RGS: Recursive Gradient Scanning Algorithm

Provides a recursive gradient scanning algorithm for discretizing continuous variables in Logistic and Cox regression models. This algorithm is especially effective in identifying optimal cut-points for variables with U-shaped relationships to 'lnOR' (the natural logarithm of the odds ratio) or 'lnHR' (the natural logarithm of the hazard ratio), thereby enhancing model fit, interpretability, and predictive power. By iteratively scanning and calculating gradient changes, the method accurately pinpoints critical cut-points within nonlinear relationships, transforming continuous variables into categorical ones. This approach improves risk classification and regression analysis performance, increasing interpretability and practical relevance in clinical and risk management settings.

Version: 1.0
Imports: rms, SemiPar, survival
Suggests: testthat (≥ 3.0.0)
Published: 2024-12-19
DOI: 10.32614/CRAN.package.RGS
Author: Shuo Yang [aut, cre], Yi Fei [aut], Jinxin Zhang [ths]
Maintainer: Shuo Yang <yangsh223 at mail2.sysu.edu.cn>
License: MIT + file LICENSE
NeedsCompilation: no
CRAN checks: RGS results

Documentation:

Reference manual: RGS.pdf

Downloads:

Package source: RGS_1.0.tar.gz
Windows binaries: r-devel: not available, r-release: RGS_1.0.zip, r-oldrel: RGS_1.0.zip
macOS binaries: r-release (arm64): RGS_1.0.tgz, r-oldrel (arm64): RGS_1.0.tgz, r-release (x86_64): RGS_1.0.tgz, r-oldrel (x86_64): RGS_1.0.tgz

Linking:

Please use the canonical form https://CRAN.R-project.org/package=RGS to link to this page.