bioLeak: Leakage-Aware Biomedical Modeling
Why bioLeak | Guided workflow | Example data | Create leakage-aware splits with make_split_plan() | Validating splits with check_split_overlap() | Scalability | Strict leakage mode | Guarded preprocessing and imputation | Fit and resample with fit_resample() | Tidymodels interoperability | Nested tuning with tune_resample() | Advanced: Using Gradient Boosting with Parsnip | Advanced: Custom learners | Visual diagnostics | Audit leakage with audit_leakage() | Time-series leakage checks | Cross-validation uncertainty with cv_ci() | Delta Leakage Sensitivity Index: Quantifying Performance Inflation | Motivation | Mathematical framework | Notation | Stage 1: Within-repeat aggregation | Stage 2: Per-repeat inflation score | Point estimates | Setting up a two-pipeline comparison | Running delta_lsi() | Designing for a target inference tier | Statistical inference | Sign-flip randomization test | Huber M-estimator and robustness properties | BCa bootstrap confidence interval | Pairing condition and the inference tier system | Paired vs. unpaired designs | Metric direction with higher_is_better | Accessing the LeakDeltaLSI object | Connection to audit_leakage() | Parallel Processing | Simulation suite | Objects and summaries