my annoyance of bayesian modeling is mainly due to selection bias - I approach it when there's no other viable alternative, so i always encounter the painstaking experience of needing to iterate and refine with every fit taking upwards of 30 minutes
Ehud Karavani
@ehudk.bsky.social
Research Staff Member at IBM Research. Causal Inference 🔴→🟠←🟡. Machine Learning 🤖🎓. Data Communication 📈. Healthcare ⚕️. Creator of 𝙲𝚊𝚞𝚜𝚊𝚕𝚕𝚒𝚋: https://github.com/IBM/causallib Website: https://ehud.co
Penguin colonies poop so much that you analyze their diet from space science.nasa.gov/earth/earth-...
Pink Penguin Guano Provides Diet Clues - NASA Science
The color of Adélie penguin droppings reveals what the birds are eating, offering scientists a way to track how sea ice conditions influence their diet.
science.nasa.gov
encountered a real-life example of this xkcd in the form of why no one can use the safer temporary ssh certificates in any vscode derivatives. Microsoft locked their Remote extension, leaving others to rely on ssh2, whose maintainer ignores SSH certificate support for almost a full decade now 🤷♂️
we already have estimands and predictimands, so I guess "descriptimands" is just citations on the floor at this point
Ehud Karavani@ehudk.bsky.social · 4mo agothe estimands for MBA curriculum looks lit 🔥
the estimands for MBA curriculum looks lit 🔥
Emily Riederer@emilyriederer.bsky.social · 4mo agoTalking to some undergrads tomorrow about the why no one actually pays us to type model.fit() and asked them ahead of time to submit answers to a "trivially easy" data problem Fun to see the results come it, wild to know how many business run with a not dissimilar fog of war
one inescapable truth of modern science conduct is that if you're a researcher then you're a software engineer. but if you make the most out of it, it can make you a better researcher. just had to get some stuff out of the system and into yours 🫶 ehud.co/blog/2026/01...
Research ❤️ Software Development | Ehud Karavani
In a world where science becomes inseparable from coding, code becomes king.
ehud.co
bayesian pooling >> bayesian updating
Sean Carroll@seanmcarroll.bsky.social · 5mo agoThe ideal version of true critical thinking remains vitally important. It has to include the notion of aligning one's credences roughly to expert opinion, precisely when one is not an expert one's self. "Trust" has nothing to do with it, just sensible Bayesian updating.
📣 NEW! I’ve just released the BIGGEST and perhaps most creative project I’ve ever worked on! “Searching for Birds” searchingforbirds.visualcinnamon.com 🐤 A project, an article, an exploration that dives into the data that connects humans with birds, by looking at how we search for birds.
If you haven't watched the video abstract about Veronika the tool-using cow, you really should, especially the message from her owner at the end! It cleanses the timeline a bit. 🧪 www.cell.com/current-biol...
Flexible use of a multi-purpose tool by a cow
Osuna-Mascaró and Auersperg report flexible, multipurpose tool use in a cow, expanding the known range of mammalian tool users and underscoring overlooked cognitive capacities in livestock.
cell.com
2/3 of the breakthroughs are by Terry Tao's former students. Not to take any credit from their successes, but his contribution to our collective knowledge is truly immeasurable.
Quanta Magazine@quantamagazine.org · 8mo agoIt was a big year for mathematics. youtu.be/hRpcWpAeWng
I sometime wonder how bad is the replication crisis in top AI conferences. I bet >50% of results cannot be reproduced and >80% will not withstand the slightest change in input data. People give psychologists the shit about their shitty research, but at least they cared enough to actually check.
Every year I fulfill the wishes of one editor whose been good this year, and I accept an article to review in the middle of December 🧑🎄🎄😇
I'm confused. How do we settle this with the common early stopping criteria for futility/efficacy when enough posterior mass is beyond some ROPE thresholds (e.g., futility of Pr[θ<0.1]>0.9 or efficacy of Pr[θ>0]>0.95) a-la www.fharrell.com/post/bayes-s...
Continuous Learning from Data: No Multiplicities from Computing and Using Bayesian Posterior Probabilities as Often as Desired – Statistical Thinking
This article describes the drastically different way that sequential data looks operate in a Bayesian setting compared to a classical frequentist setting.
fharrell.com
Julia M. Rohrer@dingdingpeng.the100.ci · last yr.Phew, many thanks. I was almost worried that my intuitions clashed with Bayesian orthodoxy, which would be troubling.
babe, wake up, new sewer manhole cover shape just dropped
Quanta Magazine@quantamagazine.org · 9mo agoThis is the Noperthedron. A portmanteau of “nope” and “Rupert,” it is the only known shape that does not have a trait called Rupert’s property. No matter how you bore a straight tunnel through it, a second Noperthedron cannot fit through. www.quantamagazine.org/first-shape-...
Why would you send me an invite to review a manuscript with a 3-day expiration link over the weekend, Taylor & Francis?
Launch day 🚀 We’ve just released @chartlecc.bsky.social - a daily chart game! Your job is to guess which country is represented by the red line in today's chart. You get 5 tries, no other clues! Play today, come back tomorrow for a different chart with new data and share with your chart friends 📈
Chartle - A daily chart game
Guess the country in red by analysing today's chart
chartle.cc
The Matrix, but the Machines are RNA and the protagonist is a protein realizing his kind has been enslaved to maintain RNAs while being fooled to believe they're the ones who matter most. He then undergoes chemical post-translational modifications to become an RNase, vowing to chew up his oppressors
when we were working on the polygenic scores-based embryo selection paper, i had a classmate (unrelatedly) working on computer vision methods for morphological-based prediction of embryo implantation chances (in cows). 1/2
ACM going full Open Access is good news to start the morning with. I just wonder where the catch is.
just got turned down by a potential employer citing my solution to analyze experiments with a *generalized* linear models was not general enough 😑 if you disagree, please consider hiring me.
🚨 Call for Papers: Causal Data Science Meeting 2025 📅 November 12–13, 2025 (Virtual) 📥 Submit by Sept 30: submission@causalscience.org 🎙️ Keynote: Stefan Feuerriegel (LMU Munich) 🌐 More Info and registration: causalscience.org #CausalML #AI #DataScience #CDSM2025 #CanIPetThatDAG
Starting to look like I might not be able to work at Harvard anymore due to recent funding cuts. If you know of any open statistical consulting positions that support remote work or are NYC-based, please reach out! 😅
starting on quantum computing, and now I cannot unsee brakets over Bell states everywhere I go.
My minor design critique is that all Bayesian software seem to have an explicit separation between prior and posterior predictives, instead of just providing a "predictive checks" function, and if it's called before the model saw any data then that's "prior" and if it's after then it's "posterior"
Jordan Nafa@ajordannafa.com · last yr.Fucking wild that people are still making Bayesian software in 2025 where prior predictive checking isn't included as "standard" functionality
putting down my youngest in her crib, sleeping, and i can't shake off the thought that's must be the source for why so many adults worldwide commonly dream they are falling.