Showing 1–1 of 1 results for author: Ahdesmäki, M

Feature selection in omics prediction problems using cat scores and false nondiscovery rate control

Authors: Miika Ahdesmäki, Korbinian Strimmer

Abstract: We revisit the problem of feature selection in linear discriminant analysis (LDA), that is, when features are correlated. First, we introduce a pooled centroids formulation of the multiclass LDA predictor function, in which the relative weights of Mahalanobis-transformed predictors are given by correlation-adjusted $t$-scores (cat scores). Second, for feature selection we propose thresholding cat… ▽ More We revisit the problem of feature selection in linear discriminant analysis (LDA), that is, when features are correlated. First, we introduce a pooled centroids formulation of the multiclass LDA predictor function, in which the relative weights of Mahalanobis-transformed predictors are given by correlation-adjusted $t$-scores (cat scores). Second, for feature selection we propose thresholding cat scores by controlling false nondiscovery rates (FNDR). Third, training of the classifier is based on James--Stein shrinkage estimates of correlations and variances, where regularization parameters are chosen analytically without resampling. Overall, this results in an effective and computationally inexpensive framework for high-dimensional prediction with natural feature selection. The proposed shrinkage discriminant procedures are implemented in the R package ``sda'' available from the R repository CRAN. △ Less

Submitted 8 October, 2010; v1 submitted 11 March, 2009; originally announced March 2009.

Comments: Published in at http://dx.doi.org/10.1214/09-AOAS277 the Annals of Applied Statistics (http://www.imstat.org/aoas/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Report number: IMS-AOAS-AOAS277

Journal ref: Annals of Applied Statistics 2010, Vol. 4, No. 1, 503-519