I have a new working paper out today, looking at the impact of educational use of ChatGPT/AI on student test scores. Importantly, this is overall student-led use (incl for cheating!), not specialized AI tutoring tools. To my surprise, I found... no effect at all! arxiv.org/abs/2605.08812

Found this account @celestialsem.bsky.social and it's my new favorite thing on bluesky.

Yes good to see this. No published clusters to my knowledge have even been validated statistically (see below) much less medically. fharrell.com/post/cluster #Statistics #StatsSky

I've started including @djnavarro.net 's #rstats {sessioncheck} in code I write for various clients' projects - it helps me keep my sessions reproducible! Anyway.. a client recently asked for one of my scripts, and when they sent it back I had a good laugh seeing the session check commented out 😂

I feel like they murdered Galileo for much less than whatever this is

When I was still a neurocognitive psychologist it always seemed absurd when finding something in the brain was presented as if it validatwd the realness of the behavioral finding. Where do these people think behavior comes from? I was thought we settled the dualism/monism debate a while back...

Oh my. This is spectacular. sinceyouarrived.world/taken

the only good thing* about having to use the delta method is getting to make @drob.bsky.social’s “(Taylor’s Version)” joke when writing it up *besides the fact that it “gets me” the “answer that I need”🙄

The five stages of grief, but for when you realize that your design calls for a more complex random effects structure and now your fixed effects are no longer significant.

When I taught academic writing, I used to think that a cool drinking game would be to take a shot anytime a students' paper started with "since the dawn of time..." or "one of the most interesting unanswered questions is...". But of course I'd be dead.

Let's meet Daniela Witten ⭐️ Daniela ✨ is a biostatistician and Professor at the University of Washington ✨ focuses on high-dimensional statistical learning ✨ co-authored 'An Introduction to Statistical Learning' (with #rstats and #python) 🔗 https://en.m.wikipedia.org/wiki/Daniela_Witten

Listening to friends talking about someone who got "sucked in by a MLM and just seems like a completely different person now" Me, nodding along in agreement: Yes, multi-level modelling really is amazing

Things I'm critical of: LLMs Things I'm a fan of: LMMs aka MLMs I'm a typo away from getting a massive Silicon Valley grant for running a repeated measures ANOVA... 🤦‍♂️ The difference between a few dozen or hundred parameters and a few billion parameters is can be staggering!

Disclaimer: this is my first #quarto extension, I don't know what I'm doing That said, for who knows what reason I've always wanted a RevealJS feature where you can summon Wayne Knight to deliver his iconic "ah ah ah" line from Jurassic Park You're welcome (sound on) github.com/benharrap/ah...

Maybe teaching #rstats was a mistake? When I first advocated for teaching R in under/grad psych, it felt like the right move: SPSS was rigid and limited only to what can be selected from a stat-analysis decision tree; R on the other hand... 1/

Film you've watched more than six times with gif hard mode: no star wars, star trek, lotr, dc, marvel, disney, pixar, dreamworks, spielberg, cameron, wachowskis, or ghibli

This is such a great tutorial! If you run mixed models, learning how to tease apart state vs. trait differences is an absolute game-changer. Check out this guide from the #rstats #easystats {modelbased} package. 📊👇

What a fantastic episode. We have at our disposable so so many models to capture (co-)change over time - the second half of this episode where the research questions implied by each model is spelled out is very useful for those less experienced with these models (me). #stats

#rstats The idea of a pipe (`|`) operator goes back to early Unix. Eg: `hist | head -5` to see your most recent commands. But, where & how did arrive in R #tidyverse thinking and, importantly: in R syntax? Here's an interesting post that describes this history. adolfoalvarez.cl/blog/2021-09...

Here's a #statsmeme I sometimes use to talk about simplistic, uncritical outlier rejection rules But there's another size of the coin: Sometimes the unusual data is a true discovery

Please enjoy some of my finest work. In collaboration with my sister, Teesa D’Agositno, we bring you “I’m Just a Slope” a tribute to our love of statistics & The Lonely Island youtu.be/9EwUl6gyVGQ?...

Like to me it's a fascinating example of... capitalism is very good at giving you what you ask for, so be careful what you ask for. If you're saying "I want anti-AI images" and the best generator of those images is AI, well, the machine has no sense of irony.

27 new videos on research methods and statistics are now live on my YouTube channel! 🧵 www.youtube.com/playlist?lis...

This is important! Over the last few years I've been reporting Shrout and Lane's R{C} measure of within person change ("state") reliability and I've found that it's either very very low or very very high (rare 🥲), nut never in the middle... ofc the "trait" reliability is almost always very high...

Do you ever claim to be interested in the "association" between variables? This talk may be for you! @jeremylabrecque.bsky.social will explain when and why such associational language is not appropriate and why causal language should be used instead. Register here: ucl.zoom.us/meeting/regi...

Good news everyone 🥳 Our (w @vincentab.bsky.social) primer on models as prediction machines (with the marginaleffects package) is finally officially published!> journals.sagepub.com/doi/10.1177/...

There are only a few spots left in our Missing Data Summer School, so be quick if you want to participate! Feel free to reach out if you want any additional information on the content or the programme. If you want to sign up, you can do so here: utrechtsummerschool.nl/courses/data...

Nothing is certain except: - death - taxes - posterior intervals with overly narrow priors

If you are doing any #quarto slidecrafting, i have been trying really hard to keep this page updated with everything I have done emilhvitfeldt.com/project/slid...

This is hilarious

Yes, I am biased, but this is why, in my opinion, Apple won the AI race without being part of it 🧵👇🏼 "It's the economy, stupid." www.indiatoday.in/technology/s...

In ML, everyone’s Humpty Dumpty statmodeling.stat.columbia.edu/2026/04/17/i...

„The only remaining way to mitigate this situation is to reduce the number of applications.“ They’ve tried nothing and are all out of ideas? Whether this is good or bad, it is unjustified. There is no impact analysis, no engagement with research on funding procedures. Government by vibes

I’m hiring a PhD student! The candidate will work alongside @zefreeman.bsky.social, who is joining our research group as postdoc. jobs.unibe.ch/job-vacancie...

Staring at my mother-in-law's quilt... it feels very... #tidy... #rstats

But there's a lot of research showing that grant decisions are unreliable and a widespread feeling that the way to win is to just roll the dice often enough. So, this may not be a very selective way to reduce submissions.

When you just want heteroskedasticity-consistent standard errors, but AI thinks you're hungry for baked goods. #rstats

Science is a cumulative and collaborative process that requires human connection and interaction. I find it interesting that the 'lone genius' idea still gets romanticised in academia.

Maimonides (1190) on embracing uncertainty. This is very applicable to modern statistical analysis...

I will never not enjoy the look of sheer horror on students' faces when I show that variables with the largest standardized coefficients ("beta") aren't necessarily those with the largest unique contribution to R². Variable importance - it's hard!

After basically being closed for ~40 days due to the war, the university opened yesterday and wow, everyone was super happy to see everyone else! Like "oh, you made it too! <hugs>" May we live through some boring times for a change...

Are joint predictive models enough, or do you actually need casual inference? notstatschat.rbind.io/2026/04/13/a...

Wow, this is some wild stuff.

Just learned about the {anvil} #rstats 📦, analogous to Jax. Write R code that gets JIT compiled and has easy autodiff. Seems promising for ML and Bayesian methods. r-xla.github.io/anvil/

🚨 Come to iIIRG 2026 to hear our talk about better methods to analysing interview data! @drcodyporter.bsky.social @mattansb.msbstats.info #belgium #conference #methods #investigativeinterview

Trying to implement a statistical method in #rstats and found it really requires some structured objects and ... damn, {S7} is smooooooth!

🚨New preprint and our results are rather concerning.. We find the "boiling frog" equivalent of AI use. Using large-scale RCTs, we provide *casual* evidence that AI assistance reduces persistence and hurts independent performance. And these effects emerge after just 10–15 minutes of AI use! 1/

I'm reviewing Sapiens, one page at a time! It's a little bit fact-checking and a little-bit soul-searching combined into a lot of feelings... ecodevoevo.blogspot.com/2026/03/sapi...