Anton Strezhnev

I’m teaching a course on data wrangling with R & the tidyverse for @stathorizons.bsky.social in January. Filled to the brim with ways to do the stuff you have to do before you can run any of your spiffy models and that often breezily gets glossed over even though it’ll take up 90% of your time.

With vodka.

You have the exact same pathology in elite legal circles.

If a discipline isn't science because you can't make point predictions I've got some bad news about meteorology, medicine, pharmacy, biology, and a bunch of other physical sciences.

“¿Esos son Reeboks, o son Nikes?”

Study of the effects of officer race/gender on police interactions - rich set of controls + a story for why residual variation is exogenous (arbitrary “time off” schedules). Also careful attention to post-treatment quantities www.science.org/doi/full/10....

Lawmaking, 2035

The Taboo Against Explicit Causal Inference in Nonexperimental Psychology - Michael P. Grosz, Julia M. Rohrer, Felix Thoemmes, 2020 Causal inference is a central goal of research. However, most psychologists refrain from explicitly addressing causal research questions and avoid drawing causa...

in psych, whole subfields have decided to simply skirt the issue, so the causal inference happens in the introduction of the next paper that cites the "totally-not-causal-I-swear" paper and of course absolutely interprets it causally 🥲

I loved how they made the Coral Tower gauntlet so insufferable to incentivize you to complete that wonderful storyline instead.

Favorite boss?

The differences-in-PO estimands do have an interpretation in terms of expectations over vote share changes in hypothetical (electoral) contests, but I think this is often stretching plausibility as an estimand of interest for most scholars. www.cambridge.org/core/service...

The randomness in the design is motivated by the infeasibility of observing all profile combinations - essentially “sampling” from the distribution of choices osf.io/preprints/so...

I think the distinction here is that the conjoint is still descriptive in the sense that I could expose you to one profile with a man and another with a woman and observe both choices under a no spillover assumption. So we’re still talking about the same underlying style of task

For all values of covariates, the probability of being treated is bounded away from 0 and 1. Why, what do you think positivity is?

Auric Goldfinger

That’s fair - and 2019 Nobel wasn’t just “experiments” but experiments in the service of policy effectiveness. So maybe “program evaluation” is the common thread here - in line with Cyrus’s “problem solving” research piece.

“Credibility revolution” is really about observational designs. The “experimental revolution” I think is related but distinct (see 2019 vs 2021 Econ nobels)

This hits on a super important point - a large share of survey experiments aren’t actually targeting a causal effect; they’re preference elicitation exercises.

One of the moves needed in deepening and widening the "crediblity revolution" is *de-emphasizing* the methods!

I care a lot about causal inference - I use it in my work, I teach it, I try to (unsuccessfully) advance it - but we often focus too much on the mechanics of specific designs.

New paper! @william-dinneen.bsky.social @guygrossman.bsky.social Yiqing Xu and I use GPT to code 91k articles from 174 polisci journals (2003–2023)and track research designs, transparency practices, and citations. How has the credibility revolution reshaped the discipline? doi.org/10.31235/osf...
🧵

you cant fool me, upshot bluesky account. i know that the computer program for the "live forecast" is called needle.r. don't lie bsky.app/profile/upsh...

The UW’s Mini Movie Palace | On Wisconsin Cinematheque offers a world-class filmgoing experience.

Cinematheque offers an artfully curated selection of Hollywood classics, international cinema and challenging indie fare—for free! @uwmadison.bsky.social 's unique program reaches out to moviegoers who crave an alternative to the latest multiplex blockbuster.
onwisconsin.uwalumni.com/the-uws-mini...

2025-12-01 self-promotion Lal and Woodward is a work-in-progress poster on a decision-theoretic motivation for tiebreaker designs, which are a hybrid of RCTs and RDDs.

New linklog - tiebreaker designs, permutation testing, compressed DML estimation, new bandit library, and a bunch of interesting papers.
apoorvalal.github.io/lalgorithms/...

honestly tho, now that we've got "good enough for well-specified coding problems" language models that can run on off-the-shelf macbooks, im officially anti- all of the AI companies except Anthropic, whom i experience as an interesting digital humanities graduate department

“destructive robustness testing”

I have now read the response by Turnbull-Dugarte and López Ortega again with a clearer eye.

Multiple claims in this response are untrue or seek to obscure the truth.

My interest has always been the accuracy of the scientific record, so I feel compelled to note some of these claims here.

Today I published a replication outlining concerns with "Instrumentally Inclusive" by Turnbull-Dugarte and López Ortega (2024, APSR).

I document seemingly idiosyncratic and ad hoc choices made by the authors that create a pattern of statistically significant results consistent with their theory.

Ah, an empiricist.

Stamps featuring Kate Beaton, Jimmy Beaulieu, Guy Delisle, Julie Doucet, Bryan Lee O’Malley and Michael Nicoll Yahgulanaas

Canada is releasing a set of stamps celebrating the nation's graphic novelists.

These stamps feature Kate Beaton, Jimmy Beaulieu, Guy Delisle, Julie Doucet, Bryan Lee O’Malley and Michael Nicoll Yahgulanaas, alongside some of their famous works.

Political Deepfakes Are as Credible as Other Fake Media and (Sometimes) Real Media | The Journal of Politics: Vol 87, No 2 There is widespread concern that political “deepfakes”—fabricated videos synthesized by deep learning—pose an epistemic threat to democracy as a uniquely credible form of misinformation. To test this ...

“However, in a follow-up discernment task, subjects often confuse authentic videos for deepfakes if the video depicts an elite in their political party in a scandal.” www.journals.uchicago.edu/doi/abs/10.1...