If you currently do not have a reproducible workflow where you can share data and code where possible, I expect you will soon not be able to publish in good journals. Beyond being best practice, journals will use this to identify papers written by AI. Plan a new project accordingly.

Now published in Collabra! online.ucpress.edu/collabra/art... Using test data from TwinLife (N=9,980, Culture Fair Test CFT 20-R), we show that Regularised Prediction and Poststratification yields IQ scores that differ by up to 19 IQ points (mean 4.19 pts) from those reported in the CFT 20-R manual

New post from me on The 100% CI. Think inside the box, part 1: The guy Euclidean distance told you not to worry about www.the100.ci/2026/03/09/t...

We're thrilled to open registration for the Utrecht Replication Games. The event will be at the at the University of Utrecht on June 4th. Psych, public health, pol sci and econ studies will be reproduced! Register here: www.surveymonkey.ca/r/Replicatio...

Another day, another stupid Excel chart.

R Wars: Return of the base-r.

Conceptual and methodological flaws undermine claims of a link between the gut microbiome and autism Kevin J. Mitchell Darren L. Dahly, Dorothy V.M. Bishop

Two new papers from the lab on research practices in security & privacy research: 1) Reliability of measures and the use of Cronbach's α osf.io/preprints/ps..., 2) Practices around retraction and correction notices by ACM and IEEE (osf.io/preprints/ps...)

New post! It may seem ambitious to ask for individual patient data from clinical trials to be shared, anonymized, for use by other researchers. But the history of medicine shows us that clinical trials have already undergone a series of transformations that once seemed equally bold:

Now that the #rstats pipe wars have flared up again, it's useful to remember what the pipe (base or magrittr) was designed for, and what it was never meant to do. 🧵 1 / 7

“Training a human takes 20 years of food”

So, what's the next #rstats flame war on the schedule? Base |> vs. magrittr %>%?

Oh bla di Oh blavaan Life goes onnnnn, oh!

There’s something sad about how many blogposts are simply copy-pasted ChatGPT output now. It’s not writing, it’s content — not voices, but outputs. When the only button you press is for an LLM, blogposts don’t earn applause, they earn scrutiny.

In standard scientific reports in psych 96% of first mentioned main hypotheses are supported. In Registered Reports, this is 46%. The 96% is clearly biased (given true H1 rate and power). Lack of transparency means we do not know the true baserate. journals.sagepub.com/doi/10.1177/...

New blog post about the age-period-cohort identification problem! In which, for the first time ever, I ask "What's the mechanism?" and also suggest that sometimes you may actually *not* be interested in causal inference. www.the100.ci/2026/02/13/o...

From this follow simple recommendations: as a default, meta-scientific studies of published research artefacts need to include 1) a full, identifiable list of included studies, 2) the full coding instrument and decision rules, and 3) the individual ratings together with a codebook.

Meta-scientists cannot demand transparency from scientists while failing to meet those same standards. In this paper we make one simple recommendation: meta-scientific data should be deanonymised by default.

New paper, on a worrying trend in meta-science: the practice of anonymising datasets on, e.g., published articles. We argue that this is at odds with norms established in research synthesis, explore arguments for anonymisation, provide counterpoints, and demonstrate implications and epistemic costs.

Should meta-science articles hold themselves to the same standards of transparency and reproducibility that the field prescribes for others? We think so. New preprint and 🧵

me, today

LLMs are very good at extracting information from academic articles. They are much better than even highly trained humans (our grad RAs had hundreds of hours of practice). And of course they're ~1000x cheaper and faster

I think some of these points miss the mark. Moreso, I'm concerned that they'll be used by others to smuggle hidden moderators debates back into psychology. 🧵 www.bps.org.uk/psychologist...

What do asexual women want? Paula Bange (research.tilburguniversity.edu/en/persons/p...) first asked this question over four years ago in her Master’s thesis. Now our results are published in Archives of Sexual Behavior: link.springer.com/article/10.1...

The state of mental health care in Germany has been a huge discussion issue in my circles and these findings are fascinating but also concerning -- expanded capacity does not automatically reach those most in need.

🧵 New WP with @kaimiele.bsky.social. We document 2 unsettling patterns in mental health care: 1) despite universal coverage, few individuals with mental illness receive guideline-consistent treatment, and 2) the more severe the illness, the lower the treatment uptake & the longer the wait times. 1/n

Not sure how we are supposed to trust Ariely’s word on this unless he signed the statement at the top.

New preprint! So, what's a multiverse analysis good for anyway?> With @jessicahullman.bsky.social and @statmodeling.bsky.social juliarohrer.com/wp-content/u...

dplyr 1.2.0 is out now and we are SO excited! - `filter_out()` for dropping rows - `recode_values()`, `replace_values()`, and `replace_when()` that join `case_when()` as a complete family of recoding/replacing tools These are huge quality of life wins for #rstats! tidyverse.org/blog/2026/02...

Another reminder of why the Methods section matters and should be written in as much detail as possible, and why meta-analyses need to account for variations in procedures.

8.0% of citations in the medical literature contain major errors. E.g., the cited work makes the opposite claim or is unrelated to the claim in the citing article.

New paper from the lab! We explore the concept of red vs blue teaming to develop techniques to defend against de-anonmiysing attempts of a fairly typical Qualtrics dataset arxiv.org/abs/2601.19575

Wow this scoring chaos seems to be an extreme case of what I say about many ad hoc analyses: no derivation of method from a clear scientific theory, no assessment of statistical properties, and decades pass before someone notices. This happens in biology too, so let’s not pick on psychology only

"Honest, dedicated, skilled researchers may investigate a topic and come to opposite conclusions because of variation between how they conduct their analysis" My sense is almost all experienced researchers are intuitively aware of this, but nevertheless we tend to ignore this inconvenient truth!

Read my latest post for reflections on reproducibility, research quality and a summary of a great new study which shows how NOT to do it https://open.substack.com/pub/tomstafford/p/gambling-with-research-quality

"you could - by careful choice of an existing scoring method from the literature - find any effect, or nothing, or the reverse of any effect you choose. This is bonkers. ... so extreme as to be farcical." This is the sentiment we were hoping people would come away with! w/@anniria.bsky.social

📣Revised preprint by @cas-goos.bsky.social Measurement reliability, validity, and reporting in psychology still has a long way to go... We compared original studies w replications, and where possible recalculated reliability & unidimensionality. Some findings > osf.io/preprints/ps...

Scrutiny web app for GRIM/GRIMMER: errors.shinyapps.io/scrutiny/ GitHub docs for R package: lhdjung.github.io/scrutiny/

Great talk from @lhdjung.bsky.social at the SwissRN Research-On-Research day on his, our, and other folks' work to develop trustworthiness assessment tools. Finally his {scrutiny} R package has a logo!

Awesome to see more journals promote replication studies. They are an essential but too rare part of scientific discovery. If an effect can't be intersubjectively established, we do not understand it well enough to consider it scientific knowledge.

It will still take decades before social sciences understand with what Wittgenstein tried teach them 70 years ago bsky.app/profile/mkar...

First post of the year, new paper out today: we present possibly the biggest case of systematic Measurement Schmeasurement in tech use. It seems that most studies on gaming (videogame) addiction/disorder haven't measured gaming after all. This research took years, so long 🧵 doi.org/10.1098/rsos...

I’ve been thinking about how to find R packages by functionality when you don’t already know the package name. So over the holidays, Claude Code and I built The Warehouse: a functionality-first R package directory that helps you find packages by what they do. rwarehouse.netlify.app #rstats

Self-report scales in psychology get a lot of flak & there’s a desire for more objective, “scientific” behavioral measures. Now if you look at some of these more closely, there are hundreds of way to score them to generate some outcome variable — and many scorings are used exactly *once.*>

Our preprint has evolved! v2 of “Digital Behaviourism” is out now with a new title, new co-authors, and a deeper dive into the behavioural concepts that shape our online lives. It’s time to move beyond “screen time” and focus on function of online behaviours. osf.io/preprints/ps...

We looked at 100 IGT studies and found abysmal reporting standards, and 244 different scoring techniques. Applied to a new dataset, the average correlation between these 244 techniques was 0. 5/15 IGT meta-analyses we looked at did not account for this hetereogenity.

@davidtingley.bsky.social obsessed over this issue w.r.t “replay” for like half of our PhD, which turned into a great review with @apeyrache.bsky.social

Quickly telling someone you’ve liked something they’ve done (e.g. telling a writer you like their writing, or a researcher you like their research, or a teacher you like their teaching) has a phenomenally high effort-to-impact ratio.