Improved test portability across BLAS/LAPACK implementations,
including MKL, by comparing complete-data Q2 cross-validation summaries
with a numeric tolerance while keeping exact selected-component parity
checks.
missPLS 0.2.0
Complete-data selection now degrades cleanly when optional
plsdof is not installed, falling back to the naive
information-criterion columns when degree-of-freedom based criteria are
unavailable.
Removed fixed default RNG seeds from the public API and avoided
implicit writes to the user’s home filespace; long-run reproduction
scripts now default to tempdir() unless an explicit output
directory is supplied.
missPLS 0.1.0
Initial package release.
Added methods-first helpers for simulation, MCAR/MAR missingness
generation, imputation, component selection, and standardized study
summaries.
Added packaged real datasets for the published bromhexine,
tetracycline, octane, and ozone analyses.
Added study runners for simulation and real-data workflows, plus
diagnostics for correlation and VIF-style screening.
Added long-run reproduction scripts under tools/ for
complete-data, incomplete-data, and real-data article/thesis
reruns.
Added tests, vignette material, and package documentation for the
CRAN-ready public API.
Complete-data Q2 cross-validation now follows the documented
plsRglm::cv.plsR(object = ...) interface, while
incomplete-data Q2 selection keeps the package-side NIPALS fallback used
by missPLS.