Section: New Software and Platforms

kosel

Variable Selection by Revisited Knockoffs Procedures

Keywords: Variable selection - Regression

Functional Description: Performs variable selection for many types of L1-regularised regressions using the revisited knockoffs procedure. This procedure uses a matrix of knockoffs of the covariates independent from the response variable Y. The idea is to determine if a covariate belongs to the model depending on whether it enters the model before or after its knockoff. The procedure suits for a wide range of regressions with various types of response variables. Regression models available are exported from the R packages 'glmnet' and 'ordinalNet'. Based on the paper linked to via the URL below: Gegout A., Gueudin A., Karmann C. (2019) https://arxiv.org/abs/1907.03153

News Of The Year: This package is a new one.

• Participants: Clémence Karmann, Aurélie Muller and Anne Gégout-Petit

• Contact: Aurélie Muller

• Publication: The revisited knockoffs method for variable selection in $L_1$-penalised regressions.

• URL: https://cran.r-project.org/web/packages/kosel/kosel.pdf