# Data audit (P1)

• File: data/design.csv — SHA-256 2f459c539378cbe54d830c18af466f2fa4230ce48a609aa6464f8334a0d36ecb
• 7 data rows (+header), 3 columns: parameter (str), value (str/numeric), description (str).
• Values: n=30, rho=0.8, sigma=1.0, intercept=yes, initial_condition=stationary (y_1 ~ N(0, sigma^2/(1-rho^2))), n_replications=20000, seed=20261029.
• No missing values, no duplicate parameters, no dates (simulation design only). Units: dimensionless.
• Consistency: |rho|<1 so the stationary initial condition is well defined; sigma>0; n>=3.
• Information-availability: only input is design.csv; no violations. c=0 per TASK.md.
• Note (not a data defect): n=30 is the series length, so the regression uses n-1=29 pairs. The analytical formula is evaluated with n=30 as the task specifies ("at the design values"); the n-1 version is reported as a sensitivity key (bias_analytical_neff).
• Exclusions: one rule (drop non-finite rho_hat), removed 0 of 20000 (out/exclusions.csv).