I teach a number of methods courses mostly about the cutting edge of multilevel and panel data modeling. These are classes that are designed to be more advanced and/or specialized than most universities can offer because there isn’t really enough local demand (students who can take them) or supply (faculty who can teach them). Because they are supposed to be at the cutting edge I obviously don’t have a fully fixed curriculum of things that I always cover. Each time I teach a class it’s a bit different.
In addition to the classes that I offer on my own, I also regularly teach for the ICPSR Summer Program in Ann Arbor and the GSERM program in Switzerland. I owe a huge debt to the ICPSR Program both as a former participant and as an instructor as they’ve been instrumental in making me into a methodologist. Additionally, I’ve taught methods workshops at the University of Kentucky, UC Berkeley, the University of Ljubljana, and the University of Milan.
Advanced MLM is my first methods course and will always be my baby. I cover material from econometrics, biostatistics, statistics, political methodology, ecology, and psych methods in this class. We go through implementation of models in both Stata and R using both classical and (basic) Bayesian techniques. The class is overwhelmingly focused on the idea of model (mis) specification, omitted variable bias, and endogeneity for spatiotemporal and network models.
I teach a few different variants on this course. My four-week ICPSR course, a week-long course for GSERM, and some two-week versions that I’ll be working out on my own. To see the general content of this kind of class you can click above.
Modern DiD is my newest course and easily my most ambitious one to date. I started teaching it in summer of 2020 and am now on my fifth iteration. It changes quite a lot each time. This course is designed to cover the new flood of research on counterfactual estimation techniques in policy evaluation and econometrics. I cover topics like staggered timing, interference & spillover effects, generalized latent variable modeling for exchangeability adjustments, generalized synthetic control methods, and whatever else I run across. This course uses both Stata and R since a lot of the newer methods are built for one or the other.
Advanced Bayesian Multilevel Modeling
This is a newer course that I’m putting together based on extensions of the Bayesian content that I’ve covered in my MLM classes so far. While that class is incredibly ambitious in what it covers, it is designed as a pretty soft introduction to Bayesian methods given that most of my students haven’t studied the topic. Every time I teach it I always want to cover more. This class does that and is designed to go deep into the weeds on topics relating to Bayesian MLM.
I’ll most likely be teaching it in the fall of 2021.