[ellispatrick]

Part of a stack of 8 pull requests from the October 2026 audit of ClassifyR (AUDIT.md on branch audit-2026-10). Each part is based on the one before, so they need to be merged in order. Each one can be accepted, changed or rejected on its own merits. Draft PR #125 holds all eight together. Every part installs and passes its own tests.

Stack, merge in order: #126 tests and core fixes → #127 interface fixes → #128 metrics and plots → #129 pathways and criss-cross → #130 maintenance → #131 faster cross-validation (identical results) → #132 shared folds (changes results) → #133 release notes. The regression harness and timings are in SydneyBioX/ProjectClassifyR (private).

Summary

Release notes (inst/NEWS.Rd, version 3.17.2), help pages regenerated from the roxygen changes in parts 1–7, and the vignette's balanced accuracy computed inline instead of a fixed "error rate is about 20%".

Checks of the full stack

• R CMD check: one NOTE, already in devel (ggplot2 column names, and the unreachable ensemble-selection code, which is left for a decision).
• The vignettes build.
• 474 tests pass.
• The regression harness gives identical results for 1 and 20 cores.

Questions for Ellis and Dario

32 items in ProjectClassifyR docs/QUESTIONS.md. The main ones:

• Random forest's second forest: grown only to rank features, it doubles the fit time.
• SVM probabilities: probability = TRUE costs about 5× the fit.
• CoxNet: its lambda path is about 80% of the protocol's merge step.
• 2 dp rounding of AUC and the sample-wise C-index.
• Balancing in runTests.
• New classifiers: aorsf, glmboost, LiblineaR and ncvreg, which a benchmark showed to be fast and accurate.

🤖 Generated with Claude Code

https://claude.ai/code/session_01TfsjwkbDbHkX4kJZNe4ER4