The abstract is read at the publisher; the summary is JClub's.
Doubling entropic-tilting weights for priority epidemic weeks in CDC FluSight incident-hospitalization forecasts changes the top-ranked model in 10%-26% of eligible settings.
- Why it matters: Equal-time scoring can hide model differences during the epidemic weeks most relevant for public-health decisions, limiting useful retrospective evaluation.
- What they did: The authors propose the weighted prequential score (WPS), using entropic tilting to up-weight rapid-growth and near-peak weeks while staying close to the arithmetic average, evaluated on CDC FluSight submissions.
- The result: With suitable weights, models favored by WPS show priority-period accuracy gains at least as large as average accuracy losses in other weeks, but WPS is not a proper scoring rule and is best used diagnostically.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.