LIBRA stands for "LIBrary for Robust Analysis". It is a Matlab toolbox mainly containing implementations of
robust statistical methods. Robust statistics is involved with the detection of aberrant observations, also called outliers. The aim of robust statistical methods is to provide estimates which are not affected by the non-regular observations and which are then able to pinpoint the outliers. The robust methods implemented in LIBRA nowadays cover: univariate location, scale (Qn) and skewness estimation (MC), covariance estimation (FAST-MCD), regression (FAST-LTS, MCD-regression, depth quantiles), principal component analysis (RAPCA, ROBPCA), principal component regression (RPCR), partial least squares regression (RSIMPLS), classification (RDA, RSIMCA) and several methods to deal with skewed data. Besides that, the toolbox contains various methods for cluster analysis, many graphical diagnostic tools, and classical equivalents of several implemented methods.
Only a few of the recently added methods will be highlighted in this paper.
The features of the LIBRA functions will be illustrated by means of some real data sets.