Penalized Youden Index Estimator

Implements the Penalized Youden Index Estimator (PYE) and the Covariate-Adjusted Youden Index Estimator (covYI), providing a novel framework for feature and covariate selection and combination in high-dimensional binary classification problems. Methodologies are based on Salaroli and Pardo (2023) and an unpublished manuscript by Salaroli and Pardo (2026) under review.

pye: Penalized Youden Index Estimator R Package

The pye package implements the penalized variable selection and classification methodologies developed during my PhD research at the Complutense University of Madrid (UCM).

The package features the Penalized Youden Index Estimator (PYE) and the Covariate-adjusted Youden Index (covYI) estimator, designed for simultaneous feature selection, coefficient estimation, and threshold optimization, both in low- and high-dimensional settings.

📌 Repository Structure & Versions

• main (Version 0.1.0 - 2026): Clean production-ready codebase optimized according to CRAN guidelines. Prepared for submission.

⚙️ Core Features Included

• Primary Estimators: Smooth kernel-based estimation for standard/weighted PYE and covYI.
• Sparsity Penalties: Full implementation of $L_{1/2}$, $L_1$ (Lasso), Elastic-Net, SCAD, and MCP regularization.
• Optimization Engines: Modified monotone (mmAPG) and non-monotone (mnmAPG) Accelerated Proximal Gradient solvers.
• Cross-Validation & Simulations: Integrated automated grid search for hyperparameter tuning ($\lambda$ and $\tau$) and parallelized train-test validation splits.
• Benchmark Modules: Standardized interfaces for penalized logistic regression, penalized Support Vector Machines (SVM), and penalized AUC-based regression.

🚀 Installation (for Advisors & Collaborators)

To install the package directly from this private repository, make sure you have devtools installed and run the following command in R/RStudio:

if (!require("devtools")) install.packages("devtools")
devtools::install_github("UCMpyePackage/pye")