PredictSosDisc-class: Class '"PredictSosDisc"' - prediction of '"SosDisc"' objects
In rrcovHD: Robust Multivariate Methods for High Dimensional Data

Description Objects from the Class Slots Details Methods Author(s) References See Also Examples

The prediction of a "SosDisc" object

Objects can be created by calls of the form new("PredictSosDisc", ...) but most often by invoking predict() on a "SosDisc" object. They contain values meant for printing by show()

classification:: Object of class "factor" representing the predicted classification
mahadist2:: A "matrix" containing the squared robust Mahalanobis distances to each group center in the subspace (see Details).
w:: A "vector" containing the weights derived from robust Mahalanobis distances to the closest group center (see Details).

For the prediction of the class membership a two step approach is taken. First, the newdata are scaled and centered (by obj@scale and obj@center) and multiplied by obj@beta for dimension reduction. Then the classification of the transformed data is obtained by prediction with the Linda object obj@fit. The Mahalanobis distances to the closest group center in this subspace is used to derive case weights w. Observations where the squared robust mahalanobis distance is larger than the 0.975 quantile of the chi-square distribution with Q degrees of freedom receive weight zero, all others weight one.

show: signature(object = "PredictSosDisc"): Prints the results.

Irene Ortner irene.ortner@applied-statistics.at and Valentin Todorov valentin.todorov@chello.at

Clemmensen L, Hastie T, Witten D & Ersboll B (2012), Sparse discriminant analysis. Technometrics, 53(4), 406–413.

Ortner I, Filzmoser P & Croux C (2020), Robust and sparse multigroup classification by the optimal scoring approach. Data Mining and Knowledge Discovery 34, 723–741. doi: 10.1007/s10618-019-00666-8.

SosDisc-class