I'm excited to share my job market paper (for the 2025-26 market)!
It introduces a new extension of RDD where outcomes are entire distributions: Regression Discontinuity Design with Distributions (R3D).
Thread below ๐ (1/)
@gushamilton We do some things to restrict that like limiting web search and sandboxing them and changing years and proper nouns but probably those estimates are more of a lower bound.
@gushamilton Great point! In the simulations we do see pretty high corrections, ~19x more likely to pick an analysis when the p value is low. The applications' factors might partly be lower because they're well established papers and the models might have memorized some of the results.
New paper: Inference after coding agents (w/ Florian Gunsilius and Kaspar Wรผthrich)
We develop a method for correcting distortions to the p-values of empirical findings produced by coding agents. Tldr in video, short thread below: (1/)
https://t.co/d3TTF53qKM
Finally, one may ask: can agents self-correct?
A: no! Agents don't know what they *would* do on different data.
They do try, often correcting for multiple testing. But 1) they do so inconsistently 2) they can't correct what they don't test: the garden of forking paths! (17/)