Package: hetNSC 0.1.0

hetNSC: Heteroscedastic Nearest Shrunken Centroid Classification

Nearest shrunken centroid (NSC) classifiers for high-dimensional, low-sample-size data such as gene expression. Implements the original homoscedastic NSC with soft thresholding (Prediction Analysis of Microarrays; Tibshirani et al. 2002 <doi:10.1073/pnas.082099299>), its hard- and order-thresholding versions (Sahtout, Wang and Ghimire 2024 <doi:10.1080/03610918.2022.2047201>), and the heteroscedastic extension of Sahtout and Wang that uses class-specific variances in both the test statistic and the discriminant function (methods STh3, HTh3 and OTh3). The thresholding parameter is selected by a refined ("deep") cross-validation search. Also provides the simulation design, per-class performance metrics and the relative-heterogeneity diagnostic used in the paper.

Authors:Haiyan Wang [aut, cre], Mohammad Omar Sahtout [aut], Trevor Hastie [ctb], Robert Tibshirani [ctb], Balasubramanian Narasimhan [ctb], Gilbert Chu [ctb]