When writting an article with outputs (by outputs I mean a big dataset created), how do you share them?

Do you ask the readers to contact the main author? Should I pput them in a university website?


6 Answers 6


Take a look at http://thedata.org/, which provides free, permanent hosting of data allowing you to cite a specific iteration of a dataset in your printed work.


Depends. Some journals have online depositories for supplemental material such as data tables and multimedia files that are not appropriate or impossible to be included in main articles. For example, here's an excerpt from the official website of Physical Review Letters:

Supplementary material associated with an article (e.g., data tables, color-image files, multimedia files) may be submitted electronically for joint review. If the article is published this material will be deposited as Supplemental Material. Information about Supplemental Material is available via the Supplemental Material links on the Information for Authors subpage at http://prl.aps.org/author-information/, in the Authors, General Information section.

I've also seen many mathematics papers that cite authors' websites for computational results, and actually I've written one where I gave a link to my co-author's website for mathematical objects found by a computer search. Different journals may have different policies about how you should present and maintain such material. So you should check that on your own before submitting your work.


It depends on your field. If you're not aware of major repositories, ask a librarian who specializes in your field.

Most fields in most countries are getting field specific repositories. For example, I'm currently working on a project building a repository for Australian Archaeological Data (as well as collecting it).

We're forking our repository from tDAR in the United States, and are cooperating with the Archaeology Data Service in the UK. The list just goes on (Open Context, OCHRE, etc...).

When trying to investigate data publishing, always try to check for other countries as well. There are many data publishing projects getting off the ground. It may be worthwhile engaging a research librarian for this task.

Never try to reinvent the wheel, here. You are not experiencing a common problem, and local university library archival storage is... sometimes very much a mixed bag.

  • 1
    +1 - There are also other benefits to utilizing already available services. For instance ICPSR (for social sciences, mainly in US) they have analyst do quality control for the data, disseminate it in various statistical file formats, can restrict access to only individuals who have applied, and can even anonymize the data for you.
    – Andy W
    Commented May 14, 2013 at 11:59

In addition to the answers outlined previously, I would like to add that some authors/researchers just share the datasets freely on their own websites or on their university's data repository. For example, see this (an example of multiple datasets shared on the author's website and referred to in subsequent papers) and this (example of dataset shared on the university e-commons repository).


I like https://zenodo.org/ for the following reasons:

  • It is free (as in beer)
  • You can publish datasets up to 50GB
  • Your Dataset gets a DOI
  • The interface is clean

I've recently published a dataset there (paper, dataset)


In my field (chemistry/spectroscopy), it is so far rather uncommon to make the data directly available, so the strategy is email the author.

Here are my 2ct about the reasons for storing the data in private repositories:

  • I'm working in a research institute, not in a university. Public release of data here has to be decided the same way as licensing for produced software is decided. However, here this is rather easy from a legal point of view, as basically the institute owns the data. At universities this is often far more difficult as the ownership can be shared between several persons and institutions, e.g. if not all people involved in producing the data set had a working contract that includes this data collection (e.g. PhD students, guests).
    The data set is often the primary product of a researcher. Collecting samples and conducting experiments can take lots of effort (I'm talking of years).

  • Sometimes there is sensitive (e.g. patient) data. In that case, better be on the safe side. What if later on a flaw in data anonymization is found in a publicly accessible data set?

  • It is often difficult to share the data anyways (that is: in the way that would be most valuable for us spectroscopists) because it can be extremely hard to combine data sets acquired with different instruments.

  • These points together usually mean that if someone is using someone else's data for new original research it is usually done in the form of a collaboration.

  • People who do offer data for sharing often like to keep at least approximately track of how many people get their data because this number is useful for all kinds of end of the year / end of the project reports and number of people who asked about the data is a much more convincing number than download counts.

  • All in all, we have a central server here, where raw data and source and possibly intermediate results are stored together with the final version of the publications.
    The size of such a package can easily be of a size that would mean that a few downloads would equal the monthly internet traffic of the whole institute. This is outside the supplementary material policies of the journals in which I have published so far. Some friends in another institute in the neighbourhood share their data with collaboration partners by mailing hard disks...

  • Last but not least: there's also a question of reliability - will, say, github still be there in 20 years? Will your university web page still be there in 5 years? (OK, the corresponding author's email may vanish just as fast..., but it is usually not that difficult to track down the new email or the email of another of the co-authors).

  • 1
    Almost every chemistry journal I know allows supplementary material/supporting information and that is exactly where you should put that data.
    – user64845
    Commented Jan 31, 2017 at 9:52
  • 1
    @DSVA: this is only practical for sufficiently small data sets (I can easily acquire about 100 GB of nice FTIR images in a work day. The nonlinear imaging colleagues talk in TB). The main reason however is: I'm bound by my working contract to obey the institute policy which is: data is archived on central server at the institute (My working contract does not even allow me to keep a backup of the data and analyses I published as employee there once the contract ends - even though good scientific practice asks me professionally to be able to answer questions on the paper years later) Commented Jan 31, 2017 at 11:40

You must log in to answer this question.

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