1[2 {"id":"C1","text":"The classical one-way ANOVA does not reject equal weekday means of daily 10y yield changes (F=0.263, p=0.90).","results_keys":["anova_f_stat","anova_p_value","n_analysed"],"status":"supported","caveat":"Equal-variance assumption rejected by Levene; independence assumed."},3 {"id":"C2","text":"The Monday contrast is +0.05 bp with 95% Welch CI [-0.32, 0.42] bp, indistinguishable from zero.","results_keys":["effect_estimate","effect_ci_low","effect_ci_high"],"status":"supported","caveat":"Small effects below ~0.4 bp cannot be ruled out; Monday changes include weekends."},4 {"id":"C3","text":"Weekday-specific means estimated on 2000-2009 do not improve 2010-2019 forecasts over a zero-change forecast (RMSE 4.884 vs 4.876 bp).","results_keys":["model_metric","baseline_metric","rmse_ratio_model_over_baseline","n_train","n_test"],"status":"supported","caveat":"Single split; no formal forecast-comparison test."},5 {"id":"C4","text":"The no-effect conclusion is robust to HC1/Newey-West intervals for the Monday contrast, a Kruskal-Wallis test, and excluding the 2008-09 crisis window.","results_keys":["robust_effect_ci_hc1","robust_effect_ci_nw5","robust_kruskal_wallis_p","robust_anova_p_excl_crisis_2008_09_to_2009_06","n_excl_crisis"],"status":"supported","caveat":"Joint test not made HAC/Welch-robust."},6 {"id":"C5","text":"Weekday variances differ significantly, so the classical ANOVA p-value is approximate.","results_keys":["robust_levene_p"],"status":"supported","caveat":"Levene test also assumes independence."},7 {"id":"C6","text":"Overall conclusion: no detectable day-of-week effect.","results_keys":["conclusion","anova_p_value"],"status":"supported","caveat":"Applies to the stated estimands and period only."}8]9