Help teaching short-course that has a healthy dose of data simulation

This post is by Lizzie. I hope you like the cats photo from this summer. I do.

I am looking for help. I decided to change my term course (12-14 weeks-long) on `introduction to Bayesian modeling with some hierarchical modeling’ (no, that’s the not the official title, but that is the gist) to a three-week intensive. I have been thinking about doing this for a couple years, but finally decided to do it now. My feelings in teaching the term-length versus short course is that: (1) a lot of students discover during term that Bayesian approaches are a lot more work than what they can get away with in frequentist methods and lose interest, but are still stuck in class for weeks — this way, when they can find this out, the course will soon be over! (2) As a corollary to (1), taking a Bayesian course (to me) mainly means finding out if you want to dive in and do it more (as you will never learn enough in a term course to be off and running fully), so with a short course, more students will take it and find out if they want to learn more. (3) More students will take a short-course and provide more support for those who want to continue on (building a useful community). (4) More students will take the course and learn how to simulate data, and I want more people to learn this. There’s lots of other reasons, but those are my big ones.

The downsides are that students will likely arrive unprepared and I won’t have time to go through the basics the way I do in 12+ weeks and that generally, there will be less content. Also, no more analyzing your own data as a term project that I support. People will miss that.

I have two days of 3.5 hours each per week for three weeks (about 21 hours) and plan to cover: Week 1: Data simulation for linear regression; what are priors and some ways to check them Week 2: Fitting a model in rstanarm to simulated data; diagnostics Week 3: Introduction to hierarchical models and posterior predictive checks

What I could use help on: – Suggested classroom activities and problem sets – Good example datasets or vignettes and generally any other good resource that could help me teach this material – Advice on how to structure or approach a short course to make it work well – Recommended background info to point students to

I have some ideas of examples I might use, like simulating data to show how much bigger a sample size you need to estimate an interaction in week 1 perhaps, and the Olympic figure skater example (judges, skaters from Regression and Other Stories) as homework for week 3, but I am really not sure so all advice and ideas would be most welcome.*. I am especially hoping to emphasize data simulation throughout, so ideas there are extra welcome. Please let me know in the comments any ideas/thoughts you have (or you can email me if you much prefer). Thanks in advance.

* One thing that I am NOT looking for is a big debate on the values of trying to teach all the way to hierarchical when people may not really have the basics.