1[2 {"id":"C1","text":"At h=1 the base-effect-aware random-walk-with-drift model lowers RMSE by about 45% relative to the naive forecast, and the DM test (Bartlett LRV, lag 0) rejects equal accuracy at 5% (p=0.015).","results_keys":["rmse_model_h1","rmse_naive_h1","rel_improvement_rmse_h1","dm_stat_h1","dm_pvalue_h1","conclusion_code_h1"],"status":"supported","caveat":"Only 10 origins in one macro episode; with the HLN small-sample correction p=0.047, only marginally significant."},3 {"id":"C2","text":"At h=3 the model lowers RMSE by about 20% relative to naive, but the difference is not statistically significant (DM p=0.12).","results_keys":["rmse_model_h3","rmse_naive_h3","rel_improvement_rmse_h3","dm_stat_h3","dm_pvalue_h3","conclusion_code_h3"],"status":"supported","caveat":"Long-run variance estimated from 10 overlapping errors; low power."},4 {"id":"C3","text":"The model has the lowest RMSE and MAE among the four methods at both horizons; AR(1) improves only slightly on naive and seasonal naive is worst at h=1.","results_keys":["rmse_model_h1","rmse_ar1_h1","rmse_snaive_h1","rmse_naive_h1","rmse_model_h3","rmse_ar1_h3","rmse_snaive_h3","rmse_naive_h3","mae_model_h1","mae_model_h3","mae_ar1_h1","mae_ar1_h3","mae_naive_h1","mae_naive_h3","mae_snaive_h1","mae_snaive_h3"],"status":"supported","caveat":"Point rankings only; no significance test between baselines and the model other than vs naive."},5 {"id":"C4","text":"First-release YoY growth of the h=1 target months differs from currently published growth by 0.55 pp on average in absolute value, all revisions being downward.","results_keys":["mean_abs_revision_h1_targets"],"status":"supported","caveat":"Signed mean is in results.json supplementary; current vintage includes a large benchmark revision."},6 {"id":"C5","text":"All forecasts are real-time: each uses only the RSAFS vintage dated at its origin, and every target month lies after the last month published in that vintage; all 20 forecast-actual pairs per model are evaluated.","results_keys":["n_origins","first_origin","last_origin","n_analysed"],"status":"supported","caveat":"The gate's generic origin_date < target_date rule fails for h=1 rows by construction of the task's target definition (see report)."}7]8