EUVIMEDOpdag EUVIMED
Crossref

Hypothesis testing for the dimension of random geometric graphs

Mingao Yuan, Feng Yu

Electronic Journal of Statistics · 2026-01-01 · 20 (2) · 5431-5469

DOI: 10.1214/26-ejs2609 ↗

At a glance

Journal article in Electronic Journal of Statistics, published on 2026-01-01. Written by Mingao Yuan and 1 co-authors. 41 references.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

In a random geometric graph (RGG), nodes reside in an m-dimensional metric space, with edges connecting pairs that are closer than a specified threshold. Fitting an RGG requires determining the latent dimension m, yet validating whether a prespecified value matches the true dimension remains an open challenge. This paper introduces the first hypothesis testing framework to address this problem. Under the null hypothesis, our proposed test statistic converges in distribution to the standard normal distribution; under the alternative, it diverges to infinity in probability. We provide an efficient method to compute the test statistic directly from the adjacency matrix. Furthermore, we address the multiple testing problem, for which we propose two tests, and propose two estimators for the true dimension. Simulations confirm the excellent empirical performance of the proposed methods, and applications to various real-world networks demonstrate their practical utility.

Cite

APA

Yuan, M., & Yu, F. (2026). Hypothesis testing for the dimension of random geometric graphs. Electronic Journal of Statistics, 20(2), 5431-5469. https://doi.org/10.1214/26-ejs2609

Vancouver

Yuan M, Yu F. Hypothesis testing for the dimension of random geometric graphs. Electronic Journal of Statistics. 2026;20(2):5431-5469. doi:10.1214/26-ejs2609

BibTeX
@article{Yuan2026,
  author = {Yuan, Mingao and Yu, Feng},
  title = {Hypothesis testing for the dimension of random geometric graphs},
  journal = {Electronic Journal of Statistics},
  year = {2026},
  volume = {20},
  number = {2},
  pages = {5431--5469},
  doi = {10.1214/26-ejs2609}
}
Start for free →