Data audit
| file | SHA-256 | rows | cols |
|---|---|---|---|
| data/gc5y_daily.csv | 2cdda6ccb2147c8861997166cb0d32cf85ff2744e1e7c0f47eded0f354c2db51 (matches README) | 2498 (+header) | 3: date, yield_5y (%), post_cut (int) |
- Date range 2010-01-04 → 2019-12-31; strictly increasing; 0 duplicate dates; 0 missing / non-numeric values.
- Yield range 0.48–3.20 %, mean 1.574 %.
- post_cut: 0 for all 2498 rows; 1 for none. README confirms the Valet query had end_date=2019-12-31. The event (2020-03-04) is outside coverage → estimand not computable. Decision: premise_invalid; do not redefine post_cut nor use a proxy event.
- Calendar gaps: 1962 ×1 d, 10 ×2 d, 440 ×3 d, 79 ×4 d, 6 ×5 d (Christmas/New Year). Normal for business-day series; kept.
- One-day changes (2497): mean −0.0429 bp, sd 4.24 bp, min −27 bp (2011-08-08, plausible market move after the US downgrade; kept), max +17 bp.
- No input violates the availability rules (data end before the event date).