Thoughts on future iterations

[dajmcdon]

Potential large projects

1. Setting up JupyterHub and autograding functionality for Labs and/or Homeworks. Perhaps not that difficult using the existing DSCI infrastructure. Solutions can be validated with {digest}
2. Accomodate Python and/or Julia? Students always ask, but it would involve lots of duplication in the current state.
3. Make it partially flipped.
4. Switch HW/Labs to Quarto. Ongoing difficulty:
   • pdf output is easier for the TAs to grade, but gets long, paged, LaTeX headaches.
   • HTML output is hard to grade, but avoids LaTeX install headaches.

Potential structure changes

• Perhaps do Module 5 (Unsupervised) earlier. Or create something called "Feature creation" that bundles dimension reduction, basis expansions, some amount of the deep learning material.
• Students may need less {tidyverse} and GitHub than they did previously.
• Revisit using UBC GitHub.
• Explore changing the prerequisites (possibly just Stat 301).
• Worth incorporating other GLMs (besides logistic regression).
• Module 5 is a bit juvenile. It used to include other manifold learning stuff, but it took too much time.

Potential grading changes

• The 35% final is still a bit generous, so final grades are high (~ 82 average) even though the median on the final is about 50%.
• Could add a Midterm (and see below). Maybe 30% on the final 15% on a midterm. Effort for the rest.
• Less weight on labs (perhaps 1.5 points each)
• Some students still shirk on the clickers. The <25% penalty is very hard to enforce.
• Would be nice to have non-multiple choice exam. But horrible to mark.
• An IRT of the final would likely not be flattering (some questions are not discriminatory).

Comments from peer and student feedback

• Some students would actually like to see midterms. Though this means dealing with absences/excuses. This can also increase student anger at the instructor and lead to lower evals.
• Find a way to do more "live coding" (peer). Also, more "written" mathematics.
• Find a way to discuss HW and Lab solutions somewhere. (Perhaps every other lab "done" during the section, alternate is discussion of solutions).
• Monday labs haven't always seen the material yet. But deadlines are tied to the section meetings. I doubt this is that big a deal. Suggestion was "all due on Friday".