1[2 {"id":"C1","text":"Over the 40 origins the seasonal-AR model has lower RMSE and MAE than the seasonal-naive forecast at h = 1, 2 and 4, with relative RMSE improvements of about 23%, 13% and 19%.","results_keys":["rmse_snaive_h1","rmse_model_h1","rmse_snaive_h2","rmse_model_h2","rmse_snaive_h4","rmse_model_h4","mae_snaive_h1","mae_model_h1","mae_snaive_h2","mae_model_h2","mae_snaive_h4","mae_model_h4","rel_improvement_rmse_h1","rel_improvement_rmse_h2","rel_improvement_rmse_h4"],"status":"supported","caveat":"Point estimates on one 40-quarter window; no formal test at h = 1, 2."},3 {"id":"C2","text":"At h = 4 the mean squared-error loss differential (2.82 squared pp in favour of the model) is not statistically distinguishable from zero: Bartlett HAC SE with lag 3 gives DM = 1.35, two-sided p = 0.18.","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":"Small sample with overlapping errors; p ranges 0.04-0.23 across HAC choices in sensitivity_h4; pre-specified choice reported."},4 {"id":"C3","text":"The model's squared 4-step error was smaller than the seasonal-naive one at 29 of the 40 origins.","results_keys":["n_loss_diff_positive_h4"],"status":"supported","caveat":"Descriptive count; not a test."},5 {"id":"C4","text":"Forecasts use only quarters up to and including each origin; the last estimation sample has 121 observations.","results_keys":["n_analysed","first_origin","last_origin","n_origins"],"status":"supported","caveat":"Current-vintage data; no real-time claim."}6]7