Robust discrimination under a hierarchy on the scatter matrices

Under normality, Flury and Schmid [Quadratic discriminant functions with constraints on the covariances matrices: some asymptotic results, J. Multivariate Anal. 40 (1992) 244-261] investigated the asymptotic properties of the quadratic discrimination procedure under hierarchical models for the scatter matrices, that is: (i) arbitrary scatter matrices, (ii) common principal components, (iii) proportional scatter matrices and (iv) identical matrices. In this paper, we study the properties of robust quadratic discrimination rules based on robust estimates of the involved parameters. Our analysis is based on the partial influence functions of the functionals related to these parameters and allows to derive the asymptotic variances of the estimated coefficients under models (i)-(iv). From them, we conclude that the asymptotic variances verify the same order relations as those obtained by Flury and Schmid [Quadratic discriminant functions with constraints on the covariances matrices: some asymptotic results, J. Multivariate Anal. 40 (1992) 244-261] for the classical estimators. We also perform a Monte Carlo study for different sample sizes and different hierarchies which shows the advantage of using robust procedures over classical ones, when anomalous data are present. It also confirms that better rates of misclassification can be achieved if a more parsimonious model among all the correct ones is used instead of the standard quadratic discrimination.

[1]  G. Boente,et al.  General projection-pursuit estimators for the common principal components model: influence functions and Monte Carlo study , 2006 .

[2]  C. Croux,et al.  Principal Component Analysis Based on Robust Estimators of the Covariance or Correlation Matrix: Influence Functions and Efficiencies , 2000 .

[3]  Frank Critchley,et al.  The influence of observations on misclassification probability estimates in linear discriminant analysis , 1991 .

[4]  J. A. Branco,et al.  Partial influence functions , 2002 .

[5]  P. Lachenbruch.  Note on Initial Misclassification Effects on the Quadratic Discriminant Function , 1979 .

[6]  Frank Critchley,et al.  Influence functions of two families of robust estimators under proportional scatter matrices , 2007, Stat. Methods Appl..

[7]  Terence J. O'Neill.  Error rates of non-Bayes classification rules and the robustness of Fisher's linear discriminant function , 1992 .

[8]  Peter Filzmoser,et al.  CLASSIFICATION EFFICIENCIES FOR ROBUST LINEAR DISCRIMINANT ANALYSIS , 2008 .

[9]  Wing K. Fung,et al.  Diagnostics in Linear Discriminant Analysis , 1995 .

[10]  Liliana Orellana,et al.  Robust plug-in estimators in proportional scatter models , 2004 .

[11]  Ana M. Pires,et al.  Influence functions and outlier detection under the common principal components model: A robust approach , 2002 .

[12]  Wing K. Fung.  Diagnosing influential observations in quadratic discriminant analysis , 1996 .

[13]  H. P. Lopuhaä.  Breakdown Point and Asymptotic Properties of Multivariate S-Estimators and τ-Estimators : A Summary , 1991 .

[14]  C. Croux,et al.  Robust linear discriminant analysis using S‐estimators , 2001 .

[15]  Sanford Weisberg,et al.  Directions in Robust Statistics and Diagnostics , 1991 .

[16]  Bernhard W. Flury,et al.  Quadratic discriminant functions with constraints on the covariance matrices: some asymptotic results , 1992 .

[17]  David E. Tyler.  Radial estimates and the test for sphericity , 1982 .

[18]  Graciela Boente,et al.  A Robust Approach to Common Principal Components , 2001 .

[19]  H. P. Lopuhaä.  On the relation between S-estimators and M-estimators of multivariate location and covariance , 1989 .

[20]  N. Campbell,et al.  The Influence Function as an Aid in Outlier Detection in Discriminant Analysis , 1978 .

[21]  Christophe Croux,et al.  Influence of observations on the misclassification probability in quadratic discriminant analysis , 2005 .

[22]  Tony Springall.  Common Principal Components and Related Multivariate Models , 1991 .

[23]  V. Yohai,et al.  Robust Statistics: Theory and Methods , 2006 .

[24]  G. Boente,et al.  Robust tests for the common principal components model , 2009 .

[25]  Daniel Gervini.  The influence function of the Stahel-Donoho estimator of multivariate location and scatter , 2002 .

[26]  B. Efron.  The Efficiency of Logistic Regression Compared to Normal Discriminant Analysis , 1975 .

[27]  Bernhard W. Flury,et al.  Error rates in quadratic discrimination with constraints on the covariance matrices , 1994 .

[28]  Werner A. Stahel,et al.  Statistics in Genetics and in the Environmental Sciences , 2001, Entropy.