Partial Influence Functions
In this paper we extend the definition of the influence function to functionals of more than one distribution, that is, for estimators depending on more than one sample, such as the pooled variance, the pooled covariance matrix, and the linear discriminant analysis coefficients. In this case the appropriate designation should be "partial influence functions," following the analogy with derivatives and partial derivatives. Some useful results are derived, such as an asymptotic variance formula. These results are then applied to several estimators of the Mahalanobis distance between two populations and the linear discriminant function coefficients.
Year of publication: |
2002
|
---|---|
Authors: | Pires, Ana M. ; Branco, João A. |
Published in: |
Journal of Multivariate Analysis. - Elsevier, ISSN 0047-259X. - Vol. 83.2002, 2, p. 451-468
|
Publisher: |
Elsevier |
Keywords: | asymptotic variance influence function linear discriminant function Mahalanobis distance robust estimators |
Saved in:
Saved in favorites
Similar items by person
-
Projection-pursuit approach to robust linear discriminant analysis
Pires, Ana M., (2010)
-
Boente, Graciela, (2006)
-
Robust discrimination under a hierarchy on the scatter matrices
Bianco, Ana, (2008)
- More ...