# Data audit (P1)
- `data/DGS10.csv`: SHA-256 8187f89ab6523b3c964255e1736cfcf0e2ca854ed9e50343e04f4dc2945c5cd7 (matches README). 5217 data rows, 2 columns (`observation_date` str YYYY-MM-DD, `DGS10` percent).
- Date range 2000-01-03 … 2019-12-31; dates unique and strictly increasing; weekdays only (Mon 1044, Tue 1044, Wed 1043, Thu 1043, Fri 1043).
- Empty DGS10: 215 rows (US market holidays, plus e.g. 2001-09-11/12 closures). No non-numeric tokens other than empty.
- Available yields: 5002; range 1.37–6.79 %. Changes: 5001; mean −0.093 bp, sd 5.72 bp, min −51 bp (2009-03-18, FOMC QE announcement — genuine, kept), max +28 bp.
- Anomaly/feature: changes after a holiday span >1 calendar business day (e.g. Tuesday after a Monday holiday); per TASK they are attributed to the weekday of t. Kept, documented.
- `data/README.md`: provenance only, not analysed.
- No input violates the information-availability rules; the only rule-sensitive step is that weekday forecast means use changes dated ≤2009-12-31 only.