Introduction

As you now know, the systematic management and oversight of research data is rapidly becoming a core skill set for researchers at higher education institutions in Canada and internationally, as well as for the librarians and other data professionals who support researchers. While advances in how different types of data are managed benefit all research in the long term, this shift continues to raise a host of practical concerns for those responsible for Research Data Management (RDM). One core challenge is that RDM is an emergent field of practice. How data are expected to be managed varies by data type, field of study, institution/funder, and jurisdiction. Initiatives to manage and share genomic data, for example, are more advanced than efforts involving ethnographic data. Similarly, not every country attaches the same urgency to implementing strategies for advancing RDM.

This situation means that for researchers and others needing to manage data, sometimes there may be a clear path to follow with reliable signposts, while in other cases, it’s about breaking trail. What all RDM work has in common is a need to think critically about the tasks at hand. No single approach will apply across the board. The aim of this chapter is to help you develop a critical perspective when managing research data, regardless of your role in the process. As you will see, the ability to think through RDM-related challenges requires skill sets spanning multiple areas: developing a familiarity with the complexities of research data; applying current approaches in novel circumstances; knowing how and when to look to external communities of practice for support; and sharpening your own resourcefulness and creativity.

The key takeaway is that the work of managing data is both an art and a science. While there may be principles and practices to guide the way, “doing RDM” frequently comes down to time-strapped researchers new to RDM who are trying to wrangle their data into shape as best they can.

The discussion is framed around three questions:

• Why the push for RDM? This examines what is driving new requirements to manage research data more systematically and why the answer matters to you.
• Whose responsibility is it? RDM work encompasses different groups. The responsibilities and expertise of each impacts how work gets done and support is provided.
• Where is the leading edge? Because RDM is still emergent, your efforts need to be guided by current practice and by an awareness and appreciation of ongoing change, whether in Canada or abroad.

With these questions in mind, the chapter concludes with a series of practical steps to consider when managing research data for any project. Together, the questions and steps are meant to help you enhance your problem-solving skills and maximize your capacity for RDM-related work.

Why the Push for RDM?

If you’re new to research data, you may be surprised to learn how innovative a systematic, externally driven approach to RDM is. The management of research data has commonly been left to researchers, who, along with their institutions, have been responsible for how data are organized and archived, and whether they are shared with others. Research data have been frequently seen as proprietary, the product of a substantial investment of time, personal effort, education, and professional development on the part of the researcher. Data have formed the cornerstone of careers and the basis of scientific publications. They might be shared informally with close colleagues, but there has been minimal incentive — much less requirement — to organize data to external standards or to make them openly available to others.

Under this arrangement, there’s been little impetus for a more systematic approach to RDM. So what has now changed? To some extent, there has been a cultural shift in different research communities to acknowledge the impact that collaboration can have on the advancement of disciplines and the production of knowledge. While this evolution continues (more quickly in some fields than others), it alone does not entirely explain why concepts like the FAIR principles and tools like Data Management Plans (DMPs) have emerged. Two other factors have been particularly impactful: (1) changing expectations by funders, and (2) technological advances and the power of big data.

Changing Expectations

For over a decade, major research funders (e.g., the three federal research funding agencies in Canada and similar bodies internationally) have moved to maximize knowledge output from the research they support. Funders demand well-organized and (ideally) open data for several reasons. First, well-managed data reduces duplication by allowing researchers to identify what studies have already taken place on a topic. RDM opens access to research data more fully, expanding beyond what is included in the papers or books a researcher chooses (or is able) to publish. Second, greater access to extant data translates into improved opportunities for secondary analysis, which maximizes research outputs for every dollar, Euro, pound, etc. invested. Finally, funders recognize that improved data management and openness safeguards the robustness and transparency of the research they support (Pinfield et al., 2014).

Technological Advances

The second factor driving RDM is advances in computing technologies: notably, the capacity to generate, store, and work with very large datasets; the arrival of cloud computing and data sharing across the Internet; and decreasing costs of computing technologies. Such improvements were originally of greatest benefit to fields working with big data (e.g., astronomy, genomics, geospatial mapping), which explains in part why RDM has advanced more quickly in some disciplines than others (the nature of the data involved — quantitative — is another factor). Such technological progress has shaped what is possible in other fields of study, such as the digitization of humanities resources and the capacity to analyze social media data. Improvements in analytic software have also permitted research data to be linked securely to other forms of data (e.g., medical records, meteorological sources), creating still larger datasets.

Other Factors

Of course, other factors are also driving RDM. Managing data more consistently makes the research process more efficient and can lead to more robust findings. As discussed in the chapter on RDM and qualitative research, better organization of interview data enhances analysis because connections can be made within large sets of transcripts which would otherwise be difficult to make. Research is also increasingly transdisciplinary, meaning it crosses epistemological boundaries and methodologies and brings together diverse groups of researchers. RDM supports such efforts and facilitates collaboration. Lastly, some scholars at the end of their careers want to leave behind data-based legacy products that explain their data beyond what is captured by standard metadata, like how and why a particular methodology or theory was applied to generate the data. Enhanced RDM practice also allows highly experienced researchers to link together data from an array of related studies, sometimes spanning decades. Better management (especially documentation) ensures that future use of such data respects what is — and isn’t — possible in terms of secondary analysis.

Acknowledging the drivers of RDM helps us understand why data management is important and what our own data objectives are. If you’re a researcher, is your priority simply to meet the RDM requirements of your funder? Or is it also to establish a comprehensive research agenda over time? If you’re a data librarian helping researchers prepare their data for deposit in a repository, what do you need to know about the expectations for RDM in a particular discipline? What baseline are you working towards? There are many reasons for doing RDM and many levels at which it can be done. It’s therefore critical to align the RDM strategy and endpoint for a particular project with wider contextual factors.

Whose Responsibility is It?

The push for more systematic approaches to RDM brings with it questions about who is responsible for what. Who organizes data and how? Who decides what metadata standards to observe? Who selects a repository? The list of tasks and decisions is extensive. As a rule, final responsibility for managing data rests with the most senior researcher involved on a project, namely the principal investigator (PI). In practice, PIs routinely delegate most RDM-related work (e.g., data collection, cleaning, organization, archiving) to other members of their teams, notably post-doctoral researchers and research associates. This is where most data management in research takes place.

Delegating responsibility brings with it at least two complications. First, the individuals who are closest to and often most familiar with a dataset are frequently employed on shorter-term contracts. When they move on to other opportunities (as many do), their knowledge goes with them unless measures are taken in advance to document that knowledge as fully as possible. Unfortunately, this doesn’t always happen, impacting how effectively and consistently data are managed across the length of a study. Second, depending on their experience and training, such team members may be adept at RDM and require minimal guidance, or they may be new to RDM principles and best practices, meaning they require close oversight, effective training, and support from data management experts outside the research team.

Internationally, two models have emerged for providing RDM support services: librarian-led and researcher-led. Both seek to upskill researchers at all levels and to facilitate data management in line with funder expectations, journal demands, and the evolving practice of specific disciplines. A key difference between the models is who provides the support.

A librarian-led approach to RDM is most common in North America. It allocates responsibility for RDM services to academic libraries, where data librarians and other professionals help train and assist researchers with the management of their research data and support RDM strategy at an institutional level.

The researcher-led approach is seen frequently in Europe, where responsibility for RDM services is assigned to newly created divisions within universities. Such offices may be located in — but not necessarily of — the academic library, meaning RDM services are developed and managed separately from library services. RDM support work is typically delivered by individuals with doctorates (or at least research-based master’s degrees).

Each model has its strengths and limitations. Academic librarians are experts in managing information and, in North America, typically share a common qualification (i.e., an ALA-accredited MLIS degree). As such, they have a comparable grounding in information management principles and practices. On the other hand, while some academic librarians conduct research or hold PhDs, their primary professional role is research support, meaning they may have a limited background in conducting larger research projects or collecting and analyzing complex data.

By comparison, researchers become data experts as they earn their doctorates. Across their careers, they spend years entrenched in particular research cultures, working directly with data. But they are less likely to have proficiency in standardized ways of managing and organizing information and data. It’s not uncommon for researchers to develop idiosyncratic systems that work best for themselves and their teams.

There is little examination in the literature of the models or their effectiveness. This likely reflects the degree to which RDM services are still being established and integrated into academic structures and cultures. What the models tell us already is that RDM involves multiple groups. While final responsibility for RDM rests with PIs, routine data management and the oversight and delivery of support services typically fall to others. It is important to acknowledge that researchers, librarians, and other data professionals each bring their own expertise and perspectives to how data can be managed and to how RDM should develop going forwards.

Where is the Leading Edge?

As much as RDM is a relatively recent phenomenon, it’s important to remember that research and data evolve. New topics of inquiry and research technologies continue to appear (e.g., the opioid crisis, gene editing), as do novel types of data and ways to analyze them (e.g., social media data, augmented analytics). Such advances allow research that wouldn’t have been possible even a few years ago. While the pace of such change varies over time and by field of study, it impacts how we approach RDM, including what support services are available and how they are provided. This section highlights two examples where it’s important to reflect critically on current practice and stay attuned to developments taking place elsewhere.

The first involves the frequent use in RDM training of a lifecycle infographic to represent the research process and data management within it — discussed in chapter 1. Such images are meant to spotlight standard steps in research, from initial planning to data archiving and reuse. Lifecycle models are effective because they’re accessible, but such imagery does not always align well with how some forms of research unfold. This may lead to flawed understandings of how RDM can or should take place. For instance, much social science data is collected iteratively, meaning researchers undertake real-time reflection and methodological modification throughout the data collection process. A sociologist, for example, may introduce new lines of questioning or new participant groups during a series of focus groups. As a result, such studies do not unfold in ways similar to lab-based research.

Despite their circular imagery, lifecycle models are curiously linear and imply a start-to-finish process that doesn’t map well onto some methodologies. Such models also fail to highlight the importance of relationships in data collected across interrelated studies or over periods of time (e.g., in longitudinal research). They struggle to represent the ways extant data are increasingly used to generate new data through secondary analysis and data linkage, with the results then augmenting the capacity for still more research. The challenge for RDM practice and service delivery becomes how to keep up with the expanding ways research is conducted and the types of data generated.

The second example demonstrating the need to monitor the leading edge of RDM involves administrative and governance structures. A refrain heard frequently in data sharing, especially regarding the use of repositories, is that data should be as open as possible but as restricted as necessary. But how does this translate into practice? Current approaches typically include open access, the imposition of an embargo period (i.e., access is not allowed for an initial period of time before being made openly available), or maybe requisite permission from the original researcher. What other options are possible?

In some areas of research, infrastructures have been developed to review proposed uses of data prior to data being released. Such governance systems help ensure compliance with original ethics restrictions, prevent damage being done to the original researcher’s (or research team’s) intellectual property, and prevent harm to study participants, such as the re-identification of individuals (Murtagh et al., 2018). Enhanced forms of review can also facilitate data access by non-researchers (e.g., journalists, political groups, citizen scientists) while also ensuring that a researcher’s work is not intentionally or unintentionally brought into disrepute (Murtagh et al., 2018).

One such form of governance is the Data Access Committee (DAC). Still found mostly in Europe and the United States, DACs are independent decision-making bodies whose purpose is to oversee access to datasets for research purposes. They act somewhat like an ethics committee but at the tail end of research, regulating access to data that have already been collected. DACs are most common in human biomedical research, where combining data allows for more advanced analysis of much larger samples. For example, a team may want to pool data from several biobanks internationally to study the link between a genomic variant and a particular health condition. Because such data are highly disclosive, they are unlikely to ever be openly available. Some DACs use machine-based decision-making tools to render decisions based on level of risk, while others rely on reviews by experts in the field. Human committees are typically preferred where research is leading edge, where data are being used in novel ways, or when the subject area is particularly sensitive.

It’s important to be aware of critical work, like that of Cox and Tam, or new data access infrastructures, like METADAC, because such knowledge helps inform how to manage data and organize RDM support services. The focus of RDM will shift as data management expands to encompass more disciplines and types of data and as the nature of research and data progresses. Your work in this field needs to be guided by current best practices and a need to accommodate change and stay abreast of developments elsewhere.

The Realities of Managing Research Data

Even with well-systematized processes in place, managing data will never be a checkbox exercise. There are always decisions to be made and ways in which data don’t quite fit existing practice. In this final section, we consider the reality of undertaking RDM on the front line of research. How do researchers (and those supporting them) collect, access, process, organize, analyze, describe, and archive research data in ways that meet the requirements of funders and host institutions but also fit pragmatically into the work of the research team?

Depending on the complexity of a project, the methodical management of research data can become overwhelming, disorganized, or overlooked altogether at any step in the process. Things can go awry whether you’re strategizing RDM at the start of study, troubleshooting issues in the middle of one, trying to make sense of data that already exist, or helping others in any of these situations. Despite such hurdles, research data can be managed effectively even in the most difficult situations if you think critically, act consistently, document what you do, and search out best practices and support where needed.

The following is a broad outline of things to consider when actively managing research data on the ground. It reads somewhat like a DMP, though this outline was developed from the perspective of doing RDM rather than planning for it. That is, while DMPs are meant to be living documents, nothing is as immediate for researchers as having to manage data alongside any number of other pressing tasks. The fact is that RDM is being shoehorned into a world that is already time deficient. The ideas put forward here are based on what I have learned directly and from colleagues over a decade-plus career as a researcher and data manager. The seven points raised below focus on researchers but should also help data librarians and other professionals gain greater insight into RDM.

1. RDM is as much about thinking and problem solving as it is doing. Managing data is a big-picture activity. It’s not only about the data and how to manage them. It’s one (relatively new) element within a much larger research process. Conversely, when undertaking a specific study, there may not be much to indicate exactly how data should be managed. To use a research-based analogy, analytic software like SPSS and NVivo make working with data more manageable. However, such programs do not analyze data. That is the job of the researcher. While RDM approaches can guide data management, researchers and those who support them must think critically about the data at hand and how to apply principles and practices in ways that are practical (and often novel), and that improve research processes and outputs. Put simply, when doing RDM, in addition to acting, allow enough space to reflect and think critically.
2. Write a Data Management Plan. DMPs are useful because they make researchers think through critical aspects of how they will manage data generated during a study. Even if a grant application doesn’t require a DMP, consider writing one. When you’re finished, ask yourself what your DMP does and doesn’t include. Remember that DMPs are goal oriented and aspirational: they tell you where you want to go and how you hope to get there. They do not address the realities of managing data in everyday research life, like dealing with a co-investigator who isn’t naming data files correctly or struggling to identify a suitable repository. This is where point 1 comes into play.
3. Consider what is driving your RDM efforts. These days, most of us involved in research are upskilling ourselves in RDM because we feel we have to, particularly since funders increasingly require evidence of good data management. But what other factors are at play in your project? As a mid-career researcher, you may realize that better organization of your data can positively impact your research findings or capacity to collaborate with others. As a post-doctoral researcher, you may notice that your study’s PI is also new to RDM, making for a great opportunity to beef up your skill set to help lead in this area. If you’re undertaking a secondary analysis, maybe you’re required to return your data to the original study and need to know what level of data management is expected. Whatever the situation, there are benefits to identifying why RDM is important to you.
4. What would the perfect outcome look like? This step is important for anyone working in disciplines that have few well-defined RDM guidelines or best practices. Spend time reflecting on the ideal approach for managing and archiving your data. If you were a researcher looking for data with which to conduct a secondary analysis, what would the perfect dataset look like in terms of its organization, documentation, metadata, access arrangements, and so forth? While such a picture-perfect solution may not exist, there are likely excellent close examples somewhere in the world. Again, think critically about where you might find them and start looking. Keep asking questions until you get answers you can work with.
5. Be prepared to approach your data and its management iteratively. Research data are almost never collected in their final state. Data variously need to be cleaned, reformatted, anonymized, aggregated, and so on, before being suitable for analysis and archiving. As a researcher, you must decide whether all your data are equally useful (to the project, to other researchers). It’s essential to document your data and their provenance because such details provide others (team members, secondary users) with critical information, including what analysis the data can and cannot support. All such efforts are dynamic, meaning what you think and do early in a study may change as the project unfolds. RDM is seldom a once-and-done undertaking. Something as straightforward as a file-naming protocol may no longer function properly at the analysis stage.
6. Who is doing what? Effective data management, especially when it involves research teams, requires defined roles and responsibilities as well as continual review to ensure what is meant to take place is indeed happening. Upskilling may be required for some or all team members, so assess the situation and identify outside resources early. Meetings may eat into precious time, but bringing team members together regularly to exchange information about RDM on a project helps address challenges when they inevitably arise, such as a post-doctoral researcher leaving for a tenure-track position. As always, make sure you and your team document RDM efforts systematically using resources like audit trails and standard operating procedures.
7. Accept that things may not go smoothly — but you’ll get someplace reasonable in the end. RDM is like the research processes it supports: ever changing and never perfect. Do the best you can and apply what you learn going forward.

Conclusion

This textbook is an excellent primer on critical issues in the management of research data in Canada. The various chapters introduce a wide cross section of valuable RDM principles, policies, strategies, and practices that you will need to know as a researcher, academic librarian, or data professional. The main takeaway from this chapter is simple: data management will always require reflection and an openness to new ideas and practices. For the most part, RDM remains the responsibility of researchers working in the trenches, most of whom are still new, not so much to managing research data but to managing it in line with emerging external requirements. Unfortunately, such requirements often do not translate readily to research as practiced, resulting in any number of ongoing challenges. Librarians and other data professionals offer valuable support in this work, although their efforts must be assessed critically as different service models arise. RDM is neither a singular nor a static enterprise. What you learn in this textbook is fundamental, but a critical perspective and curiosity about how things might be different elsewhere are equally essential.

Additional Readings and Resources

Cheah, P. Y., & Piasecki, J. (2020). Data access committees. BMC Medical Ethics, 21(12), 1-8. https://doi.org/10.1186/s12910-020-0453-z

Kruse, F. & Thestrup, J. B. (2017). Research data management – A European perspective. De Gruyter Saur. https://www.degruyter.com/document/doi/10.1515/9783110365634

Pasek, J. E. (2017). Historical development and key issues of data management plan requirements for National Science Foundation grants: A review. Issues in Science and Technology Librarianship, 87. https://doi.org/10.5062/F4QC01RP

Rice, R., & Southall, J. (2016). The data librarian’s handbook. Facet Publishing.

Thompson, K., & Kellam, L. M. (Eds.). (2016). Introduction to databrarianship: The academic data librarian in theory and practice. In L.M Kellam & K. Thompson (Eds.), Databrarianship: The academic data librarian in theory and practice. Association of College and Research Libraries. https://scholar.uwindsor.ca/cgi/viewcontent.cgi?article=1047&context=leddylibrarypub

Whyte, A., & Tedds, J. (2011). Making the case for research data management. Digital Curation Centre.

Reference List

Cox, A. M., & Tam, W. W. T. (2018). A critical analysis of lifecycle models of the research process and research data management. Aslib Journal of Information Management, 70(2): 142-157. https://doi.org/10.1108/AJIM-11-2017-0251

Murtagh, M. J., Blell, M. T., Butters, O. W., Cowley, L., Dove, E. S., Goodman, A., Griggs, R. L., Hall, A., Hallowell, N., Kumari, M., Mangino, M., Maughan, B., Mills, M. C., Minion, J. T., Murphy, T., Prior, G., Suderman, M., Ring, S. M., Rogers, N. T., … Burton, P. R. (2018). Better governance, better access: practising responsible data sharing in the METADAC governance infrastructure. Human Genomics, 12(1), 1-12. https://doi.org/10.1186/s40246-018-0154-6

Pinfield, S., Cox, A. M., & Smith, J. (2014). Research data management and libraries: Relationships, activities, drivers and influences. PLoS ONE, 9(12): e114734. https://doi.org/10.1371/journal.pone.0114734

Plomp, E., Dintzner, N. J. R.,. Teperek, M., & Dunning, A. (2019). Cultural obstacles to research data management and sharing at TU Delft. Insights, 32(1). https://doi.org/10.1629/uksg.484