Variable selection using stepdown procedures in high-dimensional linear models
Volume 43 / 2016
Applicationes Mathematicae 43 (2016), 157-172
MSC: Primary 62J15, 62F03; Secondary 62J05.
DOI: 10.4064/am2286-6-2016
Published online: 25 August 2016
Abstract
We study the variable selection problem in high-dimensional linear models with Gaussian and non-Gaussian errors. Based on Ridge estimation, as in Bühlmann (2013) we are considering the problem of variable selection as the problem of multiple hypotheses testing. Under some technical assumptions we prove that stepdown procedures are consistent for variable selection in a high-dimensional linear model.
