Changelog • IQCC

IQCC 0.8.0

Add the double-sampling np chart family for high-quality processes: dsnp_prob_accept(), dsnp_arl(), dsnp_ass(), dsnp_limits(), dsnp_design(), and cchart.DSnp().
Add generalized variance monitoring through gv_stat(), gv_limits(), gv_alpha_risk(), and cchart.GV(), including normal, Cornish-Fisher, selected exact, and simulation-based limits.
Add auxiliary multivariate variability monitoring based on tr(V) through trv_stat(), trv_limits(), trv_alpha_risk(), and cchart.trV().
Add pure numerical limit and risk functions for p and u charts: pchart_limits(), pchart_alpha_risk(), uchart_limits(), and uchart_alpha_risk().
Add pure numerical functions for exact and conventional range-chart limits: r_exact_limits() and r_shewhart_limits().
Add pure numerical functions for exact and conventional S-chart limits: s_exact_limits() and s_shewhart_limits().

Replace the unweighted mean of subgroup proportions with the pooled binomial estimator in cchart.p().
Replace the unweighted mean of subgroup rates with the pooled Poisson estimator in cchart.u().
Correct standardized p and u charts so standardized statistics are not divided by subgroup size a second time.
Correct and harden Cornish-Fisher p- and u-chart implementations and their exact binomial/Poisson false-alarm calculations.
Correct the DS-np limit search so every candidate second-stage threshold is evaluated independently; harden decision regions, ARL, ASS, curtailed inspection, and plotting.
Correct the Hotelling T-squared robustness simulation so the documented multivariate-t scenario uses the intended elliptical construction, validates admissible dimensions/correlation, preserves RNG state, and reports Monte Carlo uncertainty explicitly.
Preserve the historical qcc-compatible center convention for normalized S charts while separating its numerical limits from plotting.
Preserve documented legacy aliases and positional calls for existing chart wrappers.

Reproduce published p-chart values from Joekes and Barbosa (2013) with explicit provenance, tolerances, and independent binomial-risk oracles.
Reproduce the published R-chart quantiles, false-alarm risks, ARLs, and d2()/d3() constants from Barbosa, Gneri, and Meneguetti (2013), including independent Tippett-distribution integration checks.
Reproduce published DS-np ARL and ASS values from Joekes, Smrekar, and Barbosa (2015), with independent probability and small-sample enumeration checks.
Validate the u-chart Cornish-Fisher expansion independently from Poisson cumulants and exact Poisson false-alarm probabilities.
Reproduce published generalized-variance tables where the source parametrization is identifiable; record unresolved source conventions explicitly rather than treating them as implementation errors.
Add a consolidated executable numerical-validation vignette that classifies evidence as published reproduction, independent derivation, exact discrete calculation, numerical evaluation of an exact distribution, Monte Carlo, property testing, or partial reproduction.
Expand regression and property tests for scaling, monotonicity, boundary cases, invalid inputs, random-number-state preservation, and wrapper/core equivalence.

Add executable vignettes for getting started, package positioning, high-quality processes, univariate dispersion, multivariate monitoring, statistical foundations, numerical validation, and comparison with the R statistical-process-control ecosystem.
Add paper/statistical-foundations.md as article-oriented technical source material for future manuscript development.
Expand the README and pkgdown reference index to reflect the public numerical API and validated methods.
Add or revise documentation for public functions, Phase I/Phase II conventions, exact-versus-approximate terminology, and S3 methods.

Separate numerical kernels from chart construction and plotting across the audited p, u, R, S, DS-np, generalized-variance, and tr(V) families.
Add GitHub Actions for R CMD check on Linux (R-devel, release, oldrel), Windows, and macOS, plus pkgdown, URL, spelling/documentation-quality, and test-coverage checks.
Add Codecov reporting and substantially expand the automated test suite.
Improve input validation and replace silent error-like return values with explicit errors where audited.
Remove attach()-based examples from audited interfaces and keep vignettes executable in the documentation build.

Fix d3() operator precedence bug: d2[1] was always used instead of d2[i], producing wrong results when called with a vector of sample sizes (#2)
Fix stats() operator precedence bug: 1:m-1 evaluated as (1:m) - 1, producing index 0 on first iteration (affects all n=1 multivariate calculations)
Fix cchart.T2.2() operator precedence bugs: is.null(x) == FALSE patterns were fragile (5 occurrences)
Fix cchart.T2.1() incorrect UCL formula for Phase I with n=1: now uses correct beta distribution parameters per Montgomery 7th ed. (#1)
Fix cchart.S() type parameter mismatch: code checked for "c" but documentation specified "e" for exact limits
Fix cchart.Xbar_R() typo: add.stat corrected to add.stats
Fix standardized p-chart and u-chart: removed erroneous multiplication by sample size
Fix cchart.T2.1() error handling: sprintf() replaced with stop()
Fix error messages throughout: return("msg") replaced with stop("msg")
Correct spelling in standardized, Shewhart, and approximation documentation.
Add input validation (n >= 2) to d2(), d3(), c4()

Create constants.R with ALPHA, SIGMA_MULT, Q_LOWER, Q_UPPER constants
Replace 15+ hardcoded magic numbers across 7 files
Standardize T/F to TRUE/FALSE throughout
Standardize = to <- for assignment
Replace is.null(x) == FALSE with !is.null(x)
Fix table.qtukey(): remove unnecessary function definitions, add invisible() return
Pre-allocate vectors in d2() and d3() (was O(n²) with append())
Replace matrix(nrow=m, ncol=1) with numeric(m) for control limits