Dear R users, if you are using machine learning algorithms and have a lot of data (especially a lot of predictors, e.g. in genome-wide association studies), check the package bigstep. You can easily run the stepwise procedure with e.g. BIC or mBIC (which is modified version of BIC designed for a large number of predictors). Selection can be performed on data which exceed RAM capacity. Vignette: https://cran.r-project.org/web/packages/bigstep/vignettes/bigstep.html Best, Piotr Szulc [[alternative HTML version deleted]]