Reproducibility

Principles and Practice

Author
Affiliation

Daniela Palleschi

Humboldt-Universität zu Berlin

Published

April 30, 2024

Learning Objectives

Today we will learn about…

  • reproducibility rates in linguistics
  • FAIR principles
  • concepts for building a reproducible workflow

Reproducibility

  • generating the same results with the same data and analysis scripts

    • seems obvious, but requires organisation and forethought

  • bare minimum: share the code and the data (Laurinavichyute et al., 2022)

  • rates of reproducibility vary across fields (Bochynska et al., 2023)

    • open access: 25-65%
    • data and analyses sharing: 11-33%
    • pre-registrations: 0-3%

  • what constitutes “reproducibility”?

What should (ideally) be shared?

  • materials
    • protocols
    • stimuli
    • experiment set-up
  • documentation
    • README
    • metadata
  • data
    • raw
      • e.g., text files, audio, video, or images
    • processed
  • analysis code
    • pre-processing
    • analyses
  • materials are helpful for replication
    • but also for inspection of e.g., design
  • data and code are necessary for reproducibility
    • along with proper documentation of software used

Reproducibility rates in linguistic research

Figure 1: Source: Bochynska et al. (2023), p. 11 (all rights reserved)
  • meta-analysis of 519 randomly sampled articles from various linguistic journales
    • pre- and post-reproducibility crisis (2008/9, 2018/19) (Bochynska et al., 2023)
    • differentiated between primary (collected for study) and secondary (pre-existing) data
  • reported a post-RC increase in shared materials, data, and analyses
    • but still low rates of each
  • higher rates of secondary data sharing, presumably due to publicly available corpora
  • data shared more often than analyses, pre- and post-RC

Journal of Memory and Language

Figure 2: Source: Laurinavichyute et al. (2022), p. 5 (all rights reserved)
  • code and data availability improved
  • but reproducibility rate ranged from 34-56%, depending on criteria
  • higher rates compared to field-wide meta-analysis (Bochynska et al., 2023)

FAIR principles

  • guidelines for sharing digital resources
  • refers broadly to data, but we’ll consider it in terms of analyses
Figure 3: Source: National Library of Medicine (all rights reserved)
  • findable and accesssible refer to where materials are stored
    • in findable repositories
    • that are accessible, i.e., do not require an account
  • interoperable and reusable emphasise the format of data (and code)
    • the importance of future use
    • and use beyond your precise computational environment
  • a great way to test the FAIR principles
    • code review!
    • i.e., have a colleague try to access your data/run your code
      • either via an online repository
      • or send them your project folder

Findable

  • refers to data and supplementary materials

  • materials should have a “persistant identifier”

    • e.g., Digital Object Identifier (DOI) for scholarly articles

  • a digital, long-term storage of data

    • not on a personal or professional website
    • GitHub files don’t typically have sufficient metadata
    • ideally: OSF, Zenodo or some other repository

  • in recent papers, an OSF link is typically provided

  • also: discoverable

    • e.g., in data-specific search engines (Google’s Dataset search)

Accessible

  • data (and code) should be
    • machine- and human-readable
    • available on a trusted repository, e.g., the OSF
    • Open Access
      • not behind a paywall
      • nor require a login

Interoperable

  • data (and code) should
    • not dependent on an operating system
    • nor entirely on software/package versions
  • easiest work around:
    • document your software versions
    • this doesn’t automatically facilitate interoperability
    • but may help pinpoint where problems are coming from

Reusable

  • data (and code) should
    • be reusable for future research
  • data format should be generic
    • i.e., not tied to a specific program
    • for tabular data, I recommend .csv format
  • we can swap with ‘reproducible’ in the context of analyses

Task: finding data

Go to datasetsearch.research.google.com/

  • do a search for data related to a topic of interest to you

  • what type of information does the search provide?

  • what type of links?

  • do you find analysis code, or just data?

  • do the same search at osf.io

  • and at zenodo.org/

    • are there the same amount of hits?

Data and code availability

  • “data available upon (reasonable) request”
    • generally not true
  • data was not available in 68% of the most cited psychology studies (2006-2016) (Hardwicke & Ioannidis, 2018)
    • a further 18% were available with restrictions
    • only 11% available without restriction
  • data alone is not sufficient
    • ‘Data Analysis’ sections are rarely exhaustive/unambiguous
    • very difficult to re-create analyses without code
    • e.g., is data trimming explicitly defined?
      • this will even affect descriptive statistics
Figure 4: Source: Hardwicke & Ioannidis (2018), p. 6 (all rights reserved)

Data and code \(\neq\) Reproducibility

  • even including code does not guarantee reproducibility

  • access to data and code do not mean analyses are reproducible

  • what can go wrong? Examples from Laurinavichyute et al. (2022)

  1. Data problems
    • inaccessible data
    • incomplete data (e.g., 2/3 experiments)
  2. Code problems
    • incomplete code
    • error messages
    • code rot: outdated syntax or environment
    • proprietary software
  1. Documentation problems
    • data difficult to interpret
    • no README file/data dictionary
    • unclear folder/file/variable naming convention
    • manuscript contradicts code
  2. Unclear terms of use
    • no licence specification

Share the code, not just the data

  • Why?
    • key details are often missing from ‘Methods’ sections
  • suggestions for researchers from Laurinavichyute et al. (2022)
  1. Share data in usable form
    • with pre-processing code
  2. Use publicly accessible repositories
    • e.g., OSF
  3. Use non-proprietary data formats
    • e.g., not .xls files (Excel)
  1. Provide documentation
    • e.g., README, data dictionaries
  2. Share code and data
    • they estimate a 38% increase in reproducibility
  3. Teach data management and computing skills
    • that’s what this course is for!

Building a reproducible workflow

  • there are different levels of reproducibility
    • the bare minimum is sharing the code and data
    • and including session information:
      • which operating system was used
      • which software/package versions were used
  • going bigger:
    • project-oriented workflow
    • project-specific filepaths
    • contained in a single project folder
  • we will be using RProjects to achieve this

Project management

  • folder structure
  • project-relative file paths
  • appropriate documentation
    • e.g., README
  • it’s great to map out your project structure early on
    • but it will grow as you go along
    • reproducible principles facilitate adapting as it grows

Literate programming

Instead of imagining that our main task is to instruct a computer what to do, let us concentrate rather on explaining to human beings what we want a computer to do.

— Knuth (1984), p. 97

  • originally used to refer to writing programs
  • but also applies to analysis code
    • especially if we’re aiming for reproducibility

  • main concepts:

    • code is linear (this pre-dates Knuth, 1984)
    • informative but concise commenting

  • main benefits:

    • facilitates maintenance
    • helpful for future-you, collaborators, etc.

Documentation

  • metadata

    • project README
    • codebook/data dictionary

  • README should contain

    • a project description
    • relevant links
    • description of folder structure

  • can be updated as the project develops

  • README.md files in GitHub/Lab are automatically used as a project description

    • .md is a plaintext document
    • uses markdown syntax

Version control

  • git: local tracking
  • useful for the analysis and writing phases
    • but can be tricky for collaboration
  • GitHub/GitLab: remote tracking
    • store your changes to your local git repository
    • then push them to your remote repository
  • safe guards against local hardware/software issues
    • lost or damaged computer or local files
  • and allows for collaboration or sharing

Persistant (public) storage

  • GitHub/Lab are sub-optimal
    • developer-focused
    • typically lack thorough documentation/metadata
    • not very user-friendly for non-users
  • OSF, Zenodo
    • Open Science-focused
    • can be linked to a GitHub/Lab repository
    • facilitate thorough documentation
    • user-friendly

Writing

  • dynamic reports with Markdown syntax

    • e.g., Rmarkdown, Quarto
    • integration of data, code, and prose
      • facilitates cross-referencing within document
      • integration of citation management tools
      • supports LaTeX syntax for example sentences and tables

  • papaja package for APA-formatted Rmarkdown documents

  • challenge: collaboration

    • not all collaborators know these tools
    • track changes not currently possible

Setting up a project

Learning objectives 🏁

Today we learned…

  • reproducibility rates in linguistics ✅
  • FAIR principles ✅
  • concepts for building a reproducible workflow ✅

References

Bochynska, A., Keeble, L., Halfacre, C., Casillas, J. V., Champagne, I.-A., Chen, K., Röthlisberger, M., Buchanan, E. M., & Roettger, T. B. (2023). Reproducible research practices and transparency across linguistics. Glossa Psycholinguistics, 2(1). https://doi.org/10.5070/G6011239
Corker, K. S. (2022). An Open Science Workflow for More Credible, Rigorous Research. In M. J. Prinstein (Ed.), The Portable Mentor (3rd ed., pp. 197–216). Cambridge University Press. https://doi.org/10.1017/9781108903264.012
Hardwicke, T. E., & Ioannidis, J. P. A. (2018). Populating the Data Ark: An attempt to retrieve, preserve, and liberate data from the most highly-cited psychology and psychiatry articles. PLOS ONE, 13(8), e0201856. https://doi.org/10.1371/journal.pone.0201856
Knuth, D. (1984). Literate programming. The Computer Journal, 27(2), 97–111.
Laurinavichyute, A., Yadav, H., & Vasishth, S. (2022). Share the code, not just the data: A case study of the reproducibility of articles published in the Journal of Memory and Language under the open data policy. Journal of Memory and Language, 125, 12.