RT Journal Article
T1 Eigenvectors of a kurtosis matrix as interesting directions to reveal cluster structure
A1 Peña, Daniel
A1 Prieto Fernández, Francisco Javier
A1 Viladomat Comerma, Julia
AB In this paper we study the properties of a kurtosis matrix and propose its eigenvectorsas interesting directions to reveal the possible cluster structure of a data set. Under amixture of elliptical distributions with proportional scatter matrix, it is shown that asubset of the eigenvectors of the fourth-order moment matrix corresponds to Fisher's lineardiscriminant subspace. The eigenvectors of the estimated kurtosis matrix are consistentestimators of this subspace and its calculation is easy to implement and computationallyefficient, which is particularly favourable when the ratio n/p is large.
PB Elsevier
SN 0047-259X
YR 2010
FD 2010-10
LK https://hdl.handle.net/10016/14888
UL https://hdl.handle.net/10016/14888
LA eng
DS e-Archivo
RD 2 may. 2024