Judith ter Schure

If everything is a registered report, problem solved. Priors need review.

Simulating many scenarios is good practice, whether you're analysis is frequentist or Bayesian. Still really hard to adjust a-priori in exactly the way optimal for actual data.

That's an interesting case: whether these workflows are distinguishable from fabrication or even can be classified as informative priors. Do they write about that?

The general 'not dinstinguishable' claim of course only needs one counter example to falsify 😉.

If you define you prior before results are known, and it is evaluated by reviewers, it cannot be expected to do what post-hoc data fabrication does for your analysis.

Use researchers's attitude towards error correction as a criterium in promotion and grant decisions:

Two (paid!) science sleuths in every promotion and grant committee to sift trough CVs for errors/sloppy work

Interested in correcting their own if they get cought and cost them a promotion or grant

Yes please let's normalize errors, their discovery, their correction, and their integration into the literature. "Corrections Not Found: Post-Publication Integrity in Usable Security and Privacy Research" (osf.io/preprints/ps...)

From just the title

'Bayesian estimation with informative priors is indistinguishable from data falsification':

Would a registered report not be a counter example?

My Matters Arising concerning a paper on the legal determinants of terrorism is now out in @nathumbehav.nature.com. The original paper is now retracted. To learn why, read on for a story of irregularities, imputations, and impossible values. 1/x

doi.org/10.1038/s41562-025-02347-7

My effort to reproduce this paper began as part of the Institute for Replication’s ongoing project to systematically examine the reproducibility and robustness of papers in Nature Human Behaviour3; my participation in this endeavour was approved by the Ethical Review Board of Vrije Universiteit Amsterdam’s School of Business and Economics. Inspecting the paper’s first two figures revealed a mathematical impossibility. There are nine EU countries that experienced zero terror attacks during the study’s time frame. However, the paper reports that the inverse hyperbolic sine of these countries’ per capita attack rates are positive, and increase or decrease over time. This is impossible; the inverse hyperbolic sine of zero is zero4. The main outcome variable displayed in the paper’s second figure is hard-coded in the replication data as ‘DVSin’. Figure 1’s top row of plots shows that DVSin is negatively correlated with both terrorist attack rates (r = −0.107, two-sided P = 0.024) and their inverse hyperbolic sine (r = −0.108, two-sided P = 0.022). These plots also show that in the 305/420 country-year observations after 2006 experiencing zero terror attacks (72.6%), DVSin takes on 292 different positive values. This implies that the paper’s main outcome variable cannot possibly be constructed as described in the paper.

"the paper’s main outcome variable cannot possibly be constructed as described in the paper."

Retraction of 2023 paper that did not use the reported variables. The replication report is astonishing www.nature.com/articles/s41...

Pre-publication peer review remains undefeated in laundering bullshit

The life's work of anyone who gives a damn about humanity is to resist the economic elite, demand the taxes that reduce their power and defend the space in which the rest of us can thrive. If your "elected representatives" aren't helping with that, they are not your representatives, but theirs.

This year has been brutal for the #climate movement.

Rollbacks, repression, & outright denial, especially in the US, have made progress feel fragile & uneven.

Which is exactly why it matters to pause & look back at what did move forward in 2025.

Here are 12 real wins 🏆, one for each month 🧵⬇️

Little blogpost reporting an analysis of recent (2021-2025) retractions of highly-cited papers in relation to #PubPeer comments. deevybee.blogspot.com/2026/02/the-...
#retractions #publishing

Jonge Syrische asielzoekers leerden overleven op straat en zorgen er nu voor overlast Alleenstaande Syrische vluchtelingen veroorzaken steeds vaker overlast. Deze jongeren groeiden niet op in Syrië, maar aan de rand van Europa, waar ze leerden overleven op straat. De Volkskrant ging op...

Meer van dit soort reportages graag, over de complexiteit van straatoverlast. Door @rikkuiper.bsky.social en @lisa-1804.bsky.social. www.volkskrant.nl/binnenland/j...

I also noticed that completely rediculious papers are cited by authors working on microplastics, so they seem quite sloppy in that regard as well.

bsky.app/profile/judi...

You can sign up until January 23 for the free Paul Meehl Graduate School workshop on Scientific Criticism and Peer review taught by René Bekkers. It promises to be an extremely interesting day, so do join us in Eindhoven on January 30th! paulmeehlschool.github.io/workshops/cr...

We should really move to reporting in journals that focus on reproducibility instead of novelty.

Given that NEJM has not implemented any procedures like Registered Reports, it does not deserve our admiration.

Thanks for sharing that example! Love it!

I wrote a paper about cumulative science being biased by design (doi.org/10.12688/f10...) and ended it with this Bayesian quote:

Conclusion so far:

Statisticians probably don't complain if conventional CI is reported, even though they think it shouldn't when questioned.

Nobody is actually saying 'yes', or 'no, always report this instead'.

Curious where you stand on this!

Similar: other arms could have stopped as well but didn't. Analyze taking into account the hypothetical stopping in the sampling process?

Only answer consistent with theory: yes. But hypothetical stopping rule is not defined and could be anything.

Current solution: Go Bayes or go anytime-valid 2/2

Really difficult to judge. I don't like inconsistency between the theory of frequentist confidence intervals and their use, but do know that complete consistency can get crazy quickly.

Currently looking into a four-arm trial with one arm stopped unplanned for harm on the primary outcome. 1/2

Incredible work by these authors to continue to write to the BMJ to get this paper retracted!

I am quite in favor of the opposite: let everything be considered an interim analysis. Do as many as practically possible. Base any decision to randomize the next patient on all available data. And over all trials, less patients need to be enrolled.

The AV interval is the same whether you implement the futility interim or not: Bayes like

Because valid under any stopping rule, they don't correct bias; other stopping rules bias in the opposite direction.

Disliked by frequentists that do not want to correct for stopping that did not take place

Ik ben wel voor het afschaffen van collegegeld voor opleidingen in de zorg met tekorten. Waarbij je dan mag terugbetalen als je niet in die tekortberoepen gaat werken, en vooral terug mag betalen als je gaat ZZP'en als kwakzalver.

What would you report for the primary estimate in such a trial?

Bootstrap interval under the futility analysis sampling process?

The Bayesians have a point that this is also stupid.

If I stay frequentist, I prefer anytime-valid confidence intervals but they are disliked because they are wide.

Really curious what eveyone's take is on this. Please tag people that have an opinion.

@kertviele.bsky.social @statberry.bsky.social @statsepi.bsky.social @stephensenn.bsky.social @numbersman77.bsky.social @erik-van-zwet.bsky.social

Trial statisticians!

Should RCT pubs report conventional 95%-confidence intervals if they did a futility interim and continued the trial?

No sampling distribution can possibly be Gaussian if randomly better than average data reaches study completion while randomly worse than average doesn't.

My New Year's resolution exactly!

Julia has some great advice here:
bsky.app/profile/ding...

And I want to second @tsawallis.bsky.social's reminder of Diamond Open Access Journal of Robustness Reports (scipost.org/JRobustRep).

This one is also insane. More than 700 citations!

bsky.app/profile/judi...

A hope for 2026 is that this perspective piece with @wiringthebrain.bsky.social & @deevybee.bsky.social will serve as a template for others who are similarly frustrated with with exaggerated claims and double speak around so much of research. It's ok to point out that the emperor has no clothes!