NEWS

mimar 0.8.0 (2026-06-09)

Added superlearner and sl imputers. These construct a Super Learner-style ensemble by cross-validating candidate imputers on observed cells, assigning non-negative loss-based weights, and combining predictions inside the existing chained-imputation loop.
Added library, folds, and metalearner hyperparameters for superlearner.
Updated CRAN preparation files and vignette examples for the new release.

First public release candidate.

Added ncore to impute() for completed-dataset-level parallel imputation through functionals::fmap().
Added lightweight iteration traces to mimar_imputation diagnostics for convergence screening.
Added diagnostic plot types for boxplots, bivariate observed/imputed comparisons, categorical proportions, and trace summaries.
Updated density diagnostics to draw line-only overlays across imputations so multiple completed datasets remain visible.
Refreshed the diagnostic plotting palette to give mimar a distinct visual identity while retaining the existing plot themes.
Expanded the vignette with KNN-based diagnostic examples, parallel imputation notes, and interpretation guidance.

Initial compact missing-data grammar.
Added description, amputation, imputation, evaluation, pooling, and plotting.
Added chained native and optional learner-backed imputation adapters without a funcml dependency.