Robust Mahalanobis Distance — robust

Drop-in replacement for mahalanobis that handles singular (rank-deficient) covariance matrices by falling back to the Moore–Penrose pseudoinverse via MASS::ginv.

Usage

robust_mahalanobis(x, center = NULL, cov = NULL, inverted = FALSE)

Arguments

x: Numeric matrix (samples x features) or vector.
center: Numeric vector of column means (length = number of features). If NULL, computed from x.
cov: Covariance matrix. If NULL, computed from x.
inverted: Logical; if TRUE, cov is already inverted.

Value

Named numeric vector of Mahalanobis distances.