After Milestone 3

Mick McQuaid

2023-10-30

Milestone 3 aftermath

• Milestone 3 didn’t go exactly as I had hoped
• I expected the four models to gradually get better
• Step 4 was supposed to introduce the model you will use for Milestone 4
• Instead, some groups did the machine learning step twice and ended up without a final good model
• Those groups now have no model to use for diagnostics in Milestone 4

Milestone 3 Redo

• I accept some responsibility, even though I thought the instructions were clear
• The fact of multiple groups misreading them is counter evidence
• Therefore, in this one and only one case, I’m going to allow a redo for those groups getting below a 10

Terms of Redo (1 of 2)

• Due this Friday at 11:59PM
• Improve your trial and error models to \(R^2\sim0.4\) so that I believe you’ve done trial and error and are not just using the first model to pop into your head
• Refer to the visualizations in Milestone 2 to justify your trial and error models

Terms of Redo (2 of 2)

• Use olsrr instead of leaps on step 3 (nobody used just quantitative models)
• For step 4 create a final model manually based on what you learned from the subsets
• Comment on your final model
• Omit diagnostic plots (they are for m4)

END

Colophon

This slideshow was produced using quarto

Fonts are League Gothic and Lato