Computes robust minimum covariance determinant (MCD) distances across the observations (rows). The MCD method selects a subset of h observations whose covariance matrix has minimum determinant across all subsets of size h. The MCD distances are Mahalanobis distances using the estimates of center (mean) and scale (covariance matrix) based on that subset.
out_measures.robdist(U, quantile_cutoff = NA)| U | An n x Q matrix of PC scores. |
|---|---|
| quantile_cutoff | The F-distribution quantile cutoff. Default:
|
A list with entries
A vector of length n of with the robust distance estimate for each observation.
A list with entries "inMCD", "outMCD_scale", and "Fparam"
The robust distance cutoff value
Logical vector indicating the outliers
If is.null(quantile_cutoff) the latter two elements are omitted.