Happy to see this in print!

doi 10.1146/annurev-statistics-042324-123749
@maartenvsmeden.bsky.social @laurewynants.bsky.social @vanamsterdam.bsky.social and Ewout Steyerberg

Boinggg!

Doug Altman was an internationally renowned statistician who served as The BMJ’s chief statistical adviser.

Read about life and work that made this statistician a "citation millionaire"
#BMJChristmas
www.bmj.com/content/391/...

I can follow that. Therefore, we state that the paper does not talk about model comparison (including the addition of a predictor to an existing model).

A STRATOS paper with Ewout Steyerberg, @gscollins.bsky.social @laurewynants.bsky.social @maartenvsmeden.bsky.social @vickersbiostats.bsky.social @kdpsingh.bsky.social @gaelvaroquaux.bsky.social @davemclernon.bsky.social @lasaibarrenada.bsky.social K. Kerr T. Hernandez-Boussard C. Moons D. Timmerman

Evaluation of performance measures in predictive artificial intelligence models to support medical decisions: overview and guidance Numerous measures have been proposed to illustrate the performance of predictive artificial intelligence (AI) models. Selecting appropriate performance measures is essential for predictive AI models i...

Our guidance regarding performance measures for medical AI models is finally out!

- Stop bashing AUROC, although it does not settle things
- Calibration and clinical utility are key
- Show risk distributions
- Classification statistics (e.g. F1) are improper

www.thelancet.com/journals/lan...

Expertise is having fucked up in enough different ways that you become able to anticipate it.

We need to think more about communication in this context, that is for sure. Clinical staff should not say "your risk is 21%". Just giving an 95% CI in layman terms does not do the job. The uncertainty complicates things quite a lot imo, not sure how to best tackle that to be honest.

The fundamental problem of risk prediction for individuals: health AI, uncertainty, and personalized medicine Background: Clinical prediction models for a health condition are commonly evaluated regarding performance for a population, although decisions are made for individuals. The classic view relates uncer...

In our latest work, we show that risk estimates for patients are HUGELY uncertain due to model, data, and population uncertainty. Even for well performing models (c statistic, calibration, utility) based on large N.
@laure_wynants @ESteyerberg @lasaibarrenada.bsky.social

arxiv.org/abs/2506.17141

Universities love open science
Small print: unless money is involved

Bar chart showing % of articles retracted, with expression of concern, or no action from different publishers. 100% no action from Elsevier and Wolters Kluwer; 100% retraction from Taylor and Francis.

Huge variability documented in how publishers respond when informed about a problematic body of work by a research group. www.jclinepi.com/article/S089...
#publishers #retractions

We recently had 5 review reports for a submission. While the comments were reasonable, the 5 reports were all very similar in the issues they raised. Suspicious.

Multiple Imputation of Missing Covariates When Using the Fine–Gray Model The Fine–Gray model for the subdistribution hazard is commonly used for estimating associations between covariates and competing risks outcomes. When there are missing values in the covariates includ...

Multiple Imputation of Missing Covariates When Using the Fine–Gray Model. Edouard F. Bonneville, Jan Beyersmann, Ruth H. Keogh, Jonathan W. Bartlett, Tim P. Morris, Nicola Polverelli, Liesbeth C. de Wreede, Hein Putter. Statistics in Medicine. onlinelibrary.wiley.com/doi/10.1002/...

You could park many of them in the Journal of Unplanned Interim Analyses.

Seconded.

Any time someone uses this term, make sure they explain exactly what they mean. If they can't, they are obviously trying to bullshit you.

Quality of scientific papers questioned as academics ‘overwhelmed’ by the millions published Widespread mockery of AI-generated rat with giant penis in one paper brings problem to public attention

“Volume is a bad driver,” [Sir Mark Walport] said. “The incentive should be quality, not quantity. It’s about re-engineering the system in a way that encourages good research from beginning to end.”

www.theguardian.com/science/2025...

Graeme is clearly on to something here!

What is common knowledge in your field, but shocks outsiders?

Data isn't objective and researchers have innumerable ways to put their thumbs on the scale. Many don't understand statistics well enough to realize they're doing it.

Yes! Some years ago, The Group of Biomedical Sciences at mu uni said in front of an auditorium with many young and local researchers that they wanted to invest more in superstar transfers and compared it with football transfers.

Importance of sample size on the quality and utility of AI-based prediction models for healthcare Rigorous study design and analytical standards are required to generate reliable findings in healthcare from artificial intelligence (AI) research. On…

**New Lancet DH paper**

"Importance of sample size on the quality & utility of AI-based prediction models for healthcare"

- for broad audience
- explains why inadequate SS harms #AI model training, evaluation & performance
- pushback to claims SS irrelevant to AI research

👇
tinyurl.com/yrje52fn

Yes.... somehow they always get it wrong. You cannot get to those levels in science and politics and get it right.

yes, subscription model. Fee calculated based on character count of the main paper, including refs, tables, figs. Sounds a bit odd in this digital age.

I recently had a paper in a journal where the fee for open access was lower than the fee for closed access. Is that common? @grahamkendall.bsky.social

Why aren’t you wearing a suit?

Have you said “Thank you” to Greenland and Denmark for allowing you to have a base on Greenland?

Great question! It seems that universities would (secretly?) say yes.

Oh, there's an English version as well! DW reports about misconduct at Max Planck Institutes: www.youtube.com/watch?v=n5nE...

PROBAST+AI: an updated quality, risk of bias, and applicability assessment tool for prediction models using regression or artificial intelligence methods The Prediction model Risk Of Bias ASsessment Tool (PROBAST) is used to assess the quality, risk of bias, and applicability of prediction models or algorithms and of prediction model/algorithm studies....

PROBAST+AI is out!

www.bmj.com/content/388/...

I will be presenting our recent work on individual risk estimation uncertainty at ENAR in New Orleans. Come say hi!

Work with @benvancalster.bsky.social @laurewynants.bsky.social #DoranneThomassen #EwoutSteyerberg

Academia says 'no novelty no glory no funding'

Following the footsteps of Bink Marino!