Slides 2025-2026 notes #12
Description
- slide 19 etc: y-axis for
Diabetes test resultfeels a bit strange with it being numerical. Would just display 0 and 1. - slide 25: needs some notes for trainers on why we now display "probability of positive test result"
- slide 30: love the phrase 'wrapper' to explain 'inverse link function'
- slide 33: maybe have an icon in the top-right corner to highlight there is an audience participation question (I do this for the Data analysis course - it's a useful cue / reminder for the trainers
- coffee example is great!
- slide 44: nice that the thing we're predicting is made clear
- slide 50: excellent
- slide 51: might need more background explanation for trainers in the notes
- slide 63: multiple curve example is really nice.
- slides for section 4 need trainer notes
- slide 70: in trainer notes "likelihood != probability. needs more explanation for trainers, clear definitions for and differences between the two
- slides 72 - 76: MLE in action is great. they'd go well in the LMM slides, for sure. They need some more info for trainers in the notes: (e.g. slide 73) "maybe look a bit more plausible" - explain why.
- slide 76: log-likelihood is unitless, but might be good to point out we still get a number (the bar plot doesn't show this)
- slide 76: notes seem truncated.
- slides 78-83: fantastic idea to at MLE for logistic regression. I never thought of doing it, and it really drives home the process!
- The whole Poisson section is vastly improved, well done. I'm loving the new plots.
- slides 99 - 101: need trainer notes, why do we have multiple distributions here?
- slides 112 - 114: need trainer notes
- Might need to consider splitting the lecture into 3 parts, with Section 7 (Beyond count & binomial data) happening after they've done some of the count response practicals