Activity Number:
|
48
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 4:00 PM to 5:50 PM
|
Sponsor:
|
Business & Economics Statistics Section*
|
Abstract - #300892 |
Title:
|
Estimation of Error Rate for the Linear Discriminant Function under the High-Dimensionality with Small-Sample Setting
|
Author(s):
|
Peter Chen*+ and Dean Young
|
Affiliation(s):
|
University of Mary Hardin-Baylor and Baylor University
|
Address:
|
P.O. Box 8420, UMHB Station, Belton, Texas, 76513, USA
|
Keywords:
|
linear discriminant function ; asymptotic approximation ; Monte Carlo simulation ; conditional error rate ; Mahalanobis distance
|
Abstract:
|
In this paper, we propose a new estimator of the conditional error rate (CER) for the linear discriminant function, which performs favorably to other currently-available methods under the high-dimensionality and small-sample condition. Also, we compare various CER estimators implemented by two different estimators for the Mahalanobis distance. We find that Dorans shrunken-generalized-distance expression improves the efficacy of some CER estimators when pattern classes are close together.
|