Contingency Table

Compute the 2x2 forecast verification contingency table (TP/FP/FN/TN) and standard skill scores (Hit Rate, False Alarm Ratio, Hanssen-Kuipers, acting-in-vain) from aligned binary forecast and observation timeseries. TP (Hit): forecast=1 AND observed=1 — alarm raised, event happened. FP (False Alarm / "acting in vain"): forecast=1 AND observed=0. FN (Miss): forecast=0 AND observed=1. TN (Correct Rejection): forecast=0 AND observed=0. Hit Rate (POD) = TP / (TP + FN); fraction of events correctly forecast. False Alarm Ratio = FP / (TP + FP); fraction of alarms that were false. Hanssen-Kuipers (TSS) = HR - FP / (FP + TN); discrimination skill, [-1, 1]. Acting-in-vain rate = FAR (same arithmetic; published vocabulary for the anticipatory-action / forecast-based-financing field — Anticipation Hub 2024 Global Overview, IFRC/DRK/Climate Centre FbF Practitioner Manual). Both names are exposed so AA recipes can read either. The standard product-evaluation primitive for anticipatory action triggers.

IDevaluation_contingency_table
Domaintemporal
Categoryevaluation
Typetimeseries-to-record
Version1.0.0
view yaml ◦view github ◦