A comparison between two robust PCA algorithms

• DOI: 10.1016/j.chemolab.2003.12.011
• Publication Date: 2004-02-27T10:21:49Z
• Authors: I Stanimirova; B Walczak; D.L Massart; V Simeonov
• Keywords: 0101 mathematics; 01 natural sciences

Abstract The article reports the results of a comparative study of two robust Principal Component Analysis (PCA) algorithms based on Projection Pursuit which can be used for exploratory data analysis. The first one is proposed by Croux and Ruiz-Gazen, denoted as C–R algorithm, and the second one by Hubert et al., introducing its modified version, abbreviated as RAPCA. They are applied to uniformly distributed simulated data sets, chemical data sets [environmental and near infrared (NIR) spectra] containing various numbers of variables and objects, as well as different observations' structure. Their performance and features, what they offer, are discussed in detail.