PERMANOVA 0.2.1
New features
- Added unified distance wrapper
Distances() with
automatic variable type detection.
- Added
DistCategorical() (Cramer’s V and matching
coefficient) and DistMixed() (Gower, generalized Gower, and
robust RELMS distances) powered by dbrobust.
- Expanded
DistContinuous() with Mahalanobis and robust
covariance-based distances (Robust_GV,
Robust_MCD), as well as optional Euclidean correction
(euclidean = TRUE) across distance functions.
- Added parallel permutation support (
parallel and
cores arguments) in PERMANOVA(), compliant
with CRAN core limits.
Bug fixes and improvements
- Graphical methods (
plot.BootDisMANOVA,
plot.PERMANOVA, plot.BootCanonAnalysis,
plot.MANOVA.Biplot, plot.MANOVA):
Fixed plot boundaries so projected individuals and confidence regions
are no longer clipped; fixed cluster/group factor references
(x$Clusters) and axis indexing when
A1/A2 differ from 1 and
2; removed non-functional redundant arguments.
- Principal coordinates & bootstrap: Corrected
sign and
eigen() decomposition in Gower double-centering
within PERMANOVA(), BootDisMANOVA(), and
BootDistCanonicalAnalysis(), fixed individual projection
scaling when DatosIni = FALSE, and made stratified
bootstrap resampling independent of row ordering.
- MANOVA & contrasts: Fixed Wilks’ lambda
calculation and dimension dropping in two-group designs
(
r = 1), corrected degrees of freedom for Roy’s greatest
root in MANOVA.Estimation(), enabled
MaxOrderIter in ConstructContrasts(), and
fixed Logit and normalized residuals in
IniTransform().
- Summary & LaTeX output: Fixed field references
(
Effects) and row-wise cumulative contributions in
summary.MANOVA(), summary.BootDisMANOVA(), and
summary.PERMANOVA().