Graph-Based Framework for Multivariate Independence Testing

A 1.8 MB research publication by Yanhui Li, last updated in April 2026, proposing a new nonparametric framework for testing independence between two sets of variables. The method uses weighted and unweighted graph representations and is designed for data where only pairwise distances are observed. Simulation studies suggest the methods control type I error rates and exhibit higher power than competing approaches.

Use Cases

• Testing independence between two variable sets based on the proposed graph-based framework.
• Analyzing data where only pairwise distances are observed, as mentioned in the description.
• Comparing the power of independence tests against competing methods.
• Applying a nonparametric test that maintains high efficiency across diverse alternatives.

Strengths

• Includes theoretical properties ensuring applicability to distance-only data.
• Simulation studies suggest controlled type I error rates and higher power.
• Released under a permissive CC-BY-4.0 license.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The primary files are TXT, PDF, and ZIP formats; the core dataset structure is unclear.

Provenance

Source

figshare

Collection Method

Proposed methodological framework, likely accompanied by simulation and application data.

Time Range

Publication date suggests work circa 2026.

Freshness

Last updated 2026-04-21 19:14:35; freshness should be verified.

Geography

null