Skip to content
NEMAR NEMAR NEMAR

Dataset Submission Standards

Effective date: 2026-08-18.

This page describes the quality bar a dataset must meet to be published on NEMAR (Neuroelectromagnetic Data Archive and Tools Resource), and the metadata we strongly recommend beyond that minimum. The legal side of depositing, including ownership, de-identification, and licensing warranties, is covered by the Data Contributor Terms; this page is about completeness and quality.

We keep this bar deliberately small. It was calibrated against a scan of 577 public OpenNeuro datasets in NEMAR’s modality scope, so each requirement targets a failure mode that actually occurs, rather than aspiration.

Account requirements

Section titled “Account requirements”
  • An ORCID iD is required for every NEMAR account. Your name on NEMAR is taken from your ORCID record and is not freely editable; this keeps authorship attribution canonical across datasets and Digital Object Identifiers (DOIs).
  • Sandbox training is required before your first real upload. The nemar sandbox walkthrough exercises the full upload cycle on a throwaway dataset so mistakes happen there, not on a real deposit.
  • Upload access is granted by an administrator after account review. Signing in alone does not enable uploads.
  • Every upload carries a deposit attestation covering de-identification, the re-identification key status, and (for deposits of licensed data you do not own) an affirmation that the dataset is not already archived on NEMAR or an upstream archive. See the Data Contributor Terms for what each statement means.

Required for publication

Section titled “Required for publication”

A dataset must meet all of the following to be published. Brain Imaging Data Structure (BIDS) validation is checked automatically; the rest is the bar a NEMAR administrator holds every publication request to, and a request that falls short is returned with the reasons.

  • BIDS validation passes with zero errors. Validation runs automatically on upload and again at publication.
  • A dataset_description.json with at least Name and BIDSVersion (BIDS validation already requires this).
  • A descriptive dataset name of at least 25 characters. The title must tell a reader what was studied; the dataset identifier, a bare author-year string, or a placeholder such as DataSet1 is not accepted. If your dataset is known by a short acronym, expand it with a subtitle, for example ACRO: auditory cortex recordings during natural speech.
  • Named authors. You cannot submit anonymously: the Authors field of dataset_description.json must list the people responsible for the dataset, and placeholder entries count as missing.
  • A substantive README that describes this specific dataset: what was recorded, from whom, and how the files are organized. A placeholder, a stub of a few sentences, or boilerplate copied across a family of datasets does not qualify.
  • A declared license. You choose the license, including non-commercial ones, subject to the floor in the Data Contributor Terms: the data must remain freely usable for nonprofit research. The upload tooling helps you pick one and writes the LICENSE file.
  • An ethics approval statement, either in the EthicsApprovals field of dataset_description.json or as a statement in the README. This is the visible counterpart of the warranty in the Data Contributor Terms that the data was collected under appropriate ethical approval.
  • The data files are actually present. A dataset whose metadata references recordings that were never uploaded is not approved.
Section titled “Strongly recommended”

The following are strongly recommended; a missing item may be flagged on the dataset page or raised during review, but does not block publication on its own:

  • An ORCID iD for each author. The upload tooling collects co-author ORCID iDs so every author is unambiguously credited in the dataset’s DOI record.
  • Funding sources, in the Funding field of dataset_description.json. These become funding references in the DOI metadata, which is how funders discover the datasets they supported.
  • Acknowledgments, in the Acknowledgements field of dataset_description.json, for contributions that fall short of authorship.

How the standards are checked

Section titled “How the standards are checked”

Checks run at three points, and the earlier ones exist so you learn about a problem before it costs you time:

  1. At upload, the CLI and the web uploader validate BIDS structure and warn about missing recommended metadata.
  2. At publication request, an automated prescreen reviews the dataset, including the quality judgments above. Its judgment calls are advisory by design; they inform the reviewer, who enforces the required bar, rather than desk-rejecting automatically.
  3. A NEMAR administrator reviews every publication request. Datasets that do not meet the required bar are returned with the specific reasons, and can be resubmitted after revision.

Before a DOI is minted, the dataset’s metadata record must additionally pass validation, including checks on the enriched subject terms. A DOI is permanent, so no dataset acquires one carrying metadata nobody checked.

Mirrored datasets

Section titled “Mirrored datasets”

Datasets imported from OpenNeuro already passed that archive’s own review, so they are exempt from the required bar above; re-litigating an accepted review would block importing datasets that are already public and citable. Imported datasets still receive the advisory flags, so a missing ethics statement or a thin README is visible on the dataset page either way.

Questions

Section titled “Questions”

If you are unsure whether your dataset meets these standards, open an issue on the nemar-cli repository or write to [email protected] before uploading; we would rather help you fix a README than return a publication request.