La scelta delle variabili in un modello di regressione lineare

This article analyzes the problem of variable selection in a linear regression model. After an introduction on the different aims of a selection method, some of the most used criteria are compared. In particular, we consider the two possible approaches for the variable selection: explanation or prediction. Finally, it is shown by a simulation that using a criterion that is based on the estimation of the prediction error plus a penalty term, you can get a very effective method in identifying the ‘true’ model.