COMPSTAT 2018: Start Registration
View Submission - COMPSTAT2018
A0324
Title: Robust sparse maximum association estimators Authors:  Andreas Alfons - Erasmus University Rotterdam (Netherlands) [presenting]
Christophe Croux - Edhec Business School (France)
Peter Filzmoser - Vienna University of Technology (Austria)
Abstract: The maximum association between two multivariate random variables is defined as the maximal value that a bivariate association measure between respective one-dimensional projections of each random variable can attain. Using the Spearman or Kendall rank correlation as projection index thereby yields a more robust procedure than using the Pearson correlation. We propose a projection pursuit algorithm based on alternating series of grid searches in two-dimensional subspaces of each data set, together with an extension that allows for sparse estimation of the projection directions to increase the interpretability of the results in higher dimensions. In addition, we provide a fast implementation of the algorithm for the statistical computing environment R.