HVS: Hypergraph Variable Selection
Performs hypergraph-based setwise variable selection with false discovery rate control (Organ, Kenney & Gu, 2026, <doi:10.48550/arXiv.2606.20514>). The idea is, in addition to selecting individual predictors when there is sufficient evidence, to also test all pairs of predictors, and when there is insufficient evidence to be sure which is the true predictor, it will select possibly overlapping pairs, for which there is strong evidence that at least one is a true predictor. The method is designed to control a generalised false discovery rate, where discoveries are counted based on the number of independent sets. The function of this package is similar to the 'hypergraph.sizing' package, but this package is optimised for faster computation in the case where we test all pairs of predictors. The package also includes functions for counting independent sets in a graph or hypergraph, either exactly or approximately. There is also a very limited function for isotonic regression, which is designed for fast computation in the specific case needed for hypergraph variable selection, rather than for general use.
| Version: |
1.0.0 |
| Depends: |
R (≥ 3.2.3) |
| LinkingTo: |
Rcpp |
| Published: |
2026-06-30 |
| DOI: |
10.32614/CRAN.package.HVS (may not be active yet) |
| Author: |
Toby Kenney [cre],
Sarah Organ [aut] |
| Maintainer: |
Toby Kenney <tkenney at mathstat.dal.ca> |
| License: |
GPL-3 |
| NeedsCompilation: |
yes |
| CRAN checks: |
HVS results |
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=HVS
to link to this page.