Uffe Heide-Jørgensen

Hallelujah - my personal take on "real world evidence/data" (as I see it used) is that it is a euphemism for data missing not at random

This makes me think of Terry Pratchett's Hogfather

"Normalized incompetence" seems to sum up my day-to-day!

Putting myself out there a bit. Honest request from an open-source developer: Can I please ask you to encourage people you know to stop with the "any updates on this?" discussion post? Every post like that genuinely makes me die inside a little bit. I'll try to explain in this thread. 1/

We should introduce the term “real human researchers” (RHR) to go with other nonsense phrases like “real world evidence”

today's useful tip

yep - often mentioned in the stats group I work in. We don't get it, and seeing it has been a long time we have not got it, I start to think it is simply because we are not very smart when everyone else apparently gets it

I’ve seen situations in which people try to correct the scientific record and then critics act like that was superfluous or just not the right way to use one’s energy because “science is self-correcting.” This just came to my mind

Happy reinstalling-all-your-R-packages day to all those who celebrate #rstats

#uhoh Looks like OpenAI can take surveys for you… how are we going to stop this?? It’s currently only for “Pro Users” ($200/month) but eventually roll out. Big threat to survey research

"why don't you leave academia then?"

If one wanted to hire a professional #rstats #rshiny developer to work work with qualitative/text data in particular (so many involve a dashboard for search/semantic/sentiment analyses), do you have any recommendations of individuals or companies? (please repost)

Ridiculous. The total is more than 400%. Everyone knows it should be 360% for a circle.

🚨 **Challenging the Status Quo in Machine Learning: Are Class Imbalance Corrections Doing More Harm Than Good?** 🚨 In the world of machine learning, handling class imbalance has become almost second nature.

What is meant by a 'while on treatment' estimand? thestatsgeek.com/2025/03/03/w...

Scientific publishing cannot be free but an annual $2.5 billion seems excessive to me

Spread some positivity today ✨ do something that somebody with your combination of covariates is extremely unlikely to do.

To be honest I think the control group is so poorly defined here that I put next to no value into the results. Then again I am biased against Nature as a journal going more for wow than quality. No serious journal puts the methods section at the end of a manuscript

On this day in 1862 in the city of Königsberg (now Kalinigrad, Russia) was born David Hilbert, most important mathematician of his generation. Panel below from Logicomix (www.logicomix.com), which remains one of my favorite books. Team Hilbert forever.

great post - the final twist reminds me of this classic xkcd (xkcd.com/552/)

Cross-validation identifies disasters but doesn't rescue you from them.

If only more researchers realized that sometimes the right thing to do is to abandon an infeasible study, we could do things that are feasible, instead of wasting our (and reviewers'/readers') time trying to make sense of the senseless

New blogpost: deevybee.blogspot.com/2024/11/why-...

have not had time to appreciate this in detail, but I love the colour scheme

While we’re all putting #rstats opinions out there: base plot is better than ggplot because it’s default is so rubbish it forces people to think about design, whereas ggplot default is just about OK enough (but still far from good) to make people not bother.

To new users and followers: Here's a Shiny app showing the differences between three ways of including a continuous covariate in linear regression models. Feel free to use it in your teaching or elsewhere! #rstats #datascience #statssky #stats #epi #episky