Title: Variable screening for high dimensional regression problems via random projections
Authors: Laura Anderlucci - University of Bologna (Italy) [presenting]
Matteo Farne - University of Bologna (Italy)
Giuliano Galimberti - University of Bologna (Italy)
Angela Montanari - Alma mater studiorum-Universita di Bologna (Italy)
Abstract: Variable selection in high-dimensional settings characterizes many scientific problems. The concept of sure screening has been previously introduced to reduce the dimensionality, and a procedure has been proposed essentially based on the magnitude of marginal correlations between the predictors and the response variable. We propose a variable selection method for multiple linear regression which is based on axis-aligned random projections and accounts for partial correlation between each predictor and the response. Performances of the proposed method are evaluated on simulated data.