Interesting paper on how software could help analysts. ”Causal clarity in statistical software” (Korf et al., 2025). I like many of the ideas, such as clarifying estimand, using multiple estimators, and no output of coefficients for adjustment variables doi.org/10.1093/ije/... #stats #rstats

I'm exited to announce a new resource about making slides with quarto and revealjs. This book is the combination of all the work I have done in this area, reordered and polished up There isn't a lot of new information yet, but this format allows me to add more easily slidecrafting-book.com #quarto

I hope Melvin Bragg's retirement will not end In Our Times. He has a lovely voice and manner but it's the people he invites that hold the show. I'll take this chance to post a thread of some of the more memorable episodes for me. Starting with Schaffer, Worrall, and @michelamassimi.bsky.social

Indeed, and since data is always collected within a known time-frame, there is also a known upper bound, which count models generally do not respect. The impact of ignoring the upper bound (and potential one-inflation) will of course depend on the distribution.

For instance, yesterday I read a paper with a table describing participants' sickness absence days with a mean of 71 and SD = 88. Generating a random (gaussian) sample using these values produces ~20% participants with less than zero sick days.

My impression is that papers in psychology often seem to assume all variables are gaussian and well described by mean/SD as summary stats. Having actually looked at a lot of data distributions in psych has made me a proponent of Tukey’s `fivenumber()` summary statistics.

The Disney brand, and the USA brand, sinking together. Parody from Dutch TV www.youtube.com/watch?v=7vRX...

Wonderfully snarky title.

Most point estimates of reliability have two main problems. One is that they typically describe properties of a sample, not a measure. Then they make it seem like reliability is a constant across the continuum measured, which is unlikely.

Skimming a recent psychometric paper on the Perceived Stress Scale, 4-item version. Authors find 2 factors needed to fit the data, but still go ahead and use the ordinal sum score from all 4 items in various follow-up tests. Where does this practice come from? #psychometrics

been saying this for years: stop using academia[.]edu. This is real bad folks.

Whoa—my book is up for pre-order! 𝐌𝐨𝐝𝐞𝐥 𝐭𝐨 𝐌𝐞𝐚𝐧𝐢𝐧𝐠: 𝐇𝐨𝐰 𝐭𝐨 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭 𝐒𝐭𝐚𝐭 & 𝐌𝐋 𝐌𝐨𝐝𝐞𝐥𝐬 𝐢𝐧 #Rstats 𝐚𝐧𝐝 #PyData The book presents an ultra-simple and powerful workflow to make sense of ± any model you fit The web version will stay free forever and my proceeds go to charity. tinyurl.com/4fk56fc8

This is a paper I really care about. I feel the core message is very important for social scientists in general, and political scientists in particular. "Quantitative Research in Political Science is Greatly Underpowered." (with A+ co-authors)

This should have been big news! Ten funding agencies from eight European countries have pledged to support a public infrastructure that is poised to replace academic journals: FWF 🇦🇹 RCN 🇳🇴 Forte 🇸🇪 ARIS 🇸🇮 SRC 🇸🇪 FCT 🇵🇹 CSIC 🇪🇸 DFG 🇩🇪 Formas 🇸🇪 ANR 🇫🇷 Only two of them issued press releases in […]

Matching the Forbes 400 to tax data finds that they pay a total tax rate of 24 percent of economic income, lower than the 30 percent tax rate paid on average in the US, from Akcan S. Balkir, Emmanuel Saez, Danny Yagan, and Gabriel Zucman https://www.nber.org/papers/w34170

What to do with academic publishing: youtu.be/XmttiQnlJfw?... - Support open communication platforms - Support "publish, review, curate" communications I'd also add: - Get angry - Recognize that business as usual (submit to El$evier et al) is not value-free - Stop reviewing for big pub

Preprint and the website by @dingdingpeng.the100.ci and @vincentab.bsky.social are great, and I got inspired to illustrate Bayesian workflow for model checking and comparison *before* model interpretation following their friendship importance example users.aalto.fi/~ave/casestu...

What if we thought about digital behaviour not in terms of the form that it takes but the psychological functions that it serves? In this preprint we introduce technological behaviourism, a theoretical approach grounded in operant conditioning. @shuhbillskee.bsky.social

Everyone should follow this lead. Incredibly simple to switch with minimal friction. #linux

#statstab #405 Best Practices for Estimating, Interpreting, and Presenting Nonlinear Interaction Effects Thoughts: Guidance on nonlinear interactions, reporting (probabilities) and visualisations. #probit #logit #logisticregression #nonlinear #guide sociologicalscience.com/download/vol...

I haven’t been this excited for a bubble to burst since I was a toddler.

Two more useful papers regarding SuperLearner and TMLE: Phillips, et al. (2023). Practical considerations for specifying a super learner. doi.org/10.1093/ije/... Balzer & Westling (2023). Demystifying Statistical Inference When Using Machine Learning in Causal Research. doi.org/10.1093/aje/...

I wrote an R package that creates standardized R project structures that are compliant with @mekline.bsky.social's psych-DS...ish. It also creates additional features for reproducibility and teaching like a readme, license, .gitignore and Quarto templates + can validate existing projects

Long read but informative. Hadn't seen figures based on personal use before. www.technologyreview.com/2025/05/20/1...

Grateful to have this view a few minutes walk from home.

For categorical and ordinal data a series of calibration plots can be used. The plots below show one of these calibration plots for Model 1 and Model 2 (the same as in the first post in this thread). The red line going most time outside the blue envelope indicates that Model 1 is misspecifed. 3/4

Not sure how little ole me got @rmkubinec.bsky.social, @chelseaparlett.bsky.social and @matti.vuorre.com to write a thing with me, but it happened! We wrote a Beta tutorial showcasing Beta and extensions of Beta (like ordered Beta). A preprint is available here: osf.io/preprints/ps....

Reminder that all three books I've co-authored are freely available online for non-commercial use (and the fourth will be, too)

Inspired by a paper (linked) that added a response category to the PHQ-2 and GAD-2 screener questionnaires we did the same in Swedish with good results in a sample of 15-18 year olds. The uppermost categories performed worse, though (q25-26=GAD,27-28=PHQ). #psychometrics osf.io/preprints/ps...

A curious finding from using the `dynamic` R package for CFA cutoffs is with 3 unidimensional factors that each shows bad fit, but a 3-factor model is ok. I expect that this is due to multi-factor models only considering cross-loadings as source of misfit, and ignoring local dependence issues.

Today, our article "The entities enabling scientific fraud at scale are large, resilient, and growing rapidly" is finally published in PNAS. I hope that it proves to be a wake-up-call for the whole scientific community. reeserichardson.blog/2025/08/04/a...

"The problem with AI isn't that it can do your job. It can't. The problem with AI is that your MBA-brained boss's boss doesn't know how your job works and thinks AI can do your job at fractions of a penny on the dollar, and hears the siren song of 'maximize shareholder value'." MBA-brain is real.

It turns out that pretty much no one who makes decisions in tech actually WANTS to be told that the product they're thinking about buying from that nice vendor is actually money-wasting bullshit. This strikes me as dangerous.

Reading the preface to "Thinking through statistics" and this section already made me like the book.

Another reason to use DuckDuckGo as your primary search engine: lifehacker.com/tech/how-to-...

Our (@malte.the100.ci, @ianhussey.mmmdata.io) commentary "cognitive dissonance in large language models is neither cognitive nor dissonant" has just been accepted for publication at PNAS! Check out the thread below for a summary of our arguments. osf.io/preprints/ps...

the enshittification of education is underway restofworld.org/2025/colombi...

Overwhelmed by inequality or climate inaction? You're not alone—and you're not powerless. In Rebooting Capitalism, Anthony Biglan shares a plan to turn distress into effective, collective action. 📘 snapt.io/xhGEF #RebootCapitalism #ValuesToAction Photo: Vectzeey

Might be the perfect guitar string gauge? Seems ridiculous, but .005 inches thinner than the .011 I've used forever on my solid body seems really nice.

Turns out plonking a university down in the 1800s was a long-term investment in civic life. Kinda wish current policy makers would take note...

New #rstats blog up! solomonkurz.netlify.app/blog/2025-07... This is the first in a new series discussing causal inference with experimental data using multilevel models. My basic case is g-computation is the way to go.

IMPORTANT: The Dutch invest €595 million annually on urban biking, resulting in €19 BILLION saved in public health care costs alone. That’s how smart govts do the math on investing in better mobility. Let’s be clear— it wastes public money to NOT do it. #CityMakingMath HT @modacitylife.bsky.social

#statstab #389 Risk Ratio, odds ratio, risk difference: Which causal measure is easier to generalize? Thoughts: Use this with #388 for better insight into risk, odds, probabilities, and more. Courtesy of @pgmj.bsky.social #odds #oddsratios #riskratios #probability arxiv.org/html/2303.16...

*whistles* Quite the list

Thanks to everyone who responded to my question about what folks should know about using LLMs to generate #rstats: I've summarised the feedback with a few of my thoughts at gist.github.com/hadley/77d42...

library(qs2) is so much faster than saveRDS(), especially with multithreading, and achieves better compression. I ran some mice-imputed models with 20 datasets using brm_multiple() and file sizes got a bit out of hand. Saving 50% space with 10-15x save/read time is neat. #rstats

Small things can improve your work environment surprisingly much. rectangleapp.com is a free window manager for MacOS. Works really well and is easily customized.

A nice recent article on why you should abandon hazard ratios. #statssky #episky