Page 174 - Applied Statistics with R
P. 174

174                   CHAPTER 9. MULTIPLE LINEAR REGRESSION


                                 # SSE
                                 sum(resid(full_mpg_model) ^ 2)


                                 ## [1] 4556.646

                                 # SST
                                 sum(resid(null_mpg_model) ^ 2)



                                 ## [1] 23761.67

                                 # Degrees of Freedom: Regression
                                 length(coef(full_mpg_model)) - length(coef(null_mpg_model))



                                 ## [1] 2

                                 # Degrees of Freedom: Error
                                 length(resid(full_mpg_model)) - length(coef(full_mpg_model))


                                 ## [1] 387


                                 # Degrees of Freedom: Total
                                 length(resid(null_mpg_model)) - length(coef(null_mpg_model))


                                 ## [1] 389



                                 9.4 Nested Models


                                 The significance of regression test is actually a special case of testing what we
                                 will call nested models. More generally we can compare two models, where
                                 one model is “nested” inside the other, meaning one model contains a subset of
                                 the predictors from only the larger model.
                                 Consider the following full model,


                                                                                 
                                                  =    +       +       + ⋯ +    (  −1)   (  −1)  +      
                                                    0
                                                         1   1
                                                                2   2
                                                  
                                 This model has    − 1 predictors, for a total of      -parameters. We will denote
                                 the fitted values of this model as ̂ .
                                                                
                                                               1  
                                 Let the null model be
   169   170   171   172   173   174   175   176   177   178   179