Robust SIMCA-bounding influence of outliers

In this article a robust version of SIMCA, based on spherical principal component analysis [N. Locantore, J.S. Marron, D.G. Simpson, N. Tripoli, J.T. Zhang, K.L. Cohen, Robust principal component analysis for functional data (with comments), Test 8 (1999) 1–74], is introduced for chemometrics community. The efficiency of the new approach is compared to the classical SIMCA and to its robust version proposed by Vanden Branden et al. [K. Vanden Branden, M. Hubert, Robust classification in high dimensions based on the SIMCA method, Chemometrics and Intelligent Laboratory Systems 79 (2005) 10–21]. The performances of the presented approaches are evaluated on simulated and real data sets. The results obtained from a simulation study give evidence that the proposed robust SIMCA approach offers a satisfactory efficiency when the model set does not contain outliers and is also robust, what ensures a proper classification of new objects even, when the model set used to derive classification rules is contaminated to a large extent by outlying objects.