Model selection

Lecture 8

Dr. Mine Çetinkaya-Rundel

Duke University
STA 101 - Fall 2023

Warm up

Check-in

• Daily check-ins for getting you thinking at the beginning of the class and reviewing recent/important concepts
• Go to Canvas and open today’s “quiz” titled 2023-09-27 Check-in
• Access code: parsimonious (released in class)
• “Graded” for completion

Announcements

• My remaining office hours this week:
  • Wednesday 5-6 on Zoom (link emailed)
• Accommodation requests for exam must go in this week to SDAO
• Exam classroom – check email over the weekend, some of you will take in Gross 104 to give some elbow room to everyone
• Project 1 teams + assignment to be posted later today
  • Read before lab
  • Sit with your teams in lab

\(R^2\) and adjusted \(R^2\)

Two \(R^2\)s

• \(R^2\) measures the proportion of variability in the outcome variable that is explained by the model.

  • Always goes up with the addition of a predictor.

• Adjusted \(R^2\) is this value adjusted to penalize for the number of predictors in the model.

  • Goes up if the added predictor is “useful”, i.e. actually increases the predictive performance of the model.

  • Therefore we use this metric for choosing between models.

Application exercise

Go to Posit Cloud and start the project titled ae-08-Tips.

ICYMI

Today’s daily check-in access code: parsimonious (released in class)