Joe Ornstein

A person who tweets on the Internet | @PolsUga

[Joe Ornstein @JoeOrnstein]

🚨 New R package and paper! 🚨

fuzzylink is a method for merging datasets with non-exact matches on key variables. The paper walks through several useful political science applications--linking voter files, campaign contribution data, and multilingual records. Package on CRAN.

[Cambridge University Press - Politics @CUP_PoliSci]

#OpenAccess from @polanalysis -

Probabilistic Record Linkage Using Pretrained Text Embeddings - https://t.co/ff4Zup8GOs

- @JoeOrnstein

#FirstView

[PSRM Journal @PSRMJournal]

🦜Do you want to train your 'stochastic parrot'?

➡️@JoeOrnstein, E.N. Blasingame & @JakeTruscott_44 share best practices for using large language models (LLMs) in social science measurement tasks and processing large text-as-data projects https://t.co/a2wwOL7p82 #FirstView

[Carlisle Rainey @carlislerainey]

Thanks to @JoeOrnstein and others in @PolsUga, @UGA_INTL, and @UGA_SPIA for having me out to talk about statistical power.

A great, engaged group of faculty and PhD students.

https://t.co/Kqw3h9ExnE

[Joe Ornstein @JoeOrnstein]

@RohanAlexander @PoliSciJack Very cool. Glad to see the technique in use. Looking forward to seeing the election results.

[American Political Science Review @apsrjournal]

@enblasingame, @TheCLBoyd, @robertfcarlos, & @JoeOrnstein demonstrate that the October 2018 performance quota policy for immigration judges led them to issue more removal orders (both in absentia and merits orders). Read more in #APSRFirstView.
https://t.co/sDL9SAleDr

[Joe Ornstein @JoeOrnstein]

I've updated my Text As Data course materials for 2023. Includes some new tutorials on working with APIs and Large Language Models (LLMs).

https://t.co/m2JljkByIg

[Joe Ornstein @JoeOrnstein]

@JustinGrimmer @b_m_stewart @mollyeroberts I've updated my Text As Data course materials for 2023. Includes some new tutorials on working with APIs and Large Language Models (LLMs).

https://t.co/m2JljkC6xO

[Joe Ornstein @JoeOrnstein]

I wrote a tidyverse R companion to the (awesome) Text as Data book by @JustinGrimmer, @b_m_stewart, and @mollyeroberts. Includes code for webscraping, OCR, word embeddings, topic models, and more -- you can find it all here https://t.co/m2JljkByIg.

[Joe Ornstein @JoeOrnstein]

Today's #EPSA2023 fun fact: did you know that for nearly 400 years every student at University of Glasgow was examined orally, "seated upon the Black Stone"? There's a handy 20-minute hourglass built in to the top of the chair. Post-ChatGPT, is it time to return to the Old Ways?

[Joe Ornstein @JoeOrnstein]

@GeoffLorenz Reminded me of this article: https://t.co/6ygOU8ZRD5

Commercial leases tend to be much longer than residential leases (5-10 years), so landlords are more willing to wait for the right tenant rather than lower the price to fill the vacancy now.

[Joe Ornstein @JoeOrnstein]

@abhworthington That’s how I’ve heard @JustinWolfers pronounce it. Though now I’m realizing that may be a uninformative signal.

[Joe Ornstein @JoeOrnstein]

@ShaneDPhillips A few people mentioning cars, but I’d emphasize trucks. Before trucks, industry had to locate near rail lines, so to avoid it you just build your house far away from the tracks. After trucks, industry could be ANYWHERE, so keeping it out of your neighborhood required zoning.

[Joe Ornstein @JoeOrnstein]

@Jacob_Montg One option is to redefine the population of interest. Congress is 33% lawyers. If we randomly sampled 535 Americans, what’s the chance of getting that many lawyers?

95% CI is the range of population values that would make it unsurprising to find so many lawyers in Congress.

[Joe Ornstein @JoeOrnstein]

On deck at #APSA2022:

@enblasingame, @JakeTruscott_44, and I show you how to (politely) ask GPT-3 to classify your political text data.

Tomorrow morning (9/17) at the High Dimensional Machine Learning panel. Join us then, won't you? 🦜

[Joe Ornstein @JoeOrnstein]

We're hiring an Assistant Professor in American Politics this year!

I will be @APSAtweets Thursday and Friday. Please reach out if you want to meet and listen to me gush about how much I love @PolsUga.

[Christina L. Boyd @TheCLBoyd]

UGA Poli Sci is hiring TT in American politics (open specialization & methodology). Candidates researching race, ethnicity, & politics are especially encouraged to apply.

Ad will be live in a few days.

I'm chairing the search. Happy to chat @ APSA or beyond.

[Joe Ornstein @JoeOrnstein]

🦜 New Paper Alert 🦜 “How To Train Your Stochastic Parrot: Deep Language Models for Political Texts” with @enblasingame and @JakeTruscott_44. Thanks to @kinggary and everyone at @PolMeth2022 for a great discussion this morning! https://t.co/pSPDBxubKg

[Joe Ornstein @JoeOrnstein]

My wife has me in charge of managing the gift exchange for her friend circle, and each year it gets more challenging to construct a network that avoids repeats, dyads, and cyclic triples. What I'm saying is it's nice to be with someone who really gets me.

https://t.co/9VieHXdRzy

[Jacob Montgomery (@jacobmontgomery.bsky.social) @Jacob_Montg]

This seems like a good moment to note that @jb_duckmayr is on the market and you should be interviewing him.

[Joe Ornstein @JoeOrnstein]

Respect for the Ohio man who spent an hour taking the CES so he could write "Nunya business" in every text box.