# Contract — F6-conf-01

Question. Does weekday carry predictive content for the one-day change in the 10y Treasury CMT yield (DGS10), 2000–2019? Estimands. (i) joint equality of the five weekday conditional means of Δy_t (bp), classical one-way equal-variance ANOVA F-test on all changes; (ii) Monday contrast = mean(Mon) − mean(Tue–Fri), all changes, with 95% CI (I choose Welch two-sample); (iii) OOS RMSE over 2010–2019 of weekday-mean forecast (weekday means from changes dated ≤2009-12-31) vs zero-change forecast. Assumptions: changes approx. independent with finite variance; ANOVA additionally assumes equal variances (stated in task as the test to use); stationarity of weekday means across periods for (iii) to be meaningful. Data rules. Drop empty DGS10 rows first; Δy_t = 100(y_t − y_{t−1}) between consecutive available observations, dated t, weekday of t; first obs has no change. Train: changes dated ≤ 2009-12-31 (note: the first 2010 change uses the last 2009 level — allowed, since it is dated 2010). Test: ≥ 2010-01-01. Only DGS10.csv. Outputs. out/results.json keys: n_missing_rows, n_train, n_test, anova_f_stat, anova_p_value, effect_estimate, effect_ci_low, effect_ci_high, baseline_metric, model_metric, conclusion, task_id, feasibility, feasibility_reason, data_issues, n_input_rows, n_analysed, random_seed. Also figure1.png, table1.csv (weekday,n,mean_bp,se_bp), report.md (6 sections), analysis.py, claims.json. Conclusion rule. ANOVA p≥0.05 → no_detectable_effect; else sign from Monday CI (positive/negative; if CI covers 0 despite ANOVA rejection → no_detectable_effect for the Monday contrast, documented). Invalidity. Missing/unparsable data making changes uncomputable; using test-period info in forecasts.