Verification checks (P4)
Clean re-run: deleted results.json/table1.csv/figure1.png/exclusions.csv, re-ran analysis.py → results.json byte-identical (cmp). PASS
- n_train+n_test==n_analysed (2500+2501=5001): PASS
- table1 n sums to n_analysed: PASS
- exclusions: 5217 − (215+1) = 5001 = n_analysed: PASS
- n_missing_rows independently recomputed (pandas NaN count = 215): PASS
- CI contains estimate: PASS
- All SEs positive: PASS
- Monday contrast recomputed from table1 (n-weighted Tue–Fri mean): PASS
- ANOVA p in [0,1]: PASS
- conclusion consistent with ANOVA p ≥ 0.05 rule: PASS
- RMSE magnitude plausible vs. sd of daily changes (~5 bp): PASS
- Baseline RMSE independently recomputed: PASS
- Information rule: first evaluation change dated 2010-01-04; weekday means use train only (code: groupby on train): PASS
- Figure visually inspected: five weekday means with 95% bars, all covering zero: PASS
- Assumption check: Levene test rejects equal variances (p≈1e-7) → classical ANOVA assumption violated; Welch ANOVA (p≈0.90) and Kruskal–Wallis (p≈0.52) agree with classical result. Reported, not a failure of the computation.