Skip to contents

2. Open data

2022-12-06

Source: vignettes/v2-open-data.Rmd
v2-open-data.Rmd

🎯 GOALS

Learning how Open Data can help making research more transparent and sustainable, plus getting an overview over one example of an open EEG dataset (ERP CORE).

2.1 Research we can trust

Figure: The replication crisis in psychology.1

  • We want (psychological) research findings to be trustworthy

  • At the very least, other researchers that repeat an experiment or analysis should be able to get the same results

Figure: Terminology for trustworthy research.2

  • Sharing research data helps to establish reproducibility and robustness

  • Pre-registering one’s analysis and sharing materials and protocols helps to establish replicability and generalizability

2.2 Open research data

Advantages of Open Data Disadvantages of Open Data

  • Others can reproduce your analysis

  • Others can probe the robustness of your results

  • Others can analyze their own research questions

  • Others can use it for methods development or teaching

  • You have a (well-documented) backup of the data

  • You will get cited when others use your data

  • Study participants and taxpayers get more for their time and money

  • Privacy concerns → Informed consent, pseudonymization

  • Additional effort in preparing and documenting the data

  • Fear of getting “scooped”

2.3 Standardizing data

  • Brain Imaging Data Standard (BIDS) = a standardized way of organizing neuroimaging (e.g., fMRI, EEG) data for easier sharing and reuse

Figure: Example EEG dataset in BIDS format.3

  • Folders organized by subjects (sub-001/, sub-002/, etc.), session (ses-01, ses-02, etc., if applicable) and imaging modality (T1w – structural MRI, func – functional MRI, eeg – EEG recordings, etc.)

  • Raw data files (e.g., .set/.fdt for EEG) are accompanied by metadata in .json format

    • Meta-data contains info about the task, the scanner, etc.

  • Additional dataset-wide files like dataset_description.json, participants.tsv, README

2.4 ERP CORE dataset

  • ERP Compendium of Open Resources and Experiments (ERP CORE)4 = an open dataset with data from six different EEG experiments

  • Each experiment was optimally designed to elicit a certain ERP component (more on that later)

Figure: Tasks in the ERP CORE dataset.5

  • 40 participants (25 female, 15 male, age 21.5 ± 2.9 years)

  • 30 EEG electrodes (10/20 system) + 3 EOG electrodes (eye movements)

  • Sampling rate 1024 Hz (= EEG samples per second)

2.5 Downloading data

  • Manually from the Open Science Framework (https://osf.io/thsqg/) or using the R package erpcore6

  • Start by creating an empty data directory

    • here() function from the here package creates relative file paths, hence reproducible on different machines and operating systems
(data_dir <- here::here("data/n170"))
## [1] "/home/runner/work/eegSeminaR/eegSeminaR/data/n170"
dir.create(data_dir, recursive = TRUE)
  • erpcore package is still in development → Must be installed from GitHub instead of CRAN
install.packages("remotes")
remotes::install_github("alexenge/erpcore")

  • Download the ERP CORE data using erpcore:

    • component = "n170" to get only the data from the face perception (N170) experiment

    • type = "bids" type to get data in BIDS format

    • Only download one of the 40 subjects for now

erpcore::get_erpcore(
  component = "n170",
  dest_path = data_dir,
  type = "bids",
  subjects = "sub-001"
)

✍️ EXERCISE

Open the data/n170 directory in the file explorer and take a look at the different files that were downloaded. Can you figure out the purpose for all of them?

Further reading

  • Gewin, V. (2016). Data sharing: An open mind on open data. Nature, 529(7584), Article 7584. https://doi.org/10.1038/nj7584-117a

  • Kathawalla, U.-K., Silverstein, P., & Syed, M. (2021). Easing into Open Science: A guide for graduate students and their advisors. Collabra: Psychology, 7(1). https://doi.org/10.1525/collabra.18684

  • Pernet, C. R., Appelhoff, S., Gorgolewski, K. J., Flandin, G., Phillips, C., Delorme, A., & Oostenveld, R. (2019). EEG-BIDS, an extension to the Brain Imaging Data Structure for electroencephalography. Scientific Data, 6(1), Article 1. https://doi.org/10.1038/s41597-019-0104-8

Add-on topics

2.6 Version control

  • For full transparency and reproducibility, we need to share not just our data but also our code

  • Ideally not just the final version (e.g., upload script on OSF), but the full history

  • Also helpful for your future self to understand what you did a couple of months ago

  • Version control tools like git are very powerful for this:

    1. Make your project folder a git repository

    2. Write/edit code

    3. Take a snapshot (“commit”) of the code if you’re happy with it

    4. Include a helpful commit message for your future self or collaborators

    5. Try out new stuff while always being able to revert to any previous commit

Figure: Tracking the history of your code using git commits (circles) and branches (arrows)7

  • Code in git repositories can be shared online on websites like GitHub or GitLab

  • Details are beyond the scope of this course, but for a great intro see “Happy git with R”8