Search for outliers using robust measures of location and scatter


Search for outliers using robust measures of location and scatter, which are used to compute robust analogs of Mahalanobis distance.

x is an n by p matrix or a vector of data.

The function returns the values flagged as an outlier plus the (row) number where the data point is stored. If x is a vector, indicates that the fourth observation is an outlier and outval=123 indicates that 123 is the value. If x is a matrix, indicates that the fourth row of the matrix is an outlier and outval reports the corresponding values.

The function also returns the distance of the points identified as outliers in the variable dis.

For bivariate data, if plotit=T, plot points and circle outliers.


out(x, = cov.mve, plotit = TRUE, SEED = TRUE, xlab = "X", ylab = "Y", qval = 0.975, crit = NULL, ...)



A vector, or a 2-dim matrix determines how the measure of scatter is estimated. Possible choices are:

  • cov.mve: the MVE estimate

  • the MCD estimate

  • covmba2: the MBA or median ball algorithm

  • rmba: an adjustment of MBA suggested by D. Olive

  • cov.roc: Rocke's TBS estimator


Should a plot be drawn? Only works for bivariate data


Set the random seed for reproducible results


Further arguments

Questions? Problems? Suggestions? or email at

All documentation is copyright its authors; we didn't write any of that.