iiinova — where your digital self keeps evolving

Reposted by Nathan Hoffmann

jon ben-menachem @jbenmenachem.com · 29/09/2026

"Replication crisis" reforms like preregistration & pre-analysis plans were developed to control false positives in experimental research. Does that work in observational studies? Our new paper in Sociological Methods & Research argues for a different approach. 🧵 journals.sagepub.com/doi/10.1177/...

A Replication Framework for Observational Research: From Specification Restriction to Specification Expansion
Jonathan Ben-Menachem https://orcid.org/0000-0002-8092-0848 jbenmenachem@gmail.com, Ari Galper https://orcid.org/0000-0001-7909-8837, and Nic Fishman
Abstract
Sociology has remained relatively insulated from debates about the “replication crisis.” Heeding calls to consider replication more deeply, we introduce a distinction between two types of research reforms that have emerged in the wake of the crisis: specification-restricting and specification-expanding reforms. Specification restriction is often promoted as a way to increase the repeatability of research findings by controlling false positives. We argue that this statistical justification is less compelling in observational contexts—where the assumptions licensing false-positive control are harder to sustain and rigid ex ante analytical commitments can instead enshrine fragile specifications—than in experimental contexts. We further argue that apparently low replication rates do not necessarily imply a high prevalence of false positive findings. Instead, we propose a replication framework centered on specification-expanding reforms, stronger incentives for confirmatory research, and meta-analysis. This approach equips sociologists to assess the repeatability of findings and build a more cumulative discipline.

28129

Nathan Hoffmann @nathanihoff.bsky.social · 23/09/2026

I'm excited to share my recent research on queer migrants with @demo-udem.bsky.social on Monday! Link below to join virtually. My presentation will be in English.

030

Reposted by Nathan Hoffmann

John Burn-Murdoch @jburnmurdoch.ft.com · 08/09/2026

Two extraordinary charts from today’s PISA test results: 1) School test scores continue to collapse internationally, underscoring how this is no longer a Covid effect but sustained decline. Those falls in reading and maths are equivalent to about two years of lost schooling.

41856344

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

Thanks for reading! This paper has been on quite the journey, beginning with my MS in Statistics that I did during my PhD. I deeply thank my MS committee Chad Hazlett, @jenniebrand.bsky.social, and Onyebuchi Arah. Full paper: doi.org/10.1016/j.ss... Preprint: doi.org/10.31235/osf... 7/7

0110

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

Overall, double robust methods with flexible estimators obtain lower error than linear regression or propensity score matching. However these differences are not dramatic. And non-double-robust G-computation with the same flexible estimators performs as well as the double robust methods. 6/7

150

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

I find similar results in linear simulations, a variety of data generating processes, both small and large sample sizes, and cases of misspecification. 5/7

150

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

I test these methods on 1000 cross-sectional, binary treatment simulations from the 2016 ACIC causal inference competition. Bias and error are fairly low across all methods, and the double robust ones don't necessarily come out on top. 4/7

150

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

While there's plenty of theoretical work on double robust methods, there are few applied evaluations. Double robust methods can be used with flexible or linear estimators, and I assess both cases. I also assess flexible and linear estimators in methods that are not double robust. 3/7

150

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

Social scientists have likely heard of double robust methods, but they're rarely implemented in fields like sociology. This paper provides a straightforward introduction to three of these methods — AIPW, TMLE, and DML — and compares them to simpler or traditional methods using simulations. 2/7

160

Nathan Hoffmann @nathanihoff.bsky.social · 04/09/2026

My paper "Double robust, flexible adjustment methods for causal inference: An overview and an evaluation" is out in Social Science Research. Spoiler: In simulations, double robustness appears less important than flexibility, and OLS does pretty well! 🧵👇 1/7 www.sciencedirect.com/science/arti...

24511

Nathan Hoffmann @nathanihoff.bsky.social · 02/09/2026

My department is hiring! "The Department of Sociology at Emory University in Atlanta, GA invites applications for a tenure-track Assistant Professor position in Sociology of Health, commencing in Fall 2027." apply.interfolio.com/192144

apply.interfolio.com

Apply - Interfolio {{$ctrl.$state.data.pageTitle}} - Apply - Interfolio

032

Reposted by Nathan Hoffmann

Jake Grumbach @jakemgrumbach.bsky.social · 17/08/2026

Sasha Gusev has a great new blog post on faculty hiring during the DEI era. Remarkable how much backlash there has been to what was an tiny and temporary bump in hiring Black, Latino, and Native faculty of about 1 person per R1 campus theinfinitesimal.substack.com/p/what-actua...

Demographics of R1 university faculty, 2001-2024
84.5% 80%
White 69.8%
60
40
20
8.9%
3.4%
2.9%
0.3%
2001
'05
'10
'15
'20
— Asian/PI 17.1%
lispanic 6.4%
Black 4.8%
'24 Other 1.9%

Hiring counts for tenure-track positions at R1 universities
4,000 hires
3,797
• Non-URM 3,753
3,000
2,000
2,254
1,000
623
623
298
→ URM 788
Black 348
2018
2019
2020
322
2021
2022
2023
2024

8430181

Nathan Hoffmann @nathanihoff.bsky.social · 10/08/2026

Listening to music on a train is pretty amazing

010

Nathan Hoffmann @nathanihoff.bsky.social · 20/07/2026

I contacted Stata technical support, and they sent me a link to a download a new build of Stata 15 that works on Tahoe! I can delay buying a new version for now...

001

Nathan Hoffmann @nathanihoff.bsky.social · 20/07/2026

Good to know that perpetual licenses still exist! I installed Stata 15 on my new mac a year ago and it worked fine until a recent software update...

010

Nathan Hoffmann @nathanihoff.bsky.social · 19/07/2026

I'm a total R convert and only use Stata to run other people's code (including the occasional collaborator's).

000

Nathan Hoffmann @nathanihoff.bsky.social · 19/07/2026

The perpetual license Stata 15 I bought at a student rate in 2017 is finally refusing to open on my iOS Tahoe MacBook Pro. I almost never use Stata, and the "subscription" model they use now seems like a rip-off... I might be breaking out my old MacBook when I need to run some Stata code.

300

Reposted by Nathan Hoffmann

Tuhin Chakrabarty @tuhinchakr.bsky.social · 22/05/2026

Ran some 🧪 to 🔬 why the Granta short story was certainly 🤖 generated. A lot of bad writing happens because AI hasn’t learned aesthetics. It has simply memorized the whole internet and called it a day. So sure, maybe you don't trust AI detectors. But you can trust your own eyes. #AISlop

1142

Reposted by Nathan Hoffmann

Tuhin Chakrabarty @tuhinchakr.bsky.social · 26/03/2026

🚨Paper on AI & Copyright Courts have credited AI companies' claims that alignment prevents reproducing copyrighted data. What if finetuning on a simple writing task breaks it. Worse: tuning on just one author (e.g., Murakami) unlocks verbatim recall of 30+ other authors' books (up to 90%) (1/n)🧵

25630

Reposted by Nathan Hoffmann

Public Opinion Quarterly - POQ @poqjournal.bsky.social · 12/02/2026

How do Americans view the increase in lesbian and gay migrants? Hoffman & Velasco find that Americans overall give no preference to lesbian and gay migrants; however, Democrats, atheists, and lesbian and gay Americans find them more deserving. Read now: doi.org/10.1093/poq/...

234

Nathan Hoffmann @nathanihoff.bsky.social · 20/01/2026

This paper helps disentangle Americans’ preferences for migrants’ presumed cultural similarity from economic potential and humanitarian merit. It also sheds light on an understudied but politically salient group. Thanks for reading! The full paper is open access: academic.oup.com/poq/advance-...

000

Nathan Hoffmann @nathanihoff.bsky.social · 20/01/2026

We do discover a number of heterogeneous effects: Democrats, atheists, and sexual minority respondents consider lesbian and gay migrants more deserving of admission than straight ones, while Republicans, Christians, and straight respondents see them as less deserving.

100

Nathan Hoffmann @nathanihoff.bsky.social · 20/01/2026

However, we find no preference for admission and a negative overall effect for cultural similarity: Americans do not prefer to admit lesbian and gay migrants over straight ones, and they perceive them as less culturally similar.

100

Nathan Hoffmann @nathanihoff.bsky.social · 20/01/2026

We thought Americans might be more welcoming of lesbian and gay migrants and to perceive them as more culturally similar, due to homonational norms, perceptions of lesbians and gays as economically productive, and humanitarian concerns around increasing oppression.

100

Nathan Hoffmann @nathanihoff.bsky.social · 20/01/2026

A new paper with @kvelasco.bsky.social is now online at @poqjournal.bsky.social! Using a conjoint survey experiment with 1,650 U.S. citizens, we ask, how do Americans take sexual orientation into account when evaluating immigrants' deservingness for admission? academic.oup.com/poq/advance-...

1114

Nathan Hoffmann @nathanihoff.bsky.social · 13/01/2026

Here are some photos from my mom's side, Eastern European Jews who migrated around 1900 to escape oppression in the Russian Empire.

000

Nathan Hoffmann @nathanihoff.bsky.social · 12/01/2026

I have my first lecture for my Sociology of Immigration class tomorrow, and I'm sharing my family's migration story. My dad scrounged up alien registration cards for Johann George Angerer and Kunigunda Seitz, my grandma’s grandparents, who migrated from a small German town near Nuremberg in 1854!

281

Nathan Hoffmann @nathanihoff.bsky.social · 12/01/2026

That's true, adding a don't know option would give this a lot more validity! Even so, family and other people I talk to in Kansas (where I grew up) have strikingly negative views of major cities.

010

Nathan Hoffmann @nathanihoff.bsky.social · 11/01/2026

Of course, crime in New York is at a historic low. And as a former LA resident, I'm appalled that only 34% of Americans think LA has good food! It is an incredible food city. www.nyc.gov/site/nypd/ne...

nyc.gov

NYPD ANNOUNCES SAFEST YEAR EVER FOR GUN VIOLENCE WITH FEWEST SHOOTING INCIDENTS AND SHOOTING VICTIMS

NYPD ANNOUNCES SAFEST YEAR EVER FOR GUN VIOLENCE WITH FEWEST SHOOTING INCIDENTS AND SHOOTING VICTIMS IN RECORDED HISTORY

110

Nathan Hoffmann @nathanihoff.bsky.social · 11/01/2026

Annoyed by long lines and throngs of tourists, a friend in New York feels like everyone in the U.S. wants to move there. I found this 2023 @today.yougov.com poll on major American cities, showing that only 9% of Americans think that New York is safe! Other aspects are not rated much better.

320

Nathan Hoffmann @nathanihoff.bsky.social · 01/01/2026

The two books that most blew me away were Home by Marilynne Robinson and Rejection by Tony Tulathimutte.

000

Nathan Hoffmann @nathanihoff.bsky.social · 12/12/2025

British cohort studies are excellent data and are easy to access. I've worked the most with the Millennium Cohort Study.

110

Nathan Hoffmann @nathanihoff.bsky.social · 11/12/2025

Curious if these proportions have changed over time. As a kid I would have said the future, but now I'd definitely say the past (more of an age than a period effect I think).

000

Reposted by Nathan Hoffmann

Saloni @scientificdiscovery.dev · 09/12/2025

Big new blogpost! My guide to data visualization, which includes a very long table of contents, tons of charts, and more. --> Why data visualization matters and how to make charts more effective, clear, transparent, and sometimes, beautiful. www.scientificdiscovery.dev/p/salonis-gu...

28802313

Nathan Hoffmann @nathanihoff.bsky.social · 10/12/2025

I just wrapped up teaching an intro stats course using this book, and I think it worked really well! bsky.app/profile/nath...

010

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

I deeply thank everyone who gave advice, as well as my students, who were so engaged and pleasant to teach. If anyone wants to hear more from me about putting together a new intro stats course, I'd love to talk; please message me. I'm looking forward to teaching this again soon!

010

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

In most of the five problem sets I assigned, students replicated part of a published social science paper. Below is a link to each paper: doi.org/10.1177/0003... doi.org/10.1177/0003... doi.org/10.1257/0002... doi.org/10.1257/0002...

doi.org

Sage Journals: Discover world-class research

Subscription and open access journals from Sage, the world's leading independent academic publisher.

110

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

With topics like data cleaning, visualization, and survey design, I tried to make the course useful for students who don't want to do quantitative work, but interesting for those with some background in statistics. To keep things engaging for the latter, I included Bonus problems in the group work.

110

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

As the main textbook, I chose "Data Analysis for Social Science" by @ellaudet.bsky.social and Kosuke Imai. I loved how this book introduced topics like causal inference and included accessible R code, with probability and statistics saved for the end. press.princeton.edu/books/paperb...

press.princeton.edu

Data Analysis for Social Science

An ideal textbook for complete beginners—teaches from scratch R, statistics, and the fundamentals of quantitative social science

151

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

I tried to make the course practical and hands-on. Each 75-minute class session, I lectured for about 40 minutes, then gave in-class group work exercises for the students to practice concepts and do some coding in R, graded for completion. At the end of each class, we went over solutions together.

130

Nathan Hoffmann @nathanihoff.bsky.social · 09/12/2025

With the semester finishing, I wanted to give an update about building this intro statistics course for PhD students in sociology. Thanks to everyone who gave advice! I ended up having a blast teaching it, and students seemed to like it. Here's the syllabus: docs.google.com/document/d/1...

docs.google.com

Soc 500 Syllabus F25

SOCIOLOGY 500 Research Methods and Models: Statistics Fall 2025, Tuesday & Thursday 1:00-2:15 pm Location: Tarbutton Hall Room 206 Instructor: Nathan Hoffmann Email: nihoffm@emory.edu Office: Tarbut...

1111

Nathan Hoffmann @nathanihoff.bsky.social · 24/11/2025

A few years ago I got selected for the ACS, but that was less exciting because it was just by mail.

020

Nathan Hoffmann @nathanihoff.bsky.social · 24/11/2025

Today was the most exciting day of my life: An in-person enumerator from the Census Bureau stopped by to interview me for the Current Population Survey! I can't wait to tell my students that I'm contributing to the data we use for statistics exercises.

130

Reposted by Nathan Hoffmann

Sasha Killewald @sashakillewald.bsky.social · 11/11/2025

I'm facilitating a causal inference reading group next semester for Sociology PhD students. (I will also be learning!) If there are (1) pedagogical articles or (2) empirical examples in soc that you ❤️, will you share in the comments? [And please RT to help me crowd-source!]

94427

Nathan Hoffmann @nathanihoff.bsky.social · 18/11/2025

Thanks for sharing it! It's under review...

010