Distribution-free two-sample testing with blurred total variation distance

Rohan Hore, Rina Foygel Barber · 2026

Read Distribution-free two-sample testing with blurred total variation distance by Rohan Hore, Rina Foygel Barber (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-05862.txt.

Description

Deciding whether two samples come from the same distribution, with no assumptions at all, runs into a hard wall: certifying equality is impossible in a distribution free regime. Rohan Hore and Rina Foygel Barber study a relaxation that makes the problem tractable, the blurred total variation distance. Blurring smooths the classic distance just enough to allow inference.

The authors provide distribution free upper and lower bounds on the blurred distance, with theoretical guarantees. They also examine how the quantity behaves in high dimensions, where naive distances lose meaning. The long paper, nearly fifty pages, develops the properties of the relaxation in detail. The result is a testing framework that stays honest about what samples can prove. It should interest statisticians who need two sample tests without modeling assumptions. The work reframes an impossibility into a measurable quantity, at the price of answering a slightly softer question.

Catalog record

CREATORRohan Hore, Rina Foygel Barber
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE116 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-05862

Browse the full library · About the Tower of Babel