Selecting the best-fitted regression model with stepwise regression

In order to find the best-fitted regression model, one can perform stepwise regression to step-wisely add or remove a term from a fitted model, and finally output a model with the least AIC. In the following recipe, we demonstrate how to perform stepwise regression with the step function.

Getting ready

You need to have completed the previous recipe by fitting house rental data into a multiple regression model, fit.

How to do it…

Perform the following steps to search for the best-fitted regression model with the step function:

1. First, you can use step to select the optimum model with backward elimination:

   > step(fit, direction="backward")
   Start:  AIC=12753.77
   Price ~ Sqft + Floor + TotalFloor + Bedroom + Living.Room + Bathroom
   
                 Df  Sum of Sq        RSS   AIC
   - TotalFloor   1 1.8081e+08 2.4428e+11 12752
   <none>                      2.4410e+11 12754
   - Bathroom     1 8.7580e+08 2.4497e+11 12754
   - Living.Room...