Skip to content
neuroflow0.2.31 · docs

← neuroflow

phase-preregistration

The preregistration phase produces and manages pre-registration documents that commit the research team to a specific hypothesis, design, and analysis plan before data collection begins.

Approach

  • Identify which mode applies (draft, review, deviation log, link registered report, freeze) before starting
  • Always pull the research question and hypothesis from .neuroflow/ideation/ if it exists; pull paradigm details from .neuroflow/experiment/ if it exists — do not ask the user to repeat information already in project memory
  • Ask which registry or template applies before drafting; structure adapts significantly between OSF and AsPredicted
  • In all pre-registration drafts: hypotheses must be directional and falsifiable, analysis plans fully specified, and exploratory analyses clearly distinguished from confirmatory ones
  • If a power calculation is missing, prompt the user to provide one before saving the document
  • Every draft ends with the prereg-parameters YAML block (planned N, stopping rule, alpha, filters, windows, ROIs, tests, exclusions) that restates the plan in machine-readable form
  • After data collection begins, use the deviation log mode to record any plan changes — never silently edit a submitted pre-registration document
  • Freeze the final document with scripts/freeze.py only after the person explicitly confirms in the same turn; never freeze or unfreeze on your own

Frozen preregistration files are never edited; changes go to deviations.md. A file listed under files: in .neuroflow/preregistration/status.md with status: frozen carries a > FROZEN … banner as its first line. Unfreezing is a person's action and is itself logged in deviations.md.

Relevant skills

  • neuroflow:neuroflow-core — read first; defines the command lifecycle and .neuroflow/ write rules (integrity contract C2 for status.md)

Workflow hints

  • prereg-[registry]-[date].md — the primary pre-registration document; version-control this file
  • prereg-review-[date].md — completeness and consistency review report
  • status.md — the freeze record (frontmatter: status, frozen_at, files with SHA-256 hashes, registry, doi, planned_n, set_by, set_at); written by freeze.py, never by hand
  • deviations.md — running log of post-registration deviations; append only, never overwrite past entries
  • registered-report.md — metadata for any submitted or accepted registered reports (DOI, stage, date)
  • Pre-registration documents are the scientific record — flag any ambiguity that could allow selective reporting or HARKing (Hypothesizing After Results are Known)
  • If the user is drafting for a registered report journal submission, note that stage 1 acceptance is typically contingent on the analysis plan being fully specified

Scripts

  • scripts/freeze.py — hash, freeze, verify, unfreeze for the preregistration hash lock. Hashes ignore the banner, a byte-order mark and CRLF vs LF, so Windows and macOS checkouts agree. python <skill base dir>/scripts/freeze.py verify — exit 0 = unchanged or not frozen, 1 = findings (changed, missing, banner removed, freeze not set by a person), 2 = usage error or refusal. Other checks (/sentinel, the neuroflow mod) call this script rather than re-implement the hash.