Fit robust Mahalanobis detector on compositional log-ratio coordinates

Fits a Minimum Covariance Determinant (MCD) robust Mahalanobis distance model on ILR (default) or rCLR log-ratio coordinates of a training compositional matrix. The MCD estimator is provided by the robustbase package. The function fails closed if robustbase is unavailable and require_robust = TRUE (the default).

Usage

fit_compositional_mahalanobis(
  x_train,
  transform = c("ilr", "rclr"),
  alpha = 0.75,
  pseudocount = 0.5,
  require_robust = TRUE
)

Arguments

x_train

Numeric matrix (samples \(\times\) features). All values must be non-negative.

transform

"ilr" (default, \(D \to D-1\) dimensional, full rank under healthy assumptions) or "rclr" (sum-to-zero constraint, effective rank \(D-1\)).

alpha

Numeric in (0.5, 1). MCD coverage fraction. Default 0.75.

pseudocount

Numeric \(> 0\). Added before log transform. Default 0.5.

require_robust

Logical. If TRUE (default), stops when robustbase is unavailable rather than silently downgrading to classical covariance.

Value

Object of class compositional_mahalanobis_fit.

References

Rousseeuw PJ, Van Driessen K. (1999) Technometrics 41: 212-223.

Examples

if (FALSE) { # \dontrun{
set.seed(42)
x_train <- matrix(abs(rnorm(80 * 20, 100, 30)), nrow = 80,
                  dimnames = list(NULL, paste0("miR-", seq_len(20))))
fit <- fit_compositional_mahalanobis(x_train, transform = "rclr",
                                     require_robust = FALSE)
} # }