Robust tools for three-way component analysis of compositional data

A brief overview of multilinear tools, robustification procedures and compositional data analysis for three-way arrays is followed by a detailed presentation on the use and applicability of the implemented functions by means of real data examples included in the R package rrcov3way. Two main functions are implemented, Parafac() and Tucker3(), which include alternative options to the standard least squares algorithm in order to properly calculate the models in case of data with outliers and/or characterized by a biased covariance structure. The package provides a set of tools for fitting the Tucker3 and PARAFAC models to multidimensional arrays by use of classical, robust and compositional estimating procedures. A comprehensive collection of three-way plots, diagnostics and data processing functions is also made available.