NEWS

MVTests 2.3.6

Added the X/mu0/Sigma/n/p/nper interface for one-sample cellMCD calibration. The null mean is required. When Sigma is omitted, it is estimated from raw X using the bias-corrected cellMCD scatter; a supplied Sigma takes precedence.
Validate supplied dimensions against X and retain explicit n/p/Sigma mode. Return calibration dimensions, mean, covariance, moments and success/failure counts, with safeguards for finite-variance moment matching (q > 4).
Preserve the former nrep/seed/data interface and fixed-center plug-in behavior in the deprecated simRobCellT2_onesample_legacy() compatibility function. Use set.seed() for reproducibility with the new interface.
Updated examples, help and tests. The scaled-F reference remains an empirical approximation; the RobCellT2_onesample() statistic calculation is unchanged.

Added named data-based null calibration using the cellMCD plug-in scatter, fixed null mean, matching complete-case handling, and recorded fit settings; retained the explicit-covariance positional API.
Added Sigma-aware Monte Carlo calibration for the one-sample cellMCD test, with identity-model warnings, covariance validation, and explicit safeguards for finite-variance F moment matching.

Updated AWRcca() so that score-space observation weights enter a second shrinkage-regularized canonical correlation fit. The final estimator no longer falls back to unregularized classical CCA merely because p + q > n.
Replaced the Monte Carlo consistency correction with deterministic numerical integration.
Added canonical directions and estimator diagnostics to the returned object while retaining the existing cor, shrink_used, weights, u1, and v1 components for backward compatibility.
Expanded input validation and numerical safeguards.