DATA SHARING
Data sharing is increasingly viewed as an essential step in improving research transparency and reproducibility (Taichman et al, 2016; Vickers, 2006). There has been a lot of discussion on the imperative for data sharing in the biomedical arena, particularly of publicly funded research. As a result, there are many disciplines where proposals for data sharing are being discussed.
Publishers, including PLOS and the BMJ Publishing Group, that have implemented data sharing requirements have found that it is not trivial to do. A recent post on the Scholarly Kitchen (http://scholarlykitchen.sspnet.org/2016/01/13/what-price-progress-the-costs-of-an-effective-data-publishing-policy/) discussed the costs and workforce issues that could result from making data sharing a requirement for publication of research. Some of the questions raised relate to availability of data archives and infrastructure outside of publishers, and the enforcement policies of journals to assure compliance with data archiving and sharing. If there is no enforcement of a data sharing policy, and the infrastructure to support data sharing is lacking (it is currently patchy, although well developed in Australia, for example) and the editors do not have a policy of peer review of the data prior to acceptance of the paper, how will requiring data sharing actually improve the integrity of the research? Other questions include how long should data remain available, who should manage the availability and sharing of the data, and how much will these requirements cost? In addition, there is a need to ensure that those whose data are reused get adequate credit—something that is not routinely done, but which groups such as Force 11 and others (https://www.force11.org/group/joint-declaration-data-citation-principles-final) are working towards. Most recently, questions about the legitimate requests for and re-use of data have been explored systematically and thoughtfully by Lewandowsky & Bishop (2016).
In light of the above concerns about implementing a data sharing policy, COPE invites discussion on this topic, specifically relating to the following questions:
- Should researchers be required now to make their data available as a condition of publishing?
- Who should disseminate guidelines and/or monitor data sharing practices?
- What issues surround the re-use of published data?
- Should data deposited by authors be subject to peer review? Prior to publication? To settle disputing claims about results? For use in systematic reviews and meta-analyses?
- What best practices are already used by journals, publishers and data repositories that could be adapted for use by others considering data sharing requirements?
- Since past practices do not often enable data sharing in any easy way, should there be an ‘amnesty’ for old work, but stricter standards applied to work now being done?
References
Lewandowsky S, Bishop D. (2016). Research integrity: Don’t let transparency damage science. Nature, 25 January, vol 529; http://www.nature.com/news/research-integrity-don-t-let-transparency-damage-science-1.19219
Taichman DB, et al. (2016). Sharing clinical trial data—A proposal from the International Committee of Medical Journal Editors. New England Journal of Medicine, January 20. DOI: 10.1056/NEJMe1515172
Vickers A J. (2006). Whose data set is it anyway? Sharing raw data from randomized trials. Trials 7; 15. DOI: 10.1186/1745-6215-7-15
Wager, E. (2016). Sharing data is a good thing. But we need to consider the costs. Retraction Watch, 28 January 2016; http://retractionwatch.com/2016/01/28/sharing-data-is-a-good-thing-but-we-need-to-consider-the-costs/
Comments
Without data and code, replicability is a farce. Hence, data and code should be available as much as possible. Arguments about effort are nonsense. If your data are sufficiently organized for analysis, then it is a small additional effort to upload the files to some repository.
I disagree with Lewandowsky and Bishop. Irony apart -- Lewandowsky has a reputation for hiding sloppy research, Bishop played a small but key role in the harassing of Tim Hunt -- they argue for reduced transparency so as to protect researchers against naughty outsiders. This does not work. If they want to beat you, they will find a stick. Hiding your data just hands them a bigger stick. At the same time, reduced transparency is effective in protecting naughty researchers.
to post comments
"Most recently, questions about the legitimate requests for and re-use of data have been explored systematically and thoughtfully by Lewandowsky & Bishop (2016)."
This is a jaw-dropping and quite worrying remark from the anonymous COPE representative who wrote this piece.
If this is the view of COPE, there is no point in taking part in your Webinar.
Please read the comments under the Nature article, and the following blog posts by four academics:
Nicole Janz
Getting the idea of transparency all wrong
Judith Curry
Violating the norms and ethos of science
James Coyne
Further insights into the war against data sharing: the Science Media Centre's letter writing campaign to UK parliament
and myself
Nature on research integrity?
(links removed as they triggered the sp@m trap)
to post comments
1. I would distinguish between data source and data---especially important in qualitative research. The data are what is used to support claims made, whereas a video or transcript is a data source, and different pieces may be extracted to support very different claims.
2. Multiple uses of large data sets are already common in the statistical analysis of PISA and other social science data. In any case, use of such data sets, at least in Canada, has to be approved by the local Research Ethics Board.
3. Provisions need to be made during recruitment and informed consent that data, once published, can no longer be withdrawn, as is current practice ("may withdraw at any time")
to post comments
Three issues; First, what constitutes a 'clinical trial" - - only publicly funded trials? Many residents and fellows undertake trials which could be construed as clinical (comparison groups, interventions, specific outcome measures). These trials are often small in number making potential patient identification more problematic. Not certain that these trials could even be blinded sufficiently to protect patients.
Second, who 'owns' the data? What prevents another individual from making a career wading through these public data sets publishing like crazy using the data inappropriately in manners in which the data was never intended?
Third, what it the time limit for sharing data. Some data will clearly have a shelf life due to advancements in the field and will become out-of-date or otherwise inappropriate. What prevents someone from using older data?
to post comments
I appreciate the comment of Wolff-Michael Roth about the distinction between data source and data but I think there remains a lack of confidence on the part of many in the data source for some qualitative research. For example, Carlos Castaneda received his PhD in Anthropology from UCLA based on "field notes" from his experiences with shamanism and peyote and went on to write numerous "non-fiction" books about the Yaqui Shaman Don Juan Matus (https://en.wikipedia.org/wiki/Carlos_Castaneda). Critics largely claim Castaneda's work was fiction not an ethnography, and although his PhD was never revoked, a former Chair of Yale's Anthropology Department claims: "to me it remains a disturbing and unforgivable breach of ethics” (http://www.salon.com/2007/04/12/castaneda/). Perhaps this would not occur in today's academic climate, and perhaps a more rigorous peer review would have noted discrepancies in Castaneda's work, but the data and the data source can be difficult if not impossible to disambiguate.
to post comments
From the computer science perspective, I'd like to add:
1. Publishing software should be separated from publishing data. Here is some info on experimenting with sharing software and checking it during review: http://cacm.acm.org/magazines/2015/3/183593-the-real-software-crisis/
2. When we move to other disciplines (e.g., computer science), formal peer-reviewed conference proceedings play a very important role, may be even bigger role than many journals. Thus, data sharing should be also supported for conferences and in proceedings, not only for journals.
Aliaksandr Birukou
name.surname AT springer.com
to post comments
I'm aware the date of the forum was last Friday, but I hope it is still OK to comment.
My interest is whether journals should require reviewers to review the deposited data. As we already ask so much of reviewers I fear asking them to review large datasets will take up more of their time and perhaps make it even harder for Editors to find willing reviewers.I'm certainly not against sharing data, I'm just trying to get an understanding of how others feel data should be reviewed.
to post comments