R Data Mining

From the full model to the n-1 model

We start with the full model, that is, the model with all variables included:

With the first step we try all possible combinations of 5-1 = 4 variables:

Based on the R-squared of each model from 1A to 1E, we choose the best model, which here we are supposing to be the 1E. In the next step we, therefore, remove the violet variable:

This second step is exactly equal to the first: we try all possible combinations ...