Title: Robustness of resampling-based error rate estimators in two class discrimination under non-normal population

Authors: Kozo Yamada, Hirohito Sakurai, Hideyuki Imai, Yoshiharu Sato

Addresses: Graduate School of Information Science and Technology, Hokkaido University, Kita 14, Nishi 9, Kita-ku, Sapporo, Hokkaido 060-0814, Japan. ' Research Division, National Center for University Entrance Examinations, 2-19-23 Komaba, Meguro-ku, Tokyo 153-8501, Japan. ' Graduate School of Information Science and Technology, Hokkaido University, Kita 14, Nishi 9, Kita-ku, Sapporo, Hokkaido 060-0814, Japan. ' Graduate School of Information Science and Technology, Hokkaido University, Kita 14, Nishi 9, Kita-ku, Sapporo, Hokkaido 060-0814, Japan

Abstract: This paper numerically investigates robustness of resampling-based error rate estimators to kurtosis when Fisher|s linear discriminant function is used. In order to control the population kurtosis and to examine the robustness of estimators, we assume Pearson|s type VII and exponential power distributions as a population distribution. The robustness study is carried out for several resampling-based estimators based on graphical and quantitative approaches.

Keywords: error rate estimators; robustness; resampling-based estimators; non-normality; kurtosis; linear discriminant function.

DOI: 10.1504/IJRIS.2011.037737

International Journal of Reasoning-based Intelligent Systems, 2011 Vol.3 No.1, pp.14 - 27

Published online: 26 Dec 2010 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article