[
{"id":"C1","text":"Over 35 evaluated origins at h=1, the AR(2)-on-differences model has higher RMSE than the naive forecast (relative improvement about -2.3%).","results_keys":["rmse_naive_h1","rmse_model_h1","rel_improvement_rmse_h1","n_evaluated_h1"],"status":"supported","caveat":"Final-vintage data; small sample."},
{"id":"C2","text":"At h=3 the model's RMSE is also slightly higher than naive (relative improvement about -0.7%).","results_keys":["rmse_naive_h3","rmse_model_h3","rel_improvement_rmse_h3","n_evaluated_h3"],"status":"supported","caveat":"Difference not statistically significant (DM p=0.125)."},
{"id":"C3","text":"MAE gives the same ranking as RMSE at both horizons (naive better).","results_keys":["mae_naive_h1","mae_model_h1","mae_naive_h3","mae_model_h3"],"status":"supported","caveat":""},
{"id":"C4","text":"The h=1 loss difference is statistically significant (DM p=0.017); the h=3 difference is not (p=0.125).","results_keys":["dm_stat_h1","dm_pvalue_h1","dm_stat_h3","dm_pvalue_h3"],"status":"supported","caveat":"Asymptotic test, n=35, HLN-corrected; economically negligible magnitudes."},
{"id":"C5","text":"One target month (2025-10-01) is missing, leaving 35 evaluated pairs per horizon out of 36.","results_keys":["n_evaluated_h1","n_evaluated_h3","n_origins","data_issues"],"status":"supported","caveat":""},
{"id":"C6","text":"Results do not establish real-time forecasting performance.","results_keys":["data_issues"],"status":"not_evaluated","caveat":"Only the final vintage is available."}
]