Exercise 8.7

2 approaches can be used:

• Backward Elimination

  • begining: model with all variables
  • variable elimination (one by one)
  • end: satisfaction that all remaining variables are important to the model
  • with p-value
    • identify predictor corresponding to the largest p-value
    • drop variable, refit the model

• Forward Selection

  • begining: no variable included to the model
  • variables add according to their importance
  • end: no other important variables are found
  • with p-value
    • model has no predictors
    • fit a model for each possible predictor
    • identification of smallest predictor's p-value

• Strategy within each elimination step
  • eliminate the variable that leads to the largest improvement in adjusted R-squared

• The variable age can be elminated first