Closed JEK-III closed 10 years ago
JEK-III
commented
10 years ago JEK-III
commented
10 years ago Unforuntately, actual practice lags far behind this consensus. Not all article publishers allow data citations in the references and, even when permitted, most authors refer to data in the text without a formal citation\cite{mooney_anatomy_2011}. Many data publishers provide no guidance on citation; others ask users to cite a proxy publication (e.g., from the \emph{NAR} database issue). However, a growing number of data publishers do supply users with explicit citation instructions. Dryad, Figshare, and Zenodo dataset landing pages all display a formatted citation and links for import into reference managers.
Yes, data citations are important but some datasets and web-based resources do not show how they should be cited, some journals do not allow citations to web resources in the Reference list, and even were both possible, too many authors neglect to cite actual datasets and instead cite a web site (which may have many datasets) or a related print paper.
I agree that a DOI is not enough and only permanent if it is updated when documents are moved. A full author-tile-publisher citation as you suggest is more informative and human readable.
I do not think it is problematic to cite parts of datasets. Pages and chapters in books are already cited for example. In most datasets it is also possible to identify individual dat records. Also, the actual data used could be provided in an Appendix so the reader is left in no doubt. Neither do I think 'versioning' is a problem. Where new data are added (e.g. to a time-series) then they comprise a new dataset, as they would if published in print. Where many corrections are made they a dataset can be treated like a paper; i.e. the original can be 'retracted' and replaced, or the new version be published with the metadata stating that it is more accurate.