Paper Title
Augmented Shrunken Centroid Regularized Discriminantanalysis

Abstract
A regularized extension of linear discriminant analysis (lda) calledregularized shrunken centroid discriminant analysis (scrda) was developedfor class prediction with microarray data; but it is likely of more general interestbecause of its application to other classification problems where thereare a large number of features with only a smaller number of observations. Inour article an extension is proposed that incorporates the nearest neighbour method. We show that when nonlinearity is present, the new method can substantiallyoutperform scrda. This extension also provides a method of testingfor the presence of nonlinearity. Numerical examples presented in this articlealso demonstrate how inadequate replication in cross-validation estimates ofthe test misclassification rate may result in poor quality estimates. Keywords: big-p-small-n problem;delete-d cross-validation;knn; linearand non-linear discriminant analysis; microarray data; nearest-neighbourautocovariate; synthetic microarray data;