NEWS

---

qcaERT 0.1.2 (2026-07-30)

• incl.test() and ncut.test() now leave first_failing missing when the next proposed cutoff lies outside the admissible search range, because that value was not analyzed. Inclusion-cutoff trace plots likewise omit untested boundary proposals and their reference lines.
• ncut.test() now rejects a baseline n.cut greater than the number of rows in the analysis data.
• altset.test() now verifies that every tested condition or outcome membership in calib.data matches the reference calibration reconstructed from raw.data and calib_spec.
• Corrected alternative-set fit reporting so n_fit_deltas = 0 identifies a completed comparison with no differences beyond fit_tol, whereas NA identifies a draw in which fit was not compared. Printed summaries now show the separate denominators for draw availability, formula preservation, and fit preservation.
• Removed the unused necessity argument from cluster.test(). The function now documents its scope explicitly as the sufficiency assessment of selected solution formulas and prime implicants, and its displayed configuration values no longer include a redundant solution-type prefix.
• Added centralized row-alignment checks for paired raw, calibrated, reduced-run, and truth-table inputs. Identifiable case labels must now be unique and appear in the same order; otherwise pairing remains explicitly positional.
• Corrected cluster.test() pooled, cluster-specific, and within-unit calculations. Missing unit identifiers no longer remove rows from pooled or cluster fit, and duplicate unit-cluster observations are rejected.
• Isolated cosmetic progress-message selection from R's analytical random-number stream, so progress display cannot change seeded results.
• Standardized validation of integer-valued controls, including search budgets, sample sizes, replications, seeds, and display digits. Intentional sample_prop rounding is unchanged.
• Declared R (>= 3.5.0), matching the serialization format used by the shipped demonstration data.

qcaERT 0.1.1

• Added the World Happiness Report demonstration objects whr_raw, whr_calibrated, and whr_calib_spec.
• Fixed calibration trace plots so plot(..., type = "trace") no longer duplicates an evaluated failing threshold.

qcaERT 0.1.0 (2026-05-27)

This is the first public version of qcaERT.

Package scope

• Added a family of robustness tools for QCA workflows:
  • calib.test() for calibration-threshold robustness.
  • incl.test() for inclusion-cutoff robustness.
  • ncut.test() for frequency-cutoff robustness.
  • loo.test() for leave-one-out case influence.
  • subsample.test() for repeated subsample stability.
  • altset.test() for sampled alternative analysis settings.
  • theory.test() for comparing theoretically motivated condition-set specifications.
  • cluster.test() for cluster, group, and repeated-unit heterogeneity.
  • sol.df() for compact QCA solution tables.
  • sol.chart() for visual presentation of sol.df() tables.

Shared API and output conventions

• Standardized the function family around common solution controls: solution, include, dir.exp, which_M, and i_mode.
• Standardized exclusion handling across applicable functions with exclude_mode, exclude_recompute, and exclude_static.
• Standardized most result objects around:
  • diagnostics for detailed/internal results.
  • results for the clean table.
  • settings for the analysis settings.
  • supporting fields such as baseline, bounds, by_direction, by_case, by_run, by_draw, and summary where relevant.
• Standardized print() methods to show concise summaries.
• Standardized as.data.frame() methods to return the clean table.
• Documented the structured-result exceptions explicitly:
  • cluster.test() returns results$overview, results$clusters, and results$units; as.data.frame() returns results$overview.
  • theory.test() returns results$models, results$solutions, and results$pairwise; as.data.frame() returns results$models.