Wrapped up Stanford CS336 (Language Models from Scratch), taught with an amazing team @tatsu_hashimoto @marcelroed @neilbband @rckpudi. Researchers are becoming detached from the technical details of how LMs work. In CS336, we try to fix that by having students build everything:

This is a really good set of notes, by @urfriendlen https://t.co/6RirceLYUi

@ElizabethSetren @TuftsUniversity @nberpubs @BlueprintMIT Next, in 5th place, was @Jeonghyeok__Kim's "The Long Shadow of School Closures: Impacts on Students’ Educational and Labor Market Outcomes" https://t.co/bxUl9jfVA1 @UH_Econ @AnnenbergInst

Folks seemed to like these before, so I'm posting the full set of slides from my advanced undergrad 'metrics course at @BrownUniversity (drawn from an earlier class developed at @UChicago, with parts co-developed by @jondr44) https://t.co/4MpVjmFuhM https://t.co/dl4W5p36mP

Colin Cameron, econometrician and author with Trivedi of a fantastic econometrics textbook, has a deck of slides on causal inference for those interested. https://t.co/UpkPuOvhro

A new randomized controlled trial comparing breastmilk to formula among extremely preterm infants was published just yesterday. Short🧵 They found no effect whatsoever on neurodevelopmental outcomes at 22-26 months of age.

Hi #EconTwitter! Looking for a long list of “public goods”, with zillions of useful 𝐫𝐞𝐬𝐨𝐮𝐫𝐜𝐞𝐬 𝐟𝐨𝐫 #Economists? Don’t miss @Christine_Cai27’s amazing list - with everything, from applied #econometrics to #python coding, latex, job market, etc. Bookmark it now👇

Hi #EconTwitter! Looking for a nice course on 𝐂𝐚𝐮𝐬𝐚𝐥 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 at the master/PhD level? 📚🔬 Check out this insightful one by @MC_Knaus (@uni_tue) where the focus is on program evaluation. R code/notebooks included! Don't miss out! 👀

Constructing schools in #Senegal 🇸🇳 reduces fertility before age 25 by 39% --> mechanisms are delays in [early] marriage & increases in infant health by @Hamidou_Diallo_ [@WorldBank ] & Delaunay [@ird_fr] #EconTwitter https://t.co/aqsTieSeUc

Hi #EconTwitter! Looking for graduate #Econometrics lecture notes, covering a lot - GMM, bootstrap, program evaluation, treatment effects and much more? Check out these @MITOCW notes by Victor Chernozhukov @VC31415 (@MIT) & Ivan Fernandez-Val (@BU_Tweets)! Very useful stuff!

ChatGPT is everywhere and everyone is using it. But most academics don't know how to use it *smartly.* Here's how to use ChatGPT intelligently (and ethically) for academic purposes:

My student is teaching Public Econ at grad level (expenditures not tax). He is looking for slides to use as a model. If anyone has some publicly available, please reply and he will be immensely grateful or DM me!! I haven't taught public at grad level so couldn't share mine.

"Who Benefits From Attending Effective High Schools?" now available ahead of print at the Journal of Labor Economics https://t.co/YnxbTu3YhZ

The new version of #sdid is out! Now featuring event study graphs. Great stuff! @DanielPailanir I wrote about sdid here: https://t.co/ZH0f5v8yQp Will be updating this page and other packages soon!

Synthetic difference-in-differences, in Stata! Check out our article describing the package. Makes nice graphs illustrating time trends and unit weights. Comments welcome! https://t.co/itOs5Jl9on

Over 7 years as (Managing) Editor of the Review of Finance, I had to reject 1,000 papers given our increase in standards. "Lessons From 1,000 Rejections" aims to use them constructively by distilling common reasons for rejection to guide future research.

If one more article is written talking about all the data on student outcomes and doesn't mention PSEO, my head is going to explode. On a related noted, new data coming on Thursday... https://t.co/dBVq93vqtN

@estebanjq3 For any power analysis beyond basic RCT setups, simulation is almost always the answer! I walk through an interaction term power calculation by simulation in my book

2 useful papers today on how you can get basically any answer you want from regressions using arcsinh(y) or log(y+1) I made these graphs to show the problem: The mean of the transformed var (black lines) varies like crazy by what units you choose for y 🧵 on my own take-aways

I have a report out today via @BrookingsInst that analyzes new (Oct 2017) data from US Dept of Ed on student loan repayment. Ten-tweet summary in this thread. [START]