🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

An entire field built on nothing. We should always be careful with hype in science and stick to solid data instead.

The UK Biobank (UKB) has been an amazing field changing resource for the past decade. Unfortunately, and very suddenly, they have decided that to work with their data you now *must* use their “Research Analysis Platform” (RAP). UK Biobank I guess there have been general warning

Just a reminder that if you can not specify which effects are too small to matter, and 'any effect matters', no effect can falsify your predictions. And if nothing can falsify your statistical predictions, Sir Karl Popper would have thought you are not doing science.

If you're not interested in the null hypothesis of zero effect then you shouldn't be testing it. But then you have to specify a minimum effect size of interest. If you can't do that maybe you should be testing the null.

Challenge accepted. Here are a few comments on the paper after starting to wade through its massive content. The paper in question is nature.com/articles/s4158… 1/🧵

Big News! On behalf of the ASA, Taylor & Francis are launching a new journal: 𝗦𝘁𝗮𝘁𝗶𝘀𝘁𝗶𝗰𝘀 𝗮𝗻𝗱 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗶𝗻 𝗜𝗺𝗮𝗴𝗶𝗻𝗴 The journal goes live for submissions later this month, but take a look at the cover! T&F STEM tandfonline.com/journals/usdi20

We have updated NextBrain with: - new atlas with improved brainstem. - updated segmentation code. - new labeled ex vivo dataset. - new preprint. Please see the project website for links and resources: github-pages.ucl.ac.uk/NextBrain

Just submitted this review. Is my request reasonable? The paper consists mostly of hierarchical regression analysis, with a surprisingly large coefficient for one IV. (There were several other questions, all of which I answered in the same way as points 2 and 3 here.)

1/ Can Large Language Models (LLMs) truly reason? Or are they just sophisticated pattern matchers? In our latest preprint, we explore this key question through a large-scale study of both open-source like Llama, Phi, Gemma, and Mistral and leading closed models, including the

The real work of science is generating a well-posed hypothesis that can be tested, not predicting the results of a study to test that hypothesis.

It's an absolute scandal and tragedy that India is forced to spend $250m/year to provide access to the latest science to its people. A complete and total failure of global science leadership and of science writ large.

We are witnessing a very sad moment in American history. The President of the United States is aligning himself with the dictator of Russia, Vladimir Putin, to undermine the independence of Ukraine and its democracy. Here are my thoughts:

Whenever you welcome guests, treat them with kindness, grace and respect. Simple as that.

If ICE can legally deport students for expressing the wrong political views, we just don't have academic freedom at US universities, at least not for int'l students. I'm not a lawyer and I sincerely hope this law gets struck down on 1A grounds, but it looks very bleak.

⚡️Conformal prediction for uncertainty quantification with #rstats⚡️ Tutorial by my PhD student Jesus Uriel Diaz Martinez on how to obtain conformal intervals in regression problems 🔷 Data splits {rsample} 🔷Linear & GAM {parnsip} 🔷Conformal intervals {probably} 👉 uriel278.github.io/tutorials-cp/c…

I wrote about how population stratification in genetic analyses led to a decade of false findings and almost certainly continues to bias emerging results. But we are starting to have statistical tools to sniff it out. A 🧵:

a nice result, but the title should have been "large language models pass a five-minute turing test". arxiv.org/abs/2503.23674

The university will not surrender its independence or relinquish its constitutional rights. Neither Harvard nor any other private university can allow itself to be taken over by the federal government. hrvd.me/ResearchFundin…

In this thread I'll record some brief impressions from trying to use o3/o4-mini (the new OpenAI models) for mathematical tasks.