The paper introduces “confidence horizons,” large-sample confidence sequences restricted to a bounded time horizon, aiming to reduce the conservativeness of methods that must remain valid indefinitely. The authors connect the construction to classical Pocock, O'Brien–Fleming, and Wang–Tsiatis group-sequential boundaries. According to the supplied abstract, closed-form distribution functions allow exact calculation of certain asymptotic quantiles without the repeated numerical integration commonly used in group-sequential analysis. An application considers treatment-effect estimation in sequentially randomized experiments using adaptive Neyman allocation.
No heat snapshots are available in the last 24 hours.