Requesting raw data from previously published research

Question

I am working on a mathematical model to help explain some experimental results and generate new hypotheses. Unfortunately, I have neither the resources nor interest to gather experimental data, but in this particular sub-field publishing a model without showing its usefulness on some experimental data is not common.

There are several existing experimental findings that can be explained by my model. However, they are presented in other modeling papers and the raw data is not available with the paper or on the authors' websites. In the papers they only present partially-analyzed data (for instance, they show results averages over participants, but not individual participant's results; or sometimes they only give the results of statistical tests).

I want to contact the authors for their raw data and have 3 related questions:

1. What is the protocol for contacting by email to ask for authors' raw data? Is this common?
2. Will the researchers expect to be invited on-board as co-authors? Or is a citation to their papers, and an acknowledgement of the form "AK would like to thank X, Y, Z for providing their raw data" sufficient?
3. If my model (without fits to specific data) is in a pre-print state then should I send a pre-print to the authors I contact? What if the pre-print points out weaknesses in their approach to modeling similar problems?

Answer 1

This will, as it seems nearly all questions on this site, vary based on field. My answer applies to Epidemiology and medical research only. Your mileage may vary.

• It is very common for this to happen in my field. There has been an increasing emphasis on using meta-analysis and systematic reviews to summarize bodies of work, and with those there is almost always a need for some more information, raw data, etc. to come from study authors. It's fairly routine now. They may not say "yes" for a number of reasons. One may be that they're working on their own projects in a similar direction. But there are others - that privacy laws prohibit releasing data to just anyone or their funding dictates similarly is a fairly common one. So be prepared for "no".

• As for the authorship question, it likely depends on the extent of your data request. Generally, if you're just asking for a few numbers that go into a reported value, then in all likelihood, an acknowledgement is more than sufficient. If, on the other hand, you're asking for access to the raw data from their 5 year, many thousands of dollars cohort study? You're likely going to have a member of their study team be an author on your paper, and there will be far more extensive approval processes than just "Sure, .csv file is attached."

• "If my model (without fits to specific data) is in a pre-print state then should I send a pre-print to the authors I contact? What if the pre-print points out weaknesses in their approach to modeling similar problems?" For the initial contact, I'd suggest it isn't necessary, because what you're really trying to find out is "Is the release possible". At later stages, I would expect to see what you were doing, either in a full explanation, or as a pre-print. As to your "What if?" question - so what if it does? Science is about improving the methods we use.

Answer 2

Many journal formats (and the arXiv) indicate a "corresponding author". That's the person you'll want to contact.

Doing so isn't a everyday occurrence, but it does happen.

In some fields access to the data may be covered by agreements to keep it confidential for a certain period of time (or until first publication). These agreement often specify the answer to the authorship question as well.

Answer 3

This may or may not be helpful in this case, but you may be able to persuade them to publish the data separately, in a dedicated data journal or other data archive. Depending on the field, they may already have done so (some UK funders, for example, are starting to require this as a condition of funding).

This gives you the additional option to give credit by citing the dataset directly as well as relevant papers, and also opens up the possibility that others will use and cite the data as well, bringing more prestige to the original researchers.

Currently, some researchers are open to this type of data publication, while others are not, so your mileage may vary.

If you want to find data to reuse, or to suggest somewhere for them to deposit, there are some lists of archives available to look through, such as the DataCite repository list and Databib.

Answer 4

I talk about biological sciences, but this can probably apply elsewhere. In theory, the moment a set of analysis is published, the data associated with it should also be public or available for:

1. Other researchers who want to use them
2. Other researchers who want to evaluate (i.e. repeat) the experiments and verify the initial findings.

So, in principle, if the publication does not have a link to the public data, then you could contact the journal and complain. Of course it depends on the data, but dna sequencing or protein analysis data are usually available. There might be legal or other limitations for patient, medical or other types of data.

This is the formal way.

There are some exceptions: The data are public but no publication is out yet. Because of policy they want/have to provide the data to the public, but the publication is in preparation. In this case you cannot use the data and you have to contact the PI to see how your analysis comes in conflict (or not) with theirs. Every institution has different guidelines.

Back to the initial issue, in reality, you might find resistance in getting access to the raw data of a published work (which, as I said it shouldn't be the case if the data are not sensitive, because anyone should be able to evaluate and validate their analysis).

You have two options: Check publicly available data (depending on your field - I can propose some in biological sciences I'm familiar to, if needed) and work your models on them. That would be easier for you, as you could avoid peculiar situations.

Contact the authors and propose a collaboration to do a different type of analysis than theirs (with your model) on those data. They would be happy to collaborate with you and they might even provide some insights on your analysis that would make it even better.

To the initial questions:

1. I would say that it's common. It's more common to look directly at the public repositories and see data that are already published. The collaboration proposal might be a safer approach (in terms of results and good relationship and future collaborations)

2. They might be expecting to be invited. It depends on the terms of the collaboration. In any case, it should be agreed upon on the beginning, so you can avoid the frustration at the time of publication (and after you have spent time on working on the data).

3. I'm not familiar, but if you go for a collaboration, then it makes sense to explain your method at the first meeting. If you just need the data, or you take the data from a public repository, you don't have to send anything.