A0466
Title: Efron-Stein type inequalities for randomly stopped processes
Authors: Victor H de la Pena - Columbia University (United States) [presenting]
Abstract: Efron-Stein type inequalities are established for randomly stopped processes, extending classical concentration results to settings with data-adaptive stopping times. The main result provides variance bounds for fixed functions of independent random variables evaluated at random stopping times that are adapted to the data but independent of the target function values. It carefully distinguishes between trivial cases (where stopping times are independent of the entire process) and genuinely non-trivial applications. The theoretical foundation relies on recent refinements of maximal inequalities for randomly stopped sums.
