NeurIPS 2020

Confidence sequences for sampling without replacement

Meta Review

This paper develops confidence sequences in the setting of sampling without replacement. I want to highlight some particular strengths of the paper. (*) The reviewers are in agreement that the writing is excellent. (*) The work is theoretically strong and practically relevant. (And yet the authors are also very clear to the reader what they are accomplishing in each bit; the prose is easy to follow.) (*) The theory approach is novel and interesting in its own right. (*) The authors start with great motivating examples (election opinion surveys, permutation tests, Shapley values, understanding the effect of interventions).