NEWS

---

gipsDA 1.0.0 (2026-10-01)

Bug fixes

• Fixed prediction from formula fits when fitting-only arguments such as show_progress_bar and store_probabilities are supplied.
• Fixed prediction from matrix fits using na.action.
• Improved handling of edge cases involving one-predictor models and unused factor levels.
• Added validation for method = "debiased" in small-sample settings.

Documentation

• Clarified the interpretation of selected permutation structures in gipslda() versus QDA models.
• Updated the manual reconstruction vignette to match the current covariance scaling and class-specific QDA sample-size handling.
• Documented fitting metadata stored in fit_info and summary objects.

Maintenance

• Updated the minimum supported R version.
• Improved build ignores for local coverage files, plots, and generated artifacts.