TwoCutoff News

TwoCutoff 0.1.0 (2026-06-06)

New Features

• Added low_pct and high_pct arguments to derive_cutoffs_percentile() for flexible percentile specification (instead of hardcoded 80th/20th).
• Added loess_span argument to derive_cutoffs_sensspec() to allow user control of smoothing in LOESS regression.
• Added risk_col argument to dca_analysis() to support custom risk score column names.
• Dynamic lower bound for decision curve plots (ylim_lower) to avoid clipping when net benefit is negative.
• Biomarker name is now stored in result objects (cutoff.percentile, cutoff.sensspec) and referenced directly in print methods.

Improvements

• evaluate_performance() documentation updated to clarify that indeterminate subjects are excluded from sensitivity/specificity calculations.
• Print methods (print.cutoff.percentile, print.cutoff.sensspec) now display biomarker name explicitly instead of inferring from dataset.
• Cleaner output from adjust_score() when validate = FALSE (no test_auc included).
• Expanded test coverage in tests/testthat/:
  • Validation and error handling in adjust_score.
  • Edge cases and fallback cutoffs in derive_cutoffs_*.
  • Performance metrics validation for both cutoff methods.
  • New test files for compare_performance(), dca_analysis(), and plot_two_cutoff().

Bug Fixes

• Fixed fragile biomarker inference in print methods.
• Improved error messages for missing biomarker or non-binary outcome in adjust_score().

This NEWS file summarizes the major changes, new arguments, documentation updates, and test coverage improvements for this release.