, 2016
The case statement describes well the importance of this topic, but the terminology surrounding the issue of data quality needs clarification and the case statement doesn't help much. I would contend that data quality, data usability, fitness for use, and re-usability are not all equivalent, but they seem to be used interchangeably in the statement. Moreover, I would say that certification of a repository has little direct bearing on fitness for use, in the way that I understand it. There are many factors going into my assessing whether a given dataset is fit for my purposes. Data may be of high accuracy, but not at the time resolution or latency that I need it. Conversely, data may be of less than optimal quality (e.g. crowd sourced) but still very useful to me because it provides information that I can quality check myself based on other data that I possess or might be the only source of much needed information. Thus fitness is highly dependent on the intended purpose, which will vary from user to user. While I believe that a repository can do much to gauge user needs and make recommendations based on that understanding, I do not believe that compliance of a data repository is, in itself, a useful proxy for data fitness for use. Thus, I recommend that the case statement be revised to include as the first deliverable of the working group to be a definition of terms, keeping these as tight and distinct as possible, with reference to established standards and vocabularies, such as those from ISO and W3C. I also think, but could be convinced otherwise, that trustworthiness of repositories not be conflated with data quality.
Author: Peter Doorn
Date: 16 Jan, 2017
A couple of definitions is given in the (draft) Science Europe Data Glossary, see: http://sedataglossary.shoutwiki.com/wiki/Main_Page
Author: Peter Doorn
Date: 16 Jan, 2017
You write "I also think, but could be convinced otherwise, that trustworthiness of repositories not be conflated with data quality". Conflate no, but TDRs often play an important role as "gatekeeper" of data quality, for instance by maintaining quality control of metadata, identifiers, data formats, access licenses, etc. This is how a trustworthy repository does play a generic role in guaranteeing a certain degree of data quality. Of course this does not guarantee that any individual dataset is fit for a particular use to a certain user!
Author: Jonathan Petters
Date: 19 Dec, 2016
Enjoyed reading the case statement, and I fully support a system by which prospective users of a dataset can help themselves understand how useful that dataset may be for their purposes before they go down the wrong rabbit hole and waste research resources! If such a system was compatible with our Hydra-Fedora-based repository I’d work to get it incorporated.
As Dr. Khalsa and the case statement note, the usefulness of a dataset is strongly dependent on the prospective user and what they are trying to accomplish. However, my reading of the case statement suggests a ‘data depositor/sharer’ perspective on making data usable, and not from the prospective user’s perspective. If true, be sure to make that clear. I’ve been partial to Tim Berners-Lee’s 5 star deployment scheme for open data as one way to communicate some of this perspective.
On the other hand, one could envision a system by which users of a dataset comment on and rate how useful a dataset was for their particular purposes. This could emulate the five-star rating systems we see on Amazon.com, Yelp, etc. I understand this may be out of scope, but a data depositor/sharer perspective on data ‘fitness’ or ‘quality’ might be able to co-exist with such a data user rating/commenting system.
Author: Peter Doorn
Date: 16 Jan, 2017
DANS is working exactly on two of the comments you made on our case statement: the Tim Berners-Lee’s 5 star deployment scheme for open data, and the rating of datasets by (re-)users. See a recent webinar on implementing the FAIR principles in the context of a trusted digital repository: https://eudat.eu/events/webinar/fair-data-in-trustworthy-data-repositori...
Author: Michael Diepenbroek
Date: 06 Jan, 2017
Dear Siri and Jonathan,
thanks for your comments. Please, keep in mind that we do not present a solution in the case statement, this is the work to be done, in particular clarifying concepts and terminologies (Siri, you are right, terminology should be the first deliverable). As for some special points:
- yes, the concept of data quality is relative with quality being dependant on users and applications (in the statement) - important is that users are able to judge whether the quality fits their application. This needs to be communicated to users.
- we missed Tim Berners-Lee’s 5 star deployment scheme, but will be happy to include it into our discussions.
- commenting and rating a data set (social tagging) could be a criterion to include, however, there are transitions to altmetrics - needs to be clarified.
- certification as part of the concept: Archiving and publication workflows are not transparent to users. Workflows are related to the repository/data center as a whole. Quality of such workflows are recognized in current certification schemata (e.g. WDS), however, somehow blurred amongst other criteria. Also to be discussed.
Happy to see you in Barcelona for further discussions
Helena & Michael