New method combines dependent tests via shared permutations, avoiding Bonferroni's conservatism
_onionesque · x · 2026-10-30
A statistical methodology paper on combining dependent tests using shared permutations, sidestepping the excessive conservatism of corrections like Bonferroni.
- Finite-sample error control is guaranteed by exchangeability
- Calibration adapts to the dependence structure among tests
- The paper also presents sequential and learned aggregation variants at the end
Relevant for researchers dealing with multiple-comparison problems in large-scale experiments or model evaluation.
More from Research
- Gary Marcus and coauthors argue human imitation is a misplaced goal for AI — ValerioCapraro · 2026-10-30
- AWS's EasyProver solves 669/672 Putnam problems, plus new harder benchmark Aegis — amazonwebservices · 2026-10-30
- "More Is Different" in neural circuits: compositionality as a lens for interpretability — neurovium · 2026-10-30
- Researchers credit tabpack for making large MLP grid experiments feasible — antoine_chaffin · 2026-10-30
- Lychee-Reranker: decoupled FBNL reranker up to 4B params, strong on BEIR/MIRACL — tomaarsen · 2026-10-30
- Sentence Transformers adds tunable compression: R2 rerankers keep 92% nDCG@10 at 128x — tomaarsen · 2026-10-30