|
Activity Number:
|
25
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 3, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
IMS
|
| Abstract - #300353 |
|
Title:
|
Diagnostics for Quadratic Discriminant Analysis
|
|
Author(s):
|
Christophe Croux*+
|
|
Companies:
|
U.V. Leuven
|
|
Address:
|
Dept. of Applied Economics, Leuven, , B-3000, Belgium
|
|
Keywords:
|
multivariate analysis ; diagnostics ; robustness
|
|
Abstract:
|
In regression analysis it is common to compute diagnostic measures like the Cook Distance or studentized residuals to identify influential observations. For most other multivariate analysis techniques, such diagnostic measures are less frequently used and often not even available. Our aim is to develop tools to identify influential points in the setting of discriminant analysis. One possible approach would be to look at the influence of each observation on the parameters of this quadratic function. This is not without interest, but since the main purpose of discriminant analysis is classification, it is more interesting to look at the influence of observations from the training sample on the probability of misclassification. We derive now expressions for such diagnostics in the quadratic case, and we will illustrate their use by some real data examples. Finally, let us not forget that diagnostic measures derived from nonrobust estimators may become unreliable in presence of multiple outliers. It will be important to use robust estimators for discriminant analysis.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2003 program |
