Sequential Estimation Methods from Inclusion Principle
This work addresses the problem of ensuring rigorous confidence guarantees in sequential estimation for researchers and practitioners in statistics and machine learning, representing an incremental improvement over existing asymptotic methods.
The authors tackled the problem of sequential estimation by proposing new methods based on the inclusion principle, which reformulate estimation as constructing sequential random intervals and use confidence sequences to control coverage probabilities, resulting in procedures that rigorously guarantee pre-specified confidence levels, unlike existing asymptotic methods.
In this paper, we propose new sequential estimation methods based on inclusion principle. The main idea is to reformulate the estimation problems as constructing sequential random intervals and use confidence sequences to control the associated coverage probabilities. In contrast to existing asymptotic sequential methods, our estimation procedures rigorously guarantee the pre-specified levels of confidence.