7 Data Management

EcoHealth Alliance is committed to producing and promoting reliable and reproducible research. In order to achieve this, we have to provide data (and other research outputs) that non-team members can interpret and use; as well as promote best practices for data management among collaborators. Ideally, the framework for managing data laid out in this chapter will facilitate the creation of high quality, share-able research outputs. By focusing on Data Management Plans and the dmptool, we can build on well established workflows for producing high quality research outputs.

7.1 Data Management Plan

Data Management Plans , also called Outputs Management Plans or Data Management and Sharing Plans, are living documents that help structure the creation and management of data throughout the lifecycle of a project. DMPs are flexible and do not force researchers to choose a particular technology set but rather ask probing questions about the mechanics and ethics of data use in research projects. Organizing data management in this way provides a common framework to think about data without requiring specific technologies be used in the research workflow. Furthermore, DMPs use stable identifiers (URIs) to connect components of the research workflow, making long term data access more reliable.

The majority of funders require a DMP; however, each funder has specific expectations about what, when, and how research outputs should be shared. It is important you and your collaborators understand those expectations before submitting a DMP. Its equally important that all collaborators understand and agree to the obligations created when submitting a DMP. Early communication between collaborators is key to navigating differing expectations about data sharing from researchers in different contexts.

Data management plan as hub in knowledge management system

Important note on budgeting: Data management activities, but not necessarily infrastructure, are an allowable cost for most funding agencies (NIH, NSF, NASA). Gray areas include paying for hosting services and other infrastructure-like components of the DMP.

Benefits of using a DMP:

  1. They provide a scaffold for you to conceptualize data management for your project
    • What data do you need to answer your research question, where will it come from, what resources are needed throughout the project lifecycle, what are the mechanics of managing the data?
  2. They make it easier collaborate
    • Defining responsibilities, Committing to using data standards, Documenting how the project works
  3. They make it easier for your data to be reused
    • You get more citations, your effort contributes to knowledge creation in unexpected ways, your results become more reproducible
  4. They are a funder requirement and you want funding
    • NIH, NSF, NASA, Wellcome Trust, etc. require a DMP be submitted with a proposal.

Components of a DMP:

  1. Data Type - What will be collected or created during the project?
  2. Related Tools, Software, or Code - Whats needed to make your analysis run?
  3. Standards and Documentation - How will people/machines know what they are looking at?
  4. Where, when, and how will data be made accessible?
  5. Restrictions on data use - How will you abide by ethical standards or other restrictions on data reuse?
  6. Responsibilities - Who is supposed to do what? How will you monitor that? What do they have to do it with?

EHA DMP Philosophy:

  1. Its never too late to write a DMP
  2. Data Management Plans are living documents that change with a project
  3. DMPs are created collaboratively and stored in DMPTool.org
  4. We ensure our DMPs meet EHA best practices for FAIR data and Reproducible Science
  5. Projects should have adequate resources (personnel time, infrastructure, time in project schedule) to implement the DMP
  6. Collaborators, especially those from outside institutions, are full participants in the DMP process

7.1.1 Expectations by project phase

Proposal/Pre-Award Phase

  • Look for funder requirements and use funder specific templates for DMPs. If no template exists, use the EHA Minimal Data Management Plan or create one based on funder requirements.
  • Think about how you might make data Findable, Accessible, Interoperable and Reproducible (FAIR)
    • use the re3data data repository catalog to find identify a potential archive for your data
  • Establish expectations for data sharing and outputs with collaborators and PIs. These discussions should begin early at the same time as discussing project responsibilities and budget.
  • Consider what tools you will use throughout the lifecycle of your data 
  • Consider how data collection, analysis and management tasks will be divided among collaborators
  • Incorporate data management activities into your project staffing, budget, and schedule
  • Outline the ethical considerations for properly managing data in your project
  • Ensure collaborators and PIs understand the commitments they are making via the DMP. Request and incorporate feedback from collaborators.
  • Schedule a meeting with the Data Librarian, create a timeline for proposal submission, and have a notion of tools and standards to use

Post-Award/Early Phase

  • Review and update proposal DMP
    • include updates from IRB or IACUC
  • Refine roles and tech stack
  • Provide more detail on data collection and storage
    • Think about how measurements and primary data sets will be stored
    • Think about how statistical models and derived products will be produced and stored
    • Consider where data will be stored long term, how it will be accessed, and by whom
    • Determine backup strategy and setup backups
  • Think about how you will store artifacts of analysis
    • Where will your code live? Who will be able to access it?
    • If you’re using spreadsheets with formulas, proprietary software or other methods analyzing data, how will you make that workflow reproducible?
  • Schedule a meeting with the Data Librarian

Operational Phase

  • Link the plan to the research artifacts (data sets, publications, code repositories, etc.) being created via URI’s.
  • Review, revise, and update components of the DMP
    • check that IRB/IACUC documents match DMP and protocols
  • Make sure all SOPs and relevant data collection or analysis documents are accessible (linked) to the DMP 
  • Check that data storage locations and methods are well described and linked 
  • Check that necessary stakeholders are identified 
  • Check that appropriate privacy and security measures are working as expected 
  • Add any publications or research outputs are to DMP outputs
    • DOI’s can be assigned to code, datasets, and published articles
  • Check that your data products and code meet the needs of your proposed long term storage solution
  • Schedule a meeting with the Data Librarian

Publication and Archiving phase

  • Review and revise your DMP
  • Check that linked objects are accessible to the appropriate individuals
  • Add DOI or stable identifier for research objects 
    • DOI’s can be assigned to code, datasets, and published articles
  • Submit materials to long term storage
    • Ensure sharing and access are in agreement with requirements from IRB and/or research 
    • Use EHA institutional tags where possible e.g. Zenodo Community
  • Schedule a meeting with the Data Librarian

7.1.2 Using DMPTool to create prepare your proposal data Management plan

  1. Create an account on DMPTool.org associated with EcoHealth Alliance
  2. Identify Funder DMP requirements and Schedule a meeting with the Data Librarian
  3. Create a DMP using appropriate template given your funder. If no template is available or the funder has no requirements, use the EHA Minimal Data Management Plan. Add collaborators and complete as much of the plan as you can
  4. Principle Investigators and Project Partners explicitly agree to abide by the DMP. All collaborators should fully understand and agree with the data sharing components of the plan before approving it.
  5. Request feedback from the Data Librarian
  6. Work with the Data Librarian to incorporate feedback
  7. Export DMP for inclusion in grant

7.2 Notes on data management

Can the data be shared and published, and easily re-used in other analyses?

  • Create and maintain a data management plan
  • Store data in simple, interoperable formats such as CSV files.
  • Microsoft Excel can be a useful tool for data entry and organization, but limit its use to that, and organize your data in a way that can be easily exported.
  • Metadata! Metadata! Document your data.
  • For relational datasets you can create linked data on Airtable. For more information see 8
  • For data sets that cross multiple projects, create data-only project folders for the master version. When these data sets are finalized, they can be deposited in public or private data repositories such as figshare and zenodo. In some cases it makes sense for us to create data-only R packages for easily distributing data internally and externally.

We aim to generally work in a tidy data framework. This approach to structuring data makes interoperability between tools easier.

7.3 Backups

Cloud based storage solutions like google drive, dropbox, and aws S3 (used in airtable and ODK) are extremely reliable. Nevertheless, it is a good practice to have a backup of critical research objects like datasets and code.

7.3.1 What do we envision these backups being used for?

Backups should protect against catastrophic loss of data. Catastrophic loss includes things like losing access to a service (either because the system is down or we cut ties), deletion of a dataset, or deletion of key tables. Backups may be cycled to save on space (e.g. backs are deleted after a certain period of time). In the event of catastrophic loss, it should be possible to restore or reconstitute a dataset from one of these backups.

The snapshot and revision history in features in cloud storage should be sufficient for “time travel” type backups.

7.3.2 What counts as a back-up?

A backup is any copy or representation of the data stored outside of the cloud storage system that allows users to recover the data stored in the system and allows the structure of the full dataset to be reconstructed (e.g. can restore relationships in airtable bases).

Criteria: 1. Data are stored outside the service 2. Data are properly documented with a data dictionary and other metadata 3. Data can serve as a replacement in established research workflows

Some groups may already have versioned backups of data outside of a given service that are not necessarily pulling the whole database but are capturing essential data. Ultimately, if this type of backup is sufficient to meet research goals that is fine.

Some other groups may have a single source of truth that is split across multiple databases/workspaces. Priority should be given to the single source of truth. E.g. A central laboratory database consolidates data from several countries (USA, Canada, and Mexico). Country level mirrors of the database are created (e.g. just the data from the USA) to provide access to the data for that user group. Those country level mirrors do not need to be backed up unless they are being modified in a way that is not reflected in the central database.

7.4 Learn

7.5 Install

Get the tidyverse package for R using install.packages("tidyverse"). This will install several other relevant packages.