Just a few weeks in the past, I informed you concerning the new materials we’re including to this yr’s CodeChella — steady DiD, artificial DiD, triple variations, bounding workout routines, the entire frontier. Right now I need to make the ask extra instantly: in the event you’ve been on the fence, that is the publish the place I attempt to get you off it.
Come to Madrid.
CodeChella runs Might 25–28 at CUNEF Universidad. 4 days, 9am to 5pm, morning espresso and pastries included. Even in the event you barely know what a regression is, however you’re keen to study and get your arms soiled with code, then you definitely’re prepared for this workshop. We construct from the bottom up.
Tickets are on Eventbrite right here. Pricing:
∙ College students: $220
∙ Submit-docs: $300
∙ School: $500
If value is the impediment, e-mail me at causalinf@mixtape.consulting and we’ll work one thing out. I imply it. I don’t need that to be the rationale you don’t come.
The Claude Code Thread
I’ve been writing on this Substack for months about how Claude Code has modified the way in which I do empirical analysis. CodeChella is the place you get to see it in motion.
All through the workshop, I’ll be working my replications and demonstrations inside Claude Code environments. Which means each time we work via a brand new estimator — occasion research, Callaway-Sant’Anna, Arkhangelsky’s artificial DiD, Rambachan-Roth bounds — you’ll even be watching me work with Claude Code in actual time to construct it. The diff-in-diff content material and the AI-assisted workflow are woven collectively, not siloed.
My principle right here is fairly easy: one of the simplest ways to study Claude Code is to make use of it for one thing you had been already planning on doing anyway. Making occasion examine graphs. Working pre-trend checks. Constructing clear tables and publication-quality figures. If these are stuff you care about — and in the event you’re coming to CodeChella they most likely are — then you definitely’ll depart with each the econometrics and a working sense of tips on how to use an AI coding agent to do utilized quantitative analysis.
However I additionally need to be trustworthy about one thing. Velocity shouldn’t be the purpose. The factor I need to train — the factor I believe issues most proper now — is verification. How have you learnt what Claude Code produced is correct? How do you construct habits that catch errors earlier than they find yourself in a paper? How do you construction a workflow in order that the beneficial properties in velocity don’t come at the price of credibility?
That’s a part of what this workshop is now. Not a demo of how briskly I can run issues. A severe try to indicate you tips on how to use these instruments properly.
Madrid in Late Might
The climate is ideal. The meals is extraordinary. CUNEF is a good venue. And actually, 4 days in Madrid with a room full of people that care about causal inference is one among my favourite issues I get to do.
I’ll be again subsequent Monday with extra. However in the event you already know you need to come — seize your ticket right here.
