# Data audit (P1)
| file | sha256 | rows | cols |
|---|---|---|---|
| data/design.csv | 2f459c539378cbe54d830c18af466f2fa4230ce48a609aa6464f8334a0d36ecb | 7 (+header) | 3 (parameter, value, description) |

Hash matches bundle.json. Parameters: n=30 (series length y_1..y_n, so 29 regression observations), rho=0.8, sigma=1.0, intercept=yes, initial_condition=stationary (y_1 ~ N(0, sigma²/(1-rho²))), n_replications=20000, seed=20261029. No missing values, no duplicate parameters, no anomalies; |rho|<1 so the stationary initial condition is well defined. Units: dimensionless (sigma in units of y). No dates (simulated design).
Availability: the only input is design.csv; nothing violates the rules. Note: "n" in -(1+3rho)/n is taken as design n=30 per TASK ("at the design values"); a sensitivity with n-1=29 is in out/diagnostics.json.
Exclusion ledger: out/exclusions.csv — one rule (non-finite rho_hat), 0 removed; 20000 raw = 20000 analysed.