Speculation on the generality of the backward stepwise view of PCA

A novel backwards viewpoint of Principal Component Analysis is proposed. In a wide variety of cases, that fall into the area of Object Oriented Data Analysis, this viewpoint is seen to provide much more natural and accessable analogs of PCA than the standard forward viewpoint. Examples considered here include principal curves, landmark based shape analysis, medial shape representation and trees as data. Categories and Subject Descriptors G.3 [Probability and Statistics]: Multivariate statistics General Terms Theory

• Type: Journal
• Year: 2010
• Where: MIR
• Authors: J. S. Marron, Sungkyu Jung, Ian L. Dryden
• Keywords: Medial Shape Representation, MIR 2010, Multimedia, Principal Component Analysis, Standard Forward Viewpoint