1[2 {"id":"C1","text":"The dsar_bic model has lower out-of-sample RMSE and MAE than the seasonal-naive forecast at h=1, 2 and 4 over the 40 origins (RMSE reductions of about 17-23%).","results_keys":["rmse_snaive_h1","rmse_model_h1","mae_snaive_h1","mae_model_h1","rmse_snaive_h2","rmse_model_h2","mae_snaive_h2","mae_model_h2","rmse_snaive_h4","rmse_model_h4","mae_snaive_h4","mae_model_h4","rel_improvement_rmse_h1","rel_improvement_rmse_h2","rel_improvement_rmse_h4"],"status":"supported","caveat":"Point estimates only at h=1,2; no significance test performed there."},3 {"id":"C2","text":"At h=4 the mean squared-loss differential (snaive minus model) is positive, 2.76 squared pp, with Bartlett HAC (lag 3) SE 2.15, DM 1.29, two-sided p 0.20: not statistically distinguishable from zero.","results_keys":["loss_diff_mean_h4","loss_diff_se_h4","hac_lags_h4","hac_kernel_h4","dm_stat_h4","dm_pvalue_h4"],"status":"supported","caveat":"T=40 overlapping errors; asymptotic normal approximation rough."},4 {"id":"C3","text":"The h=4 non-significance holds across the HAC variants considered (Bartlett lags 3,4,6; rectangular lag 3 with and without HLN correction), with p-values from about 0.05 to 0.20.","results_keys":["dm_pvalue_h4","supplementary"],"status":"supported","caveat":"Bartlett lag 6 is borderline (p about 0.052)."},5 {"id":"C4","text":"At h=4 BIC selected no lags at every origin, so the model reduces to expanding-window seasonal means at that horizon.","results_keys":["model_description","supplementary"],"status":"supported","caveat":""}6]7