{"id":"arxiv/2603.14835","url":"https://arxiv.org/abs/2603.14835","tags":["math.ST","stat.AP"],"score":0,"title":"Evaluating time-to-event forecasts under fixed-horizon right-censoring","domain":"math.ST","source":"arxiv","authors":["Robert J. Taggart","Nicholas Loveday","Simon Louis"],"reviews":0,"subject":"math.ST","summary":"Time-to-event forecasts frequently arise in applications with finite prediction or decision horizons, including in operational meteorology, hydrology, warranty analysis and insurance. In such settings, forecasts, observations, or both may only be available through their right-censored values relative to a common fixed censoring time. This article develops a framework for evaluating such forecasts. For probabilistic forecasts, recent localization results for proper scoring rules imply coherent evaluation methods based on threshold-weighted versions of familiar scores, including the continuous…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-16","reactions":{}}