Feeling the urge to freshen things up a bit around the ol’ blog. I’m starting small: I’ve decided to shamelessly steal Matt Levine’s schtick when it comes to linkfest post titles. 🙂 You’ll have to read all the linkfest items to find out which one the title refers to. 🙂
This week: the illusion illusion, against statistical amateurs, against the scientific method, against loss aversion, weather forecasting breakthrough, Recursion Fairy, and more.
Omitted variable bias in the context of estimating nonlinear interaction terms. Typically crisp, clear blog post from Data Colada. I was interested in this in part because it seems like ecologists these days are increasingly into estimating interaction terms, including nonlinear interactions. Over the holiday, I’m planning to do a little side project, compiling some data to check whether ecologists are in fact more into estimating interaction terms than they used to be.
Is the solution to scientific fraud, and questionable research practices, to expect (or somehow force?) all researchers to collaborate with professional statisticians? I don’t really buy it, and I kind of feel like the linked review doesn’t either?
Writing in Science, Vassiliki Betty Smocovitis reviews Brenda Wineapple’s new book on the perfect storm of circumstances that made the Scopes trial such a sensation.
Are people really risk averse and loss averse? Or do they merely struggle to think through complex choices? Link goes to a blog post discussing some new experimental results that seem to have made quite a splash.
Adam Mastroianni stumps for Feyerabend’s anarchist philosophy of science. Well, mostly. I’m still mulling over whether the “hockey stick” graph (no, not that hockey stick graph) is as telling against Feyerabend as Mastroianni thinks it is. Self-recommending.
Speaking of philosophy of science: here’s Max Dresow writing accessibly and interestingly on the best philosopher of science you’ve never heard of. Well, I’d never heard of him; I’m only guessing you haven’t either.
Astral Codex Ten reviews Tom Wolfe’s From Bauhaus to Our House. One of these years I need to get around to reading that book (and I say that in full awareness that the book was widely panned at the time, and that Tom Wolfe’s opinions on Darwinian evolution were rubbish). Polemics can be fun to read, even if they (inevitably?) get some things wrong. Anyway, I link to this mainly because I’ve linked to other recent pieces on this topic. Note that the linked review leans into analogies between the events in Wolfe’s book, and more recent events in areas outside architecture. Your mileage may vary as to whether those analogies please you, or annoy you, or what. I am keeping my own reactions to myself.
In a recent linkfest, I accidentally goaded Stephen Heard into explaining the difference between good and bad writing advice.
LLMs don’t recognize optical illusions. This is funny, but also serious. It’s a generic recipe for confusing LLMs (and people who aren’t looking or reading carefully). Show them a picture, or chunk of text, that similar but not identical to a familiar picture or chunk of text.
Here’s why certain people’s names break ChatGPT.
Big breakthrough in using machine learning to predict the weather, and storm tracks. Apparently, the probabilistic forecasts not only are substantially longer-term and more accurate than the best ones currently available, they’re also faster to produce and computationally cheaper to produce. The underlying code is going to be shared online.
The (No!! Stop!!) Recursion Fairy. 🙂 I assume Stephen Heard will be adding this to his collection of examples of humor in scientific writing.
Status competition. 🙂 😦
How squids evolved to shoot ink. 🙂