Variable Selection and Model Building
10.1 a. With α = 0.10, the model chosen is y = β0 + β2x2 + β7x7 + β8x8 + ε.
b. With α = 0.10, the model chosen is y = β0 + β2x2 + β7x7 + β8x8 + ε.
c. With αIN = 0.05 and αOUT = 0.10, the model chosen is y = β0 + β2x2 + β7x7 + β8x8 + ε.
d. The three procedures chose the same model.
10.3 The choice of cut-off values is to prevent the circular addition-subtraction of the variables.
The model involves just x1
= 75.3%. Cp
= −1.8 and
b. Stepwise leads to the same model involving just x1.
10.7 When FIN = FOUT = 4.0, the model involves only x6. However, when FIN = FOUT = 2.0, the model involves x6 and x7.
The model involves x1
, and x4
= 95.3%, Cp
= 5 and
10.11 The model is y = β0 + β1x1 + β2x2 + β4x4 which has a PRESS = 85.35 and R2Pred = 96.86%.
From Section 3.7, we get
which means that .
b. Since we are fitting a model with only one regressor, the ...