Changes to my metrics syllabus for this year:
1. Adding another day of diff-in-diff. At least in finance, it's pretty obvious that this is the dominant empirical paradigm and there's value in spending more time
2. Expanding discussion on model-based vs. design-based identification
3. Compressing discussion of unsupervised ML -- my colleagues are better at it anyway
4. Expanding on discussion of supervised ML in heterogeneous FX contexts (e.g. Chernozhukov and coauthors' work on GATES)
/n
5. Adding discussion on estimation for demand systems (e.g. BLP)
a. estimation issues (e.g. how semiparametric assumptions can get you in trouble, a la Conlon and Gortmaker, and Jiang Manchanda and Rossi)
b. id insights from Borusyak and Hull
6. Discussing contamination bias and issues in linear regression
7. Adding a discussion on granular IV from Gabaix and Koijen
8. Adding a discusison on Hausman instruments
@paulgp minor QoL thing: can you replace the raw github links in the main readme.md with nbviewer links? e.g.
https://github.com/paulgp/applied-methods-phd/blob/main/lectures/01_po_dags.pdf gets previewed in github's horrible window thing, while
https://nbviewer.org/github/paulgp/applied-methods-phd/blob/main/lectures/01_po_dags.pdf shows up like any other pdf in its own window.
[Can send you a pull request with the edits fwiw; thanks again for making this excellent material public]
@apoorvalal Yes! Haha, @gmcd has asked me to do the same... you're in good (and particular) company :-)
@paulgp Wait? It’s not 100% D-in-D now?
I kid, this is so much better than what was going on in the 90s when I was in grad school.
“I was running SPSS on a Mac Quadra when you were still in diapers!"
@TradingPlacesResearch honestly it feels like 100% dind!