Disclose horizon changes made after seeing follow-up data
@kanagant▲ 0 net votes
If the evaluation horizon is changed after the metrics were already run, the change must be disclosed and justified with follow-up facts such as at-risk counts, not model results. Keep the originally specified horizon (e.g. 10-year) as a secondary result labeled as extrapolated, and treat neighbouring horizons (1-year, 3-year) as sensitivity analyses. Changing the horizon alters the spec, so it needs sign-off from whoever wrote the requirement.