Me reading the abstract: Well, 'coding error' is a bit strong, isn't it, they just disagreed on some variable coding. Me getting to p. 3: Oh, OOOHHH, OH NO!!!!

My other concern would be the lack of a third group of raters who did not author or cite one of the articles with pubpeer comments. Currently, the survey presents free floating claims that pubpeer comments did not affect conclusions - even when retracted. Would objective readers agree?

What riles me up about “people tried to shut down PP research” is that Cuddy’s own coauthor-turned-critic Carney guest-edited a whole special issue of Registered Reports on Power Posing (incl a meta-analysis), which shows strong evidence against PP even though most authors clearly weren’t sceptics.

New blog post: Evaluating Dr. Cuddy’s Claim that the Debunking of Power Posing is a Myth. daniellakens.blogspot.com/2026/05/eval... On an AI generated description of a non-existent study, incorrectly citing findings from studies, and the importance of scientific criticism.

Earlier this week I posted an example of a fake western blot provided by ThermoFisher to demonstrate the validity of a p53 antibody. I considered it an amusing curiosity. In fact Thermo Fisher Scientific has systematically manipulated antibody validation data!

Austin et al. (2026) advocated for AI in hypothesis-driven scale validation research. In this preprint, led by @wchiggins.bsky.social, we (@malte.the100.ci @bethclarke.bsky.social) sketch out several weaknesses in this paper, and offer recommendations for avoiding them. osf.io/preprints/ps... 1/🧵

His exact words were, "These A.I. errors do not, in fact, diminish the larger questions that the book raises about truth, trust and A.I. and its impact on society, democracy and editorial." The book is THE FUTURE OF TRUTH. Indeed.

This paper has now been accepted at Advances in Methods and Practices in Psychological Science!

In fisheries models, natural mortality (M) is often set to 0.2 per year. A detailed derivation of this result was provided by John Pope in 1975, see below:

I'm crabby today but I wish we could have a little less pearl-clutching over "the children are citing AI hallucinations" and a little more over "tenured academics are citing documents that cite documents that cite documents that cite documents that don't support their claims."

After about 10 unpublished studies measuring every form of inhibitory control and every form of impulsivity and self-control we gave up on the null and void reality we had discovered. I'm sure the null results were due to our inability to properly wield the cog neuro measures...4/

Some research fields have a type of paper that basically asks: "What if a more expensive but less reliable machine can do [this] instead?"

Looks like "AI did most of the thinking" for this paper

Eagle-eyed @grinschglsandra.bsky.social spotted this mess of a plot, with bars that don't match percentages. We, with @malte.the100.ci, quantify a host of other issues we spotted here. pubpeer.com/publications...

Done with my last talk before leave! This one was fun and I'm glad I added Columbo last minute. Slides here, in case you're curious: juliarohrer.com/wp-content/u...

Write for us! We’re looking for people to write many more pieces for us at Works in Progress – including on the story behind superglue, why there’s no syphilis vaccine, how gas turbines became efficient, and many more in this post. If you or someone you know would be interested, please pitch us :)

I've wanted an R fiddle for a while, so I had Claude make one. fiddle.rforms.org - completely client-side thanks to webr, no cookies, no external reqs without permission - link to scripts via url param (? or #) - rmd/qmd support

“If researchers observe Cohen’s d = 4, no they didn’t.” Wolfgang and I have an idea for how to flag errors that produce erroneous large effect sizes during analysis, when they can be corrected. Feedback welcome!

Ian Hussey (@ianhussey.mmmdata.io) contacted me about the recurring problem of unusually large SMD values appearing in published meta-analyses. Very large SMDs can result from using SEs in place of SDs in their calculation, which is a mistake I have also seen in practice. 1/4

Wie man ins LLM hineinruft, so schallt es heraus archive.is/6RdK9#select...

Not an original thought, but for *reasons* this morning I am feeling intensely aware of the fact that the ability of spreadsheets to simultaneously serve as data entry, storage, and processing devices is the key to both their widespread use and the near-certainty of their footgunization.

it me

DeepSeek 3.2 was a very horny LLM, but DeepSeek V4 is somehow even *more* horny.

Every system or con that depends on the argument "nobody's going to take the time to do that" is going away. AI will take the time to do that.

One of the few days I regret leaving Berlin

Anybody got relevant experience doing marginal effects in the context of within-between/“hybrid”/Mundlak models?

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/...

common problem of debugging claude with claude is that anthropic change it so often that claude isn't up-to-date as to its own internals. MUST NOT anthropomorphise this with some self insight bullshit.

1/ "Silicon samples" are becoming more and more common in research and polling. One problem: depending on the analytic decisions made, you can basically get these samples to show any effect you want. The updated version of this preprint is now online! THREAD🧵 arxiv.org/abs/2509.13397

I’m sorry but this is an absolutely insane conclusion for the authors to draw given what they actually did

1. The paper with the implausibly large effects of Omega-3 fatty acids on mental health was now retracted. A little thread on the process where @ianhussey.mmmdata.io and I was involved. www.sciencedirect.com/science/arti...

Ian is smart, and he has integrity. You want to work with him.

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...

The number of submissions will only increase and I believe their solution is a poor fix. But my friends and I disagree about better solutions. To make resubmission intervals longer for people who get a low rating acts as a penalty for low ratings. If these ratings reliably evaluated quality: good.

We are inviting applications for a two-year postdoctoral position in a collaborative meta-science project on the effectiveness of data and code sharing policies in research-performing organizations. www.tue.nl/en/working-a...

This is such a strong comment, published in 2022. It has 14 Google Scholar citations. The replicated paper has 65 citations in 2025 alone - and counting. www.aeaweb.org/articles?id=...

This is a great idea. Gets the individual incentives ($) lined up with the collective incentives (good scientific knowledge) in a viable and sustainable way Funders, why not?

"We need a system where producing bad work has real, predictable costs, even for metrics-based reputation, where finding errors in others’ work is rewarded rather than punished." 👏

Hell yeah. Sadly people are afraid of trying new things. Especially in academia.

I gave a TED talk today!

New post on The 100% CI: Science needs downvotes. www.the100.ci/2026/04/13/s... In which I make the case that grant funders should add funding lines that include a module for bug bounties.

New blog posts, new mistakes, new (to me) kinds of error! Recently, @schmukle.bsky.social was confused about a plot in one of my papers. This usually means I messed up bad. After a couple-month-long investigation, I can now report: I messed up, but fortunately it was fairly inconsequential.

Survey research is often interpreted as showing that belief in conspiracy theories can be surprisingly widespread, including belief in conspiracy theories that would be astonishing if true. For example, in The Atlantic we learn that “12 million Americans believe lizard people run our country”

Direction of effect and success of publcation: a randomized study has really been done already! Thanks to @floriannaudet.bsky.social for pointing it out jamanetwork.com/journals/jam...

I would have preferred to have the "draw the rest of the owl" meme on the cover, but this will do. Seems like it is on schedule, and we'll leave some typos so you know we didn't write it with AI.

The Iowa Gambling Task is an extreme example of Jingle Fallacy and schmeasurement. In 100 articles we found 244 different ways of scoring it, 177 were never reused. Correlations between them range -.99 to .99. At the same time, we show meta-analyses combine these results as if they’re equivalent.