We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math.ST

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Statistics Theory

Title: High Probability Lower Bounds for the Total Variation Distance

Abstract: The statistics and machine learning communities have recently seen a growing interest in classification-based approaches to two-sample testing. The outcome of a classification-based two-sample test remains a rejection decision, which is not always informative since the null hypothesis is seldom strictly true. Therefore, when a test rejects, it would be beneficial to provide an additional quantity serving as a refined measure of distributional difference. In this work, we introduce a framework for the construction of high-probability lower bounds on the total variation distance. These bounds are based on a one-dimensional projection, such as a classification or regression method, and can be interpreted as the minimal fraction of samples pointing towards a distributional difference. We further derive asymptotic power and detection rates of two proposed estimators and discuss potential uses through an application to a reanalysis climate dataset
Subjects: Statistics Theory (math.ST); Methodology (stat.ME)
Cite as: arXiv:2005.06006 [math.ST]
  (or arXiv:2005.06006v4 [math.ST] for this version)

Submission history

From: Jeffrey Näf [view email]
[v1] Tue, 12 May 2020 18:55:05 GMT (4692kb,D)
[v2] Fri, 15 May 2020 10:38:44 GMT (4804kb,D)
[v3] Fri, 14 May 2021 17:04:46 GMT (2809kb,D)
[v4] Wed, 2 Jun 2021 17:19:29 GMT (2825kb,D)

Link back to: arXiv, form interface, contact.