FWDselect: an R package for variable selection in regression models

Abstract

In multiple regression models, when there is a large number p of explanatory variables which may or may not be relevant for predicting the response, it is useful to be able to reduce the model. To this end, it is necessary to determine the best subset of q (q <= p) predictors which will establish the model with the best prediction capacity. FWDselect package introduces a new forward stepwise based selection procedure to select the best model in different regression frameworks (parametric or nonparametric). The developed methodology, which can be equally applied to linear models, generalized linear models or generalized additive models, aim to introduce solutions to the following two topics: i) selection of the best combinations of q variables by using a step-by-step; and perhaps, most importantly, ii) search for the number of covariates to be included in the model based on bootstrap resampling techniques. The software is illustrated using real and simulated data.