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

Download:

Current browse context:

cs

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

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

Statistics > Methodology

Title: Two-Sample Tests that are Safe under Optional Stopping, with an Application to Contingency Tables

Abstract: We develop E variables for testing whether two data streams come from the same source or not, and more generally, whether the difference between the sources is larger than some minimal effect size. These E variables lead to tests that remain safe, i.e. keep their Type-I error guarantees, under flexible sampling scenarios such as optional stopping and continuation. In special cases our E variables also have an optimal `growth' property under the alternative. We illustrate the generic construction through the special case of 2x2 contingency tables, where we also allow for the incorporation of different restrictions on a composite alternative. Comparison to p-value analysis in simulations and a real-world example show that E variables, through their flexibility, often allow for early stopping of data collection, thereby retaining similar power as classical methods.
Subjects: Methodology (stat.ME); Machine Learning (cs.LG); Statistics Theory (math.ST)
Cite as: arXiv:2106.02693 [stat.ME]
  (or arXiv:2106.02693v2 [stat.ME] for this version)

Submission history

From: Peter Grünwald [view email]
[v1] Fri, 4 Jun 2021 20:12:13 GMT (1609kb,D)
[v2] Sat, 9 Oct 2021 09:39:51 GMT (649kb,D)

Link back to: arXiv, form interface, contact.