22

Suppose there is data that you as an author of a journal article wants to share with the readership of the article (e.g. raw experimental data, code, gold standard / ground truth data).

What would be the best way to do this? Possibilities include:

  • Add a footnote that data is available upon request by e-mail.
  • Make the data available for download on an institutional webpage.
  • Make the data available for download on your personal webpage.
  • Make the data available through the article's publisher.
  • Something else...
5
  • 2
    Something else: Dataverse.
    – Sverre
    Feb 27, 2017 at 13:13
  • 3
    Cross-site duplicate: opendata.stackexchange.com/q/980/190
    – gerrit
    Feb 27, 2017 at 18:42
  • What does the guidelines to authors say?
    – Greg
    Mar 1, 2017 at 0:55
  • @Greg: For the sake of argument, let's assume that it does not specify anything in this regard. Mar 1, 2017 at 6:12
  • 1
    For maximum utility, be sure to publish the data under a license that clearly permits other researchers to use it. In some jurisdictions, database rights, for example, might prevent re-use unless re-use is explicitly allowed. Some licenses, e.g. the 4.0 series of Creative Commons licenses, gracefully handle this risk.
    – user10623
    May 6, 2017 at 6:43

5 Answers 5

24

The most common and sustainable thing to do is to deposit the data in a research data repository. Depending on which one you choose, the data will get a persistent identifier, e.g. a DOI, can be cited properly in publications, might be reused by other researchers, ...

You can find a list of available data repositories at re3data, the registry of research data repositories. The number of repositories is still growing. Even your institution/university might offer an institutional data repository.

13

If it is data that is of general interest, then go for a public repository as FuzzyLeapfrog mentioned (of course, only do this, if you have the legal right to do so).

If it is something that is very specific to the publication (e.g. code that produces the results/graphics/tables from the publication), then use the "additional resources" from the publisher (if available and convenient).

7

The Open Science Framework is quite good. https://osf.io/

It's not commercial, which is a nice property. It offers flexible, archived storage for a project. It has many tools tailored for storing data, code, and materials in an academic environment (for example, you can even share a link that blinds the author names when submitting as part of blind review).

For more information, check out: https://osf.io/support/

6

It depends on a field. In my area, it is nowadays quite common to provide additional information in GitHub repository. GitHub, while being primarily software sharing platform, fits quite well to such task, as research (in a form of journal/conference publication) is typically supported by some code that was used to collect data and process/analyze it, datasets themselves and the description of these datasets (i.e. metadata). On top of that it is easy to create a set of web pages or even site using GitHub Pages.

Sharing data on GitHub works like a charm if your datasets' volume is moderate (say, <100Mb). If larger, then a Github repo may contain code, metadata, some sample extracts from the datasets and, in addition, specifies how to access the datasets themselves. The latter might be direct links to the data stored in, e.g., Dropbox or some other online storage, in your department/organization storage system (if there is such), etc. And/or it might be just instructions on how to obtain the datasets (request by email, etc.). In addition, any other related information can be specified there - like copyrights, how to refer to a publication and/or its supporting materials, etc.

5
  • Is GitHub so much prevailing in comparison with other git hostings, e.g. BitBucket and others? Feb 28, 2017 at 19:45
  • The specific git platform like GitHub or BitBucket or GitLab, etc. is not that crucial, imo. Still, by default (plus considering free plans), GitHub is preferable as it has the largest user base (hence, more potential that your stuff would be noticed). BitBucket (or GitLab) overcome GitHub in allowing private repositories but for academic projects/data (which are primarily open-source, at least should be) it is not a real drawback. Bitbucket is worse than GitHub or GitLab as it has restrictions on number of collaborators (not crucial but important).
    – Denis
    Mar 1, 2017 at 11:05
  • Lastly, Bitbucket is better that the other two in terms of integration with other software development/management systems (like JIRA), but that is irrelevant in most academic projects.
    – Denis
    Mar 1, 2017 at 11:07
  • 1
    Zenodo can also mint a DOI for a Github repository, making it easier to cite: guides.github.com/activities/citable-code
    – Gaurav
    Mar 2, 2017 at 0:11
  • 1
    The problem with using a git platform is that there is no guarantee of long-term availability. Not only are you relying on GitHub to host your data for free in perpetuity, but it's possible for you as the owner to delete or replace the data with something else, accidentally or intentionally. The advantage of depositing in a specialist archive is that these usually have some technical and social guarantees of availability and fixity, which is important because this shared data is essentially becoming part of the scientific record. Guarav's suggestion gives you the best of both worlds though.
    – Jez
    Mar 11, 2017 at 9:16
-1

Almost everybody can consume EXCEL (CSV) files, so host on google drive. Share links with emails you want to be with. No hassle of 3rd party servers or data service and universally acceptable CSV format will be here. For very large data split into files.

3
  • You do not want to bother your readers with creating a Google account if they do not have one.
    – Karlo
    Feb 28, 2017 at 17:56
  • Excel files and CSV files are quite different. Feb 28, 2017 at 21:27
  • How to handle data like climate model outputs containing multidimensional data over time? Definitely not CSV! How to ensure data availability in 10 or 20 years? What about long-term archiving? Feb 28, 2017 at 23:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .