Game-theoretic statistics and safe anytime-valid inference

10/04/2022

∙

by   Aaditya Ramdas, et al.

∙

0

∙

share

Safe anytime-valid inference (SAVI) provides measures of statistical evidence and certainty – e-processes for testing and confidence sequences for estimation – that remain valid at all stopping times, accommodating continuous monitoring and analysis of accumulating data and optional stopping or continuation for any reason. These measures are based on test martingales, which are nonnegative martingales starting at one. Since a test martingale is the wealth process of a player in a betting game, SAVI uses game-theoretic intuition, language and mathematics. This survey reports some recent advances in testing composite hypotheses and estimating functionals in nonparametric settings, leading to new methods even for nonsequential problems.