stabel: An R package for variable selection and prediction in high-dimensional cross-sectional data

Variable selection is crucial in high-dimensional biomedical research, yet many methods are unstable and select irrelevant variables. Stability selection improves robustness but may miss important variables and does not offer prediction. To address these limitations, we developed stabel, an R package, a framework that integrates multiple stability selection procedures with diverse models and subsampling strategies, followed by ensemble-based prediction. The package supports binary and continuous outcomes, offering LASSO, sparseSVM, and random forest for variable selection, and any prediction algorithm available in the SuperLearner package. stabel provides a flexible and reliable tool for high-dimensional variable selection and prediction.