Exam MAS-I Practice Question 142

Study guide: Exam MAS-I

A pricing team has 250 drivers, each with 1,200 telematics features. Using all 250 drivers, it runs forward stepwise selection and keeps 30 features, then fits least squares on them. The team now wants to show management that the model will predict loss well for new drivers.

Which piece of evidence best supports that claim?

• A) The final 30-feature model's training R^2 of 0.96, the highest of all the models tried
• B) The final model's adjusted R^2 of 0.93, which already penalizes the 30 selected features
• C) Each of the 30 selected features has a p-value below 0.001 in the final least squares fit
• D) The model's error on 100 more drivers set aside before any selection or fitting began
• E) The final model's BIC, which is the lowest among the models the stepwise search compared

Answer this question and see the worked solution

Concepts tested

• High-Dimensional Data
• Test Error
• Variable Selection
• Overfitting

More questions from this set

• Exam MAS-I Practice Question 141
• Exam MAS-I Practice Question 151