What Happens If in the Principal Component Analysis the Pearsonian is Replaced by the Brownian Coefficient of Correlation?
38 Pages Posted: 1 Jun 2014 Last revised: 22 Jun 2014
Date Written: May 29, 2014
The Brownian correlation has been recently introduced by Székely et al. (2007; 2009), which has an attractive property that when it is zero, it guarantees independence. This paper investigates into the effects and advantages, if any, of replacement of the Pearsonian coefficient of correlation (r) by the Brownian coefficient of correlation (say, ρ), other things remaining the same. Such a replacement and analysis of its effects have been made by the Host-Parasite Coevolutionary algorithm of global optimization applied on six datasets.
Keywords: Brownian correlation, Principal Component Analysis, Global Optimization, Host-Parasite Coevolutionary algorithm, Iris Flower Dataset, 1985 Auto Imports Database, Levy distribution, outliers
JEL Classification: C13, C43, C45, C61, C63, C87
Suggested Citation: Suggested Citation