How to elaborate a data management plan

Editor's Notes

1. H2020 open data pilot focuses on research data specifically While focus is on digital research data, worth considering all data that you will produce and how you will maintain links between various outputs.
2. As seen in Tim Berners Lee’s model, if you want to embrace the open data approach, it involves more than just making data available. You need to consider making your approaches and methodologies more transparent; you need to provide context; you need to make that data intelligible! Worth noting here that even if you don’t plan to be open, RDM is good practice so that you are able to stand up to scrutiny if questions arise about your findings and practices.
3. Funding body mandates have been a key driver for improving RDM infrastructure and practice. RCUK issued its common principles on research data several years ago to address this need for transparency and value. Funders have an obligation to show that public funds are being allocated appropriately. Important to note that funders need to demonstrate impact too. Otherwise research budgets shrink! Not just UK, H2020 open data pilot and enthusiasm for Open science. Important to stress that funders are NOT saying that ALL data must be kept or shared. Funders expects responsible research practice. The background to this is about making the most of the data that has been created through publicly funded research. The guidelines speak of: Improved quality of results Greater efficiency Faster to market = faster growth Improved transparency of the scientific process
4. Data driven research is becoming the norm for all disciplines even arts and humanities. To be effective in this era, researchers need access to high quality data. Currently, there are very worrying stats around the percentage of published findings that are reproducible.
5. First of all, what is a Data Management Plan (dmp)? Essentially, most funders just want evidence at the grant stage that data has been considered – how much will be generated? Usually just a 1-2 page summary covering the expected data to be produced through research along with an idea of what might be shared and how and when it will be shared. It is important to stress that funders aren’t expecting something carved in stone at this stage. Projects often change quite radically from what is submitted at the proposal stage and this is ok. Researchers just need to be able to provide evidence that they have thought about the data they might be generating and how it will be managed and shared. In terms of preservation, it is important to remember that not ALL data has to be retained. Selecting what data needs to be kept is something that only the researcher can do. Essentially, he/she will need to retain any data that underpins published findings to allow for validation of results. Additional data that is not required for validation purposes but is deemed to have longer-term value might also be worth keeping. DMPs are often submitted as part of grant applications, but are useful whenever you’re creating data. Some HEIs are introducing policies that require DMPs for all research undertaken by staff – whether externally funded or not.
6. Documentation and metadata are essentially descriptive information about the information contained in a dataset. There should be good documentation at the study level, for example a description of the research methodology that created the data – [the best metadata the data can have is the publication it supports] or a data paper. There should also be documentation at the file, item and variable level suitable so that someone reusing the data can understand it – this could be ensuring that excel spreadsheets have sensible row and column descriptions or that a document is included with the dataset which properly explains any abbreviations used.
7. Metadata is a subset of documentation Documentation is a catch-all term which can include some very high-level, human-readable, loosely structured information about the dataset – for example a description of the laboratory method used to generate a set of results or a sketch book accompanying a work of sculpture. The term metadata has come to define a subset of documentation information that uses standardised terms and is presented in a structured way. This metadata will be in a form enabling it to be read by machines and/or may be presented in human readable form as well.
8. Use an existing standard whenever possible (almost always!) DCC Metadata catalogue, explaining what they are, who's using them and how to use them. Ideally, use of standard unique IDs (grant, ORCIDs, funders, institution, DOIs). Metadata should capture provenance info. WC3 PROV-O standard and Australian National Data Services (ANDS) work might be relevant. Think about what is needed in order to find, evaluate, understand, and reuse the data. Have you documented what you did and how? Did you develop code to run analyses? If so, this should be kept and shared too. Is it clear what each bit of your dataset means? Make sure the units are labelled and abbreviations explained. Record metadata so others can find your work e.g. title, date, creator(s), subject, format, rights…,
9. Potentially, the bare minimum of metadata required will be defined by the repository that accepts the dataset. This could correspond to the DataCite mandatory minimum metadata set although it is very likely that other elements will be required, whether by the repository or by the research funder. This metadata set is designed for citation and disambiguation, in other words ensuring that the dataset a researcher is reusing is identical to one cited in a journal article or online. This set of metadata does little to make the resource visible to researchers speculatively searching for relevant data to reuse.
10. Your own device (laptop, flash drive, server etc.). And if you lose it? Or it breaks? Departmental drives or university servers “Cloud” storage. Do they care as much about your data as you do? Read the fine print! Most likely a mixture so be clear on which data will be stored in which place. Remember to consider backup! Use managed services where possible (e.g. University filestores rather than local or external hard drives), so backup is done automatically Good practise is to have at least 3 copies of a file on at least 2 different media with at least 1 offsite Ask central IT team for advice
11. For those that do take part in the pilot, the starting point is to make all data that underpin publications open. After that, it’s for researchers to define what else should be shared and can be made open. This should be outlined in the DMP. Sometimes sharing is not appropriate (e.g. due to ethical rules of personal data, intellectual property protection, commercial restrictions etc). It’s fine to apply restrictions in such cases. This could be an embargo period prior to publication or while a patent is sought, or controlling access and re-use to protect participants’ identities (e.g. via the use of secure data services / data enclaves or data sharing agreements). Restrictions should be outlined up-front in the DMP.
12. Projects in these seven areas aren’t forced to participate. There are various reasons to opt out e.g. if results can’t be shared due to confidentiality, legal restrictions or commercial opportunities. Opt out is available at any point. It can also apply to certain parts of the data e.g. they could share some aspects but not others.
13. Most funders expect enough data to be kept to enable validation of published findings at a minimum. But other data may also have value. Not always the data – sometimes the code or algorithm more valuable and data can be easily recaptured.
14. International Standard Name Identifier (ISNI) is an identifier for uniquely identifying the public identities of contributors to media content such as books, TV programmes, and newspaper articles. Such an identifier consists of 16 digits. It can optionally be displayed as divided into four blocks. http://isni.org/
15. Remember that not all research data is digital. Consider how you will maintain links between research outputs and analogue data like tissue samples. How will access to these resources be provided if needed? In some cases, algorithms and models are more valuable than the data being run through them so be clear on what it is you need to keep (e.g., genomic data and models).
16. While the researchers themselves will have a key role to play in sharing data, it is important to note that they are not the only stakeholders involved in the process. Repository – how will data be made visible? Especially if physical data? Support staff – e.g., Library staff to help describe data to make it easier to find and understand. Research participants – informed consent is required. Anonymisation can be very costly. Commercial – identify what can and can’t be shared and make this explicit. Develop consortium agreements that make each partners rights explicit. Secondary user – when sharing data, researchers can ask reusers to tick that they accept terms and conditions regarding fair use
17. Ideally, we’d like to see researchers embrace Open Science and make as much of their data available with as few restrictions as possible. However, where that isn’t possible, it is important to be clear on why data can’t be shared. Provide a recommended data citation for data set. UK Data Service provides information on tools that are helpful for citing specific sections of interviews and other types of social science data that might be collected.
18. Guidance from the DCC can also help researchers to understand data licensing. This guide outlines the pros and cons of each approach e.g. the limitations of some CC options The OA guidelines under Horizon 2020 point to CC-0 or CC-BY as a straightforward and effective way to make it possible for others to mine, exploit and reproduce the data. See p11 at: http://ec.europa.eu/research/participants/data/ref/h2020/grants_manual/hi/oa_pilot/h2020-hi-oa-pilot-guide_en.pdf
19. Answer questions to determine which licence(s) are appropriate to use
20. Domain best General – need to ensure it’s suitable for your use Institutional – will keep data of low value or data with no other home but may not ensure targeted visibility Journal – fulfils journal requirements but does not offer repo functionality or longevity required by funders Web page – may target domain academics and allow manipulation of data and support for living datasets. Does not have citability, longevity or engender user trust that the resource is the same as the one described in publications.
21. Data journals require that the data being described in articles be freely available and usually mandate where it should be deposited, this can help to identify community-accepted repositories. Some journals may also offer recommendations for appropriate places to deposit research data
22. Depositing into a repository is good. But how can you make sure that your data can be most easily found and reused? A myriad of discovery mechanisms. Institutional data catalogues, national discovery services and funder registries. It would be ideal if data was entered once and shared between these systems more effectively.
23. DCC provides free access to a number of short, practical guides on various topics: -How to discover RDM requirements -How to develop RDM services -How to develop a data management and sharing plan -How to appraise and select research data -How to write a lay summary -How to licence research data -How to cite datasets and link to publications
24. Web-based tool to help researchers write Data Management and Sharing Plans according to different funder / institutional requirements There are various templates in DMP Online based on different funder requirements and institutional customisations. We’re currently enhancing it with practical examples, boilerplate text and tailored support. TEDDINET may wish to develop discipline specific guidance within the tool for future related projects.