# Data audit (P1)

• File: data/design.csv, SHA-256 b1a17a529973e7994bbedad5ec1bc73fe968c777301ca8745657ea300e3c74f0 (matches bundle.json).
• 30 rows x 3 columns: i (int, 1..30, unique, consecutive), x (float, dimensionless, min -1.3942, max 1.9919, mean 0.2823, sd 0.8212), sigma (float, dimensionless, min 0.6034, max 1.7903, all > 0).
• No date columns (synthetic fixed design). Missing values: 0. Duplicate rows: 0. Rows sorted by x.
• Anomalies: none. Note (feature, not defect): sigma increases with |x - ~0| (smallest near x≈0, largest at extreme x, esp. x=1.99) — heteroskedasticity linked to leverage, which is the setting where conventional SE is expected to be biased.
• Information-availability: only input is design.csv, allowed. No violations.
• Exclusions: none (all 30 rows used); see out/exclusions.csv.