AoV-PLS: a new method for the analysis of multivariate data depending on several factors

A new method for the analysis of a multivariate dataset depending on several factors is proposed. It is called AoV-PLS (Analysis of Variance-PLS). It is based on the decomposition of the dataset into the main effects, the interactions effects and possibly the residual matrix using a model akin to analysis of variance (ANOVA). Each effect is considered in turn and assessed through the use of a Partial Least Square regression (PLS-regression). The connection of AoV-PLS to competing methods such as ANOVA-PCA and ANOVA-Simultaneous Component Analysis (ASCA) is emphasized and these methods are compared on the basis of a dataset pertaining to metabolomics field.