Day 23 - 10/14/2024

Quick recap about model selection

• We want to select simple but complex enough models.
  
• Assumptions should be as realistic as possible.
  
• Usually, predictive ability is desired.
  • At what cost? - evaluate model fit adding a penalty for each parameter
    
  • Isn’t it unfair to use the data twice? - out-of-sample prediction accuracy
    • e.g., cross-validation (k-fold/leave-one-out)

Applied example

R script.

Reviewing the Homework

Find the guide here.

For next class

• Read chapter 9 - Shrinkage Methods
  
• Office hours today are right after class (10:30-11:30 am) & 3:00-3:20 pm