# Data audit

• File: data/panel.csv, SHA-256 968410b993df4d696194ba82b35a030d9c4f5f69e3c2de2c16c7b304d55ce5a8 (matches bundle.json)
• 2000 rows × 3 columns: firm_id (int, 0–49), x (float, dimensionless, range -4.342 to 4.988, mean 0.161, sd 1.325), y (float, dimensionless, range -3.498 to 7.203, mean 1.573, sd 1.782)
• No dates (cross-sectional panel without time index). Missing values: 0 in every column; no non-numeric entries. Exact duplicate rows: 0.
• 50 firms, each exactly 40 observations (balanced).
• Anomalies: none detected. No information-availability violations (single permitted input).
• Exclusions: none (see out/exclusions.csv; n_raw 2000 = n_analysed 2000).