Data for Data Science Week 4: Data manipulation and reproducibility
Today’s objectives
At the end of today’s class, you should be able to:
Maintain records of how data have been manipulated or altered
Plan ahead for publishing/archiving data
Identify tools and support at the Hutch for multiple parts of the data life cycle
Completing the data life cycle
Funding agencies and scientific publishers require us to submit our data to repositories. How do we plan ahead to make sure this doesn’t simply meet requirements, but also maximizes the use of data later? Primer on Data Management
How might data change during the course of a project?
Discovery (discovery.fredhutch.org)
Maintaining data integrity See this lesson for a simple model for applying quality control and assurance Quality assurance: require data to be entered appropriately Quality control: checking for erroneous data
Specify in your metadata what efforts are made at quality assurance! Look here for some ideas of best practices
Cleaning large-scale datasets Example tool: OpenRefine (lessons here) Clean, transform, and extend large, messy datasets Allows export of clean data AND reproducible methods used for cleaning
Version control
Tracking changes and additions/deletions to documents and files Can be used for data files (during the course of a project) Provides a record of what has changed Allows reversal back to earlier versions See the wiki entry for more information See the fredhutch.io course Intro to Git and GitHub!
REDCap (redcap.fredhutch.org) REDCap is a secure web platform for building and managing online databases and surveys. See the wiki for information on office hours and the user group!
LabKey (redcap.fredhutch.org) Join the Coop Community Slack to participate in the user group! Lab-based data repositories, data showcase, and electronic laboratory
Open Science Framework (osf.io)
Using and choosing licenses What license does Data Dryad use? Licenses dictate how you can use other people’s data, and how other people can use your data. Shaddim; original CC license symbols by Creative Commons CC BY 4.0
Ten Simple Rules for the Care and Feeding of Scientific Data “This article offers a short guide to the steps scientists can take to ensure that their data and associated analyses continue to be of value and to be recognized.”
Also see this quickstart guide Data Dryad Best Practices for creating reusable data publications The big picture
GEO example Look at the GEO metadata template document. When should researchers start filling in the information? Who needs to be involved in documenting the data?
FAIR data principles Findable, Accessible, Interoperable, and Re-usable
Most data management focuses on humans being able to reuse data. FAIR principles assist automated searching and reuse of data. “Data” includes algorithms, tools, and workflows! (FAIR article)
FAIR data: Findable F1. (meta)data are assigned a globally unique and eternally persistent identifier. F2. data are described with rich metadata. F3. (meta)data are registered or indexed in a searchable resource. F4. metadata specify the data identifier.
FAIR data: Accessible A1 (meta)data are retrievable by their identifier using a standardized communications protocol. A1.1 the protocol is open, free, and universally implementable. A1.2 the protocol allows for an authentication and authorization procedure, where necessary. A2 metadata are accessible, even when the data are no longer available.
FAIR data: Interoperable
I1. (meta)data use a formal, accessible, shared, and broadly applicable language for knowledge representation. I2. (meta)data use vocabularies that follow FAIR principles. I3. (meta)data include qualified references to other (meta)data.
FAIR data: Re-usable R1. meta(data) have a plurality of accurate and relevant attributes. R1.1. (meta)data are released with a clear and accessible data usage license. R1.2. (meta)data are associated with their provenance. R1.3. (meta)data meet domain-relevant community standards.
Summary Plan ahead for data management to save time and improve the quality of your research Document your metadata early and often Find tools to help automate and check your work Planning for reproducibility will help you and the broader scientific community
Best practices will vary depending on your particular field and project. Feel free to ask Hutch Data Commonwealth (Scientific Computing, The Coop) for assistance!
What are the biggest questions you still have about data management?
What resources would help you further?
Fred Hutch Data Science Wiki: Data generation Scientific Computing