Why additional rows matter

This project is most sensitive where the sample is thinnest: the post-interview tail, recent interview cohorts, and slow approvals that remain open well past the common 0-14 day window. A small number of new rows can materially change the interpretation of whether RDU is speeding up, whether a given interview month looks anomalous, and how aggressive the pending-case probabilities should be.

Canonical contribution path

The repository itself is the canonical public dataset. New rows should be submitted through reviewed pull requests, not ad hoc edits to generated artifacts.

Primary files:

  • data/canonical/rdu_timeline_data.csv
  • data/provenance.csv
  • CONTRIBUTING.md

Minimum row needed for analysis

The canonical data contract requires the following fields:

  • user
  • field_office
  • i-485 receipt date
  • interview date
  • i-130 approval date
  • i-485 approval date
  • receipt to interview
  • interview to i130
  • interview to i485
  • i130 to i485
  • days since interview
  • days total
  • case closed

If a row is added manually, the derived duration and status fields must remain internally consistent so that validation passes. The exact validation gates are documented in Data Contract.

Highest-value additions right now

  • post-interview cases still pending beyond 30 days
  • approvals that closed after an unusually long post-interview lag
  • recent 2026 interviews, which help distinguish transient clustering from a persistent slowdown
  • older open cases with verified receipt and interview dates, which stabilize the tail fit

Metadata the current schema does not yet preserve

The current canonical table does not include a formal case type field or a free-text narrative field. If that metadata matters for interpretation, record it in data/provenance.csv or propose a schema change explicitly rather than silently overloading existing columns.

Data policy

Row-level usernames and timelines in this repository are intentionally public. Do not submit private material or off-platform personal data without consent. Keep field_office explicit; the published site currently filters to Raleigh/Durham.