Christian Oswald

EPSS Travel Grants for 2026 Belfast conference:

PhD students & junior scholars @ institutions in European countries that are often underrepresented, can apply for full fee waiver & £500 stipend

Apply: lnkd.in/ehXjhCgf
Deadline: March 8, 11:59pm GMT.

More details: lnkd.in/eZTP5sWR

🧵Cans of Worms possibly opened by Trump's war of choice in Iran, a running list:

(not exhaustive, with the caveat that I am not an expert on the region so these are questions from a US foreign policy perspective, please add others) 1/?

The train has left the station: Agentic AI and the future of social science research | Brookings A new era of agentic AI agents has begun. What does it mean for social scientists? Solomon Messing and Joshua Tucker discuss.

Are you a social scientist who has been hearing about Claude Code but hasn't checked it out yet? Or are you a social scientist who has been playing with Claude Code and has been wondering what it means for the future of social science research?

www.brookings.edu/articles/the...

1/

🧺 Paper Picnic 2.0 is here! More journals. New features. An easier way to keep up with the latest research in political science and adjacent fields. 🧵👇

1/ Sorry for double-posting from X. Sharing a new working paper for the Year of the Horce 🐎:

"An AI-assisted workflow that scales reproducibility in empirical research" (bit.ly/repro-ai) w/ Leo Yang Yang

Russia, Venezuela, Iran, China, the Sahel region, the United States ...

Want to know why state agents carry out brutal repression — or participate in illegal coups?

Our new book "Making a Career in Dictatorship" provides answers — it just got published by @academic.oup.com:

tinyurl.com/ystwm3tf

It must be very hard to publish null results Publication practices in the social sciences act as a filter that favors statistically significant results over null findings. While the problem of selection on significance (SoS) is well-known in theory, it has been difficult to measure its scope empirically, and it has been challenging to determine how selection varies across contexts. In this article, we use large language models to extract granular and validated data on about 100,000 articles published in over 150 political science journals from 2010 to 2024. We show that fewer than 2% of articles that rely on statistical methods report null-only findings in their abstracts, while over 90% of papers highlight significant results. To put these findings in perspective, we develop and calibrate a simple model of publication bias. Across a range of plausible assumptions, we find that statistically significant results are estimated to be one to two orders of magnitude more likely to enter the published record than null results. Leveraging metadata extracted from individual articles, we show that the pattern of strong SoS holds across subfields, journals, methods, and time periods. However, a few factors such as pre-registration and randomized experiments correlate with greater acceptance of null results. We conclude by discussing implications for the field and the potential of our new dataset for investigating other questions about political science.

I have a new paper. We look at ~all stats articles in political science post-2010 & show that 94% have abstracts that claim to reject a null. Only 2% present only null results. This is hard to explain unless the research process has a filter that only lets rejections through.

🧵 New version of our paper (@bcegerod.bsky.social) is finally online: "How Many is Enough? Sample Size in Staggered Difference-in-Differences Designs"
We show that even well-identified DiD studies are often underpowered; sample sizes needed are surprisingly large
Paper: osf.io/preprints/os... 1/6

🚨(Software) Update:

In my PhD, I had a side project to fix an annoying problem: when you ask 5 people to label the same thing, you often get different answers. But in ML (and lots of other analyses), you still need a single aggregated answer. Using the majority vote is easy–but often wrong.

1/N

GitHub - sebastianvanbaalen/causalplot: Create Causal Diagrams Using A Simple Convencience Function Create Causal Diagrams Using A Simple Convencience Function - sebastianvanbaalen/causalplot

📦 I got tired of constantly battling TikZ or Word for decent causal diagrams, so decided to do something about it. Here is a first public version of 𝗰𝗮𝘂𝘀𝗮𝗹𝗽𝗹𝗼𝘁, an R package for creating clean, publication-ready causal diagrams with minimal effort. #rstats

github.com/sebastianvan...

Some lecture notes on spatial and temporal modelling, illustrated using R - darrenjw.github.io/spatio-tempo... - #rstats #rspatial #quarto

Life cycle of a method
1. scrappy theory
2. killer empirical application (Angrist/Krueger for IV, Card/Krueger for DID, Lee for RD etc)
3. Deep, general theory
4. Review papers
5. Bad empirical applications
6. Empirical reviews

Proximal has skipped 2 entirely. Is this typical for epi methods?

A festive digital illustration features the blue R logo wearing a red Santa hat with white trim on the left. To its right stands a green Christmas tree made entirely of curly braces arranged in a triangular shape. White and light-blue %>% pipe symbols float across a dark blue, starry background like snowflakes. A speech bubble above the R logo says “Ho Ho Ho! 📦” in red text. At the bottom, white text reads: “R Package Development Advent Calendar 2025” and below it, “25 Days of Coding Magic • Dec 1–25.

My #RPackageAdvent2025 recap is live! 📦✨ Get the complete guide to modern R package development: setup, docs, testing, CI/CD, and CRAN submission. All 25 days of tips, tools, and best practices in one place! Dive in! #RStats #RPackageDevelopment
Link: https://drmo.site/8ny3eU

Hey #rstats and #stats people, does anyone know of any references that explain the hazards of trying to run or interpret models with all combinations of many predictors? (i.e., including multiple 2-way and 3-way interactions)

course schedule as a table. Available at the link in the post.

I'm teaching Statistical Rethinking again starting Jan 2026. This time with live lectures, divided into Beginner and Experienced sections. Will be a lot more work for me, but I hope much better for students.

I will record lectures & all will be found at this link: github.com/rmcelreath/s...

🚨Jobs!🚨

3-year PostDoc positions (aka Research Officers) at @lsegovernment.bsky.social to work with me on the local consequences of border change. Please reach out for questions and apply by Jan 4th to join the team and department: I’d love to hear from you!

Job ad: jobs.lse.ac.uk/Vacancies/W/...

In light of record submission rates and a large volume of AI-generated slop, SocArXiv recently implemented a policy requiring ORCIDs linked in the OSF profile of submitting authors, and narrowing our focus to social science subjects. Today we are taking two more steps:
/1

The Politics of Human Rights | Cambridge Aspire website Discover The Politics of Human Rights, 2nd Edition, Sabine C. Carey, HB ISBN: 9781009537919 on Cambridge Aspire website

📣 out now, with @argohdes.bsky.social and Mark Gibney - and in honour of the wonderful Steve Poe, who motivated the 1st edition with @universitypress.cambridge.org
tinyurl.com/2s4kh6z3. If you enjoy reading this even a fraction as much as I enjoyed creating it, you’re in for a ride 🤩

We strongly suggest using the following labels: praise: Praises highlight something positive. Try to leave at least one of these comments per review. Do not leave false praise (which can actually be damaging). Do look for something to sincerely praise. nitpick: Nitpicks are trivial preference-based requests. These should be non-blocking by nature. suggestion: Suggestions propose improvements to the current subject. It’s important to be explicit and clear on what is being suggested and why it is an improvement. Consider using patches and the blocking or non-blocking decorations to further communicate your intent. issue: Issues highlight specific problems with the subject under review. These problems can be user-facing or behind the scenes. It is strongly recommended to pair this comment with a suggestion. If you are not sure if a problem exists or not, consider leaving a question. todo: TODO’s are small, trivial, but necessary changes. Distinguishing todo comments from issues: or suggestions: helps direct the reader’s attention to comments requiring more involvement. question: Questions are appropriate if you have a potential concern but are not quite sure if it’s relevant or not. Asking the author for clarification or investigation can lead to a quick resolution. thought: Thoughts represent an idea that popped up from reviewing. These comments are non-blocking by nature, but they are extremely valuable and can lead to more focused initiatives and mentoring opportunities. chore: Chores are simple tasks that must be done before the subject can be “officially” accepted. Usually, these comments reference some common process. Try to leave a link to the process description so that the reader knows how to resolve the chore. note: Notes are always non-blocking and simply highlight something the reader should take note of.

I recently discovered Conventional Comments (conventionalcomments.org) for providing a pseudo-standard set of labels for feedback and just tried it for an article review and it was really helpful to specify issues vs. thoughts vs. suggestions, etc. Hopefully it's helpful for the authors too!

Computational Turing Test Reveals Systematic Differences Between Human and AI Language Large language models (LLMs) are increasingly used in the social sciences to simulate human behavior, based on the assumption that they can generate realistic, human-like text. Yet this assumption rem...

LLMs are now widely used in social science as stand-ins for humans—assuming they can produce realistic, human-like text

But... can they? We don’t actually know.

In our new study, we develop a Computational Turing Test.

And our findings are striking:
LLMs may be far less human-like than we think.🧵

Spatiotemporal causal inference with arbitrary spillover and carryover effects: Airstrikes and insurgent violence in the Iraq War Social scientists now routinely draw on high-frequency, high-granularity ''microlevel'' data to estimate the causal effects of subnational interventions. To date, most researchers aggregate these data...

🚨New paper🚨: Are you interested in key debates about civil war and counterinsurgency? Like questions about causal inference but wish you had a better method for dealing with spillover and carryover effects? Then check out our new paper:

arxiv.org/abs/2504.03464

Thanks to the great help by @rstrauch.bsky.social, Gabriele Spilker and the Cluster of Excellence "The Politics of Inequality" @excinequality.bsky.social, version 5.0 of the "Mass Mobilization in Autocracies Database" (coverage until 2022) is now available at mmadatabase.org!

If your random seed is 42 I will come to your office and set your computer on fire🔥 Figuratively. More likely you'll get a stern talking to.

We need to have a conversation about random seeds. Don't use 42.
blog.genesmindsmachines.com/p/if-your-ra...

Colleagues: What’s your favorite, accessible resource (reading, video, slide deck) to convince students that regression on observational data—even with many controls—is not often causal. (Quasi-)experimental designs are preferred for causal inference.

Simulate Data from a DAG and Associated Node Information Simulate complex data from a given directed acyclic graph and information about each individual node. Root nodes are simply sampled from the specified distribution. Child Nodes are simulated according...

I'm starting to explore {simDAG}, a nifty #rstats package that has a nice API for specifying DAGs and then simulating data from them: robindenz1.github.io/simDAG/

If you have two measures of the same confounder, you can just include both of them in your regression model Sometimes, researchers worry about multicollinearity in situations where it’s actually a non-issue. Here’s one such scenario. Imagine a situation where you are interested in the effect of X on Y (X...

New blog post!

Let's say you have two measures meant to capture the same confounder. They're highly correlated. Can you still proceed with your regression analysis?

(I admit, the title is a bit of a spoiler)

www.the100.ci/2025/10/13/i...

New Working Paper 🚨

I am introducing the new dataset on Parliamentary Elites in Authoritarian Regimes (PEAR). It includes information on the composition of parliaments in 819 legislative terms in 130 autocracies from 1945 - based on over 200k individual-level observations.

tinyurl.com/48h68hmh

Why do authoritarian states charge political opponents with non-political crimes? In our @thejop.bsky.social paper with Jennifer Pan & @yiqingxu.bsky.social, we examine how *Disguised Repression* undermines opponents’ moral authority and mobilization capacity. doi.org/10.1086/7342...

GitHub - nicolaiberk/llm_ws: Materials for my Workshop on LLMs Materials for my Workshop on LLMs. Contribute to nicolaiberk/llm_ws development by creating an account on GitHub.

I had the pleasure to teach a 3-day crash course on #LLMs for PhD students.

We covered:

1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences

The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...

a table about lemurs

a table about students and schools

a table about wines

{tinytable} 0.14.0 for #RStats makes it super easy to draw tables in html, tex, docx, typ, md & png.

There are only a few functions to learn, but don't be fooled! Small 📦s can still be powerful.

Check out the new gallery page for fun case studies.

vincentarelbundock.github.io/tinytable/vi...