Overview
Applied Econometrics is concerened with the interpretation of statistical results in various contexts. In this course, we anchor ourselves around the ideas of potential outcomes framework and local variation of the treatment variable. These two concepts motivate the use of instrumental variables, regression discontinuity, and difference-in-differences and therefore will be essential to our understanding of the empirical literature.
One point to highlight is that in contrast to other notes on applied econometrics (Goldsmith-Pinkham, Hull, Roth), these slides/notes are quite brief. They’re meant to help one develop a conceptual framework that one can use to read the typical applied micro paper. We’re going to let detials slip through the cracks. The hope, as Brad Osgood says is that “If you keep the big picture in mind, in many cases the details will take care of themselves –really.”
Randomized Control Trials
In this class we discuss the ethics and responsabilities around Randomized Control Trials. In particular we discuss some of the key issues that arise when randomizing financial resources among those in great need.
Language Models
Bengio et al. (2003) highlight the challenge of learning probability distributions over high dimensional discrete spaces and the need to generalize. It emphasizes how the task of gernalization is equivalent to transfering probability mass. Randford et al. (2018) describe how the curse of dimensionality at that the task level (dataset, objective), make an explicit multi-task approach infeasible. Bertsch et al. (2024) compare the relative effectiveness of in ncontext Learning versus fine-tuning for multic-class classifaction as the size of the examples increases. Slides
