Exam 8 Practice Question 5
Study guide: Exam 8
An insurer models collision claim frequency, and vehicle make and model has thousands of levels. Following Chalk et al., the modeler uses a single-feature penalized regression with a Gaussian distribution and identity link. The predicted frequency for a level is shrunk from the level's own mean toward the overall mean, with the regularization parameter \lambda chosen by cross-validation.
- Overall mean frequency: 0.050 claims per vehicle-year.
- One make and model: 200 vehicle-years, mean frequency 0.080.
- Selected
\lambda = 600.
Calculate the predicted claim frequency for this make and model.
- A) Less than 0.055
- B) At least 0.055, but less than 0.059
- C) At least 0.059, but less than 0.065
- D) At least 0.065, but less than 0.075
- E) At least 0.075
Answer this question and see the worked solution