mlr3proba 0.8.1

• ✨ New: surv.logloss and surv.rcll now use linear interpolation of the survival function for density estimation.
• 🛠️ Fixes:
  • surv.mae, surv.mse, and surv.rmse return NA if the test set is fully censored.
  • Division-by-zero in surv.brier now correctly uses eps instead of returning Inf.
• 🔧 Refactoring:
  • Removed se and method arguments from most scores (time-weighted integration is now the default).
  • All internal/private functions now start with a dot (.) and are documented accordingly.
  • Full cleanup of internal Rcpp scoring functions.
• 💥 Removed all experimental proper scoring rule options and the remove_obs argument. Default behavior remains unchanged (proper = FALSE).
• 📚 Improved documentation across many measures.

mlr3proba 0.8.0

• Compatibility with mlr3 v1.0.0 (weights_learner) and mlr3pipelines v0.8.0
• MAJOR feat: support right-censored competing risk tasks (TaskCompRisks, LearnerCompRisks, MeasureCompRisks)
  • Implemented AUC(t) via RiskRegression package
  • Baseline Aalen-Johansen estimator via survival package
• fix: as.data.table() for PredictionSurv objects holds now one survival curve per observation as it should
• refactor: TaskSurv uses only right, left or interval censoring, simplified code a lot in the methods

mlr3proba 0.7.5

• feat: add PEM (Piece-wise Exponential Model) reduction method, via a survival => poisson regression pipeline, from Bender et al. (2018)
• feat: add na.rm parameter to msr("surv.calib_index") to avoid NaN scores
• fix: allow cloning of measures objects
• fix: survival measure labels are now printed and the obs_loss property is supported

mlr3proba 0.7.3

• feat: added new calibration measure => msr("surv.calib_index")
• refactor: autoplot.PredictionSurv
  • The default "calib" plot uses the survival matrix directly now which is faster
  • "dcalib" has extra barplot + better documentation
  • Added new type = "scalib" which constructs the smoothed calibration plots as in Austin et al. (2020)
  • BREAKING CHANGE: type = "preds" is now called "isd" (individual survival distribution). row_ids can now be used to filter the observations for which you draw the survival curves.

mlr3proba 0.7.2

• fix: lrn("surv.coxph") is now trained with model=TRUE which fixes an issue with using observation weights (see stackoverflow question).
• cleanup: remove tsk("unemployment") and associated files
• cleanup: remove unused references

mlr3proba 0.7.1

• cleanup: removed all PipeOps and pipelines related to survival => regression reduction techniques (see #414)
• fix: $predict_type of survtoclassif_disctime and survtoclassif_IPCW was prob (classification type) and not crank (survival type)
• fix: G(t) is not filtered when t_max|p_max is specified in scoring rules (didn't influence evaluation at all)
• docs: Clarified the use and impact of using t_max in scoring rules, added examples in scoring rules and AUC scores
• feat: Added new argument remove_obs in scoring rules to remove observations with observed time t > t_max as a processing step to alleviate IPCW issues. This was before 'hard-coded' which made the Integrated Brier Score (msr("surv.graf")) differ minimally from other implementations and the original definition.

mlr3proba 0.7.0

• Add mlr3pipelines to Imports, refactoring/simplify code, set minimum latest version from CRAN (0.7.0) in DESCRIPTION
• Add new reduction method, a survival => classification pipeline (via IPCW, Vock et al. 2016)
• Improved the way integrated survival scores (eg surv.graf) handles the times argument and the t_max (results are the same as before if the times argument is not used)
• Improved documentation of integrated survival scores and some pipelines (add references)
• Add experimental lifecycle badge for some pipelines and pipeops - these are currently either not supported by literature or tested enough

mlr3proba 0.6.8

• Rcpp code optimizations
• Fixed ERV scoring to comply with mlr3 dev version (no bugs before)
• Skipping survtoregr pipelines due to bugs (to be refactored in the future)
• Deprecate crank to distr composition in distrcompose pipeop (only from lp => distr works now)
• Add get_mortality() function (from survivalmodels::surv_to_risk()
• Add Rcpp function assert_surv_matrix()
• Update and simplify crankcompose pipeop and respective pipeline (no response is created anymore)
• Add responsecompositor pipeline with rmst and median

mlr3proba 0.6.6

• Some fixes from v0.6.5

mlr3proba 0.6.5

• Compatibility with paradox@v1.0.0
• t_max updates and fixes on surv.cindex and surv.ibrier metrics
• New methods to TaskSurv
• coxed task generator
• Lots of refactoring
• Support for discrete-time survival analysis