NEWS.md

In CalibrationCurves: Calibration Performance

CalibrationCurves 3.1.0

Major changes

• Archived timeROC dependency: The timeROC package (v0.4) is being archived on CRAN. Its required functions have been integrated directly into CalibrationCurves (R/timeROC_archived.R) with proper attribution. timeROC has been removed from Imports; pec has been added instead (required for ipcw()).

Bug fixes

• valProbCluster(): cl.level parameter was ignored (Issue #22): The confidence level was hardcoded to 0.95 throughout the clustering pipeline. cl.level is now correctly passed to CGC(), MAC2(), and MIXC() and propagated to metaprop(), metagen(), and rma.mv() calls. Hardcoded "95%" plot labels have been replaced with dynamic labels via the new ci_pi_labels() helper function.

• valProbSurvival(): crash near max follow-up (Issue #24): Uno's time-dependent AUC was evaluated at max(fit$y) - 0.01 instead of the user-specified timeHorizon, causing "incorrect number of dimensions" errors when the risk set was depleted. Fixed to use times = timeHorizon.

Enhancements

• New "default" approach in valProbCluster(): Combines MAC2 (splines) for the overall calibration curve, confidence intervals, and prediction intervals, with MIXC for cluster-specific curves. This is now the default when approach is not specified. The returned object contains both the MAC2 overall results (results$overall) and the MIXC cluster results (results$clusters).

• Unified the legend title to "Heterogeneity" across all valProbCluster() approaches.

• Updated plot font to sans-serif and increased base size to 11 for improved readability across MIXC, MAC2, and CGC approaches.

Minor changes

• Added input validation for cl.level in valProbCluster().
• New helper function ci_pi_labels() for formatting confidence/prediction interval labels dynamically.
• Added documentation of the "default" approach to the package vignette.

CalibrationCurves 3.0.0

• Added valProbCluster() for calibration of clustered data with three approaches: MIXC, MAC2, and CGC.
• Added valProbSurvival() for calibration of survival/time-to-event data.
• Added genCalCurve() for generalized calibration across the exponential family.

CalibrationCurves documentation built on March 27, 2026, 9:06 a.m.