2/13 We start with a Bayesian framework. Our Prior, P(H0), is the baseline assumption that these images represent the same individual. To move from Prior to Posterior, we need Evidence. In frequentist terms, we are not just looking for resemblance; we are looking for "Identity Invariance." If the probability of these specific ratios appearing by chance in three separate humans is less than 0.000001, the ghost in the machine is real. #Bayes #Math

7/10 🎲 Bayesianische Gewissheit: 94,2 % Übereinstimmung! Wenn wir alles zusammenwerfen, was wir über Gesichter wissen, spuckt Bayes eine Wahrscheinlichkeit von 94,2 % aus, dass Gesicht A gleich Gesicht B ist. Das ist fast so sicher wie das Amen in der Kirche oder die Verspätung der Bahn. Die Zahlen schreien uns förmlich an: Das ist derselbe Thomas! Wer da noch zweifelt, glaubt auch, dass die Mauer noch steht. #Bayes #Wahrscheinlichkeit

BREAKING: Henceforth, I shall be using a new technique, “division”, to determine how much an item costs per count or unit weight.

This advance is a radical and exciting departure from the current method, counting on my fingers and toes.

🤦🏼 🤦🏼 🤦🏼

These f-ing morons.

https://www.instagram.com/reel/DTdoFw7kciA/?igsh=NTc4MTIwNjQ2YQ==

(For those unfamiliar, #Bayes was an 18th c. British #mathematician who developed an equation that has been used for centuries to calculate the probability of an event given what we already know. For example: if you get a medical test with 80% accuracy, and it returns positive, what is the chance you actually have the tested condition? Not 80%, as you may assume. The answer depends on the likelihood of developing that condition in the first place, and then applying the likelihood that the test is correct. Bayes’ theorem is pervasive in #medicine and many other disciplines, and not recently discovered by nut job RFK Jr. and his empty-headed lackeys.)

#math #science #statistics #CDC #HHS

The free open source programming language #Stan helps me to make sense of data and to quantify uncertainty. The next annual Stan conference (in Uppsala, Sweden) is now open for registration and abstract submissions. #bayes https://bayes.club/@mcmc_stan/115701059853138765

When I was in high school, my driver’s ed teacher told us about when **he** learned to drive. He was all over the road. His father was teaching him and immediately saw why: the son was looking at the lines to stay between them, the lines he was touching, the close ones. The father told him "you have to look where you’re going … way down the road!"

I think they were **both** right (and of course, as always, this is not really about driving). You must keep aiming for the end goal, but you will also make constant adjustments (usually small) to get there. Both matter. This is how you solve problems with code. This is also Bayesian thinking.

#Bayes #Bayesian #Programming

#statstab #465 How to embrace variation and accept
uncertainty in linguistic and
psycholinguistic data analysis

Thoughts: An accessible paper on communicating your results with nuance.

#bayes #bayesian #uncertainty #error #bias #guide #tutorial

https://sites.stat.columbia.edu/gelman/research/published/Uncertainty.pdf

Is a tossed coin/die actually random, or is it influenced by the initial physical forces that are imparted to it? Is the outcome an indication of your state at the time of the toss, and is any interpretation of that state valid? What is the probability it is?

#physics #destiny #bayes

Is a tossed coin/die actually random, or is it influenced by the initial physical forces that are imparted to it? Is the outcome an indication of your state at the time of the toss, and is any interpretation of that state valid? What is the probability it is?

#physics #destiny #bayes

🎙️ What does it take to grow in tech? Jordan Thibodeau shares lessons from years inside top tech cultures with Alex Andorra

✅ Bayesian thinking as a practical advantage

✅ AI amplifies skill, not replaces it

✅ Networking & sharing knowledge matter

🎧 https://lnkd.in/ghk6D6nH

#bayes #career

Bayesian deep learning helps ML models understand their uncertainty

In this episode Alex Andorra talks with Maurizio Filippone about Gaussian Processes, scalable inference, MCMC, and Bayesian deep learning at scale

🎧 https://learnbayesstats.com/episode/144-why-is-bayesian-deep-learning-so-powerful-maurizio-filippone

#BayesianStats #AI #ML #Bayes

#statstab #450 Fitting GAMs with brms

Thoughts: Assuming linearity of your continuous predictors is not needed when you can add wiggles!

#gam #glmm #linearmodel #modelling #brms #rstats #bayes #tutorial #splines #r

https://fromthebottomoftheheap.net/2018/04/21/fitting-gams-with-brms/

#statstab #445 What are credible priors and what are skeptical priors?

Thoughts: An excellent thread on prior elicitation by some of the big names in the field (frequentist and bayesian).

#priors #bayesian #bayes #likelihood #medicine #clinical #debate

https://discourse.datamethods.org/t/what-are-credible-priors-and-what-are-skeptical-priors/580

#statstab #441 Bayes-by Shower

Thoughts: Comprehensive (read: long) tutorial on bayesian analysis and how to think about research.

#bayes #rstats #bayesian #tutorial #DAGs #estimand #observational #design

https://betanalpha.github.io/assets/chapters_html/fertility.html

When I was a child, I thought the world had things that were true and things that were false, i.e., things were "black and white".

Things happened to me, including reading "Gödel, Escher, Bach: an Eternal Golden Braid" #Godel #GodelEscherBach, and I realized "Oh! There’s a gray area! (and not only that, the very edges of the gray area are fuzzy!"

And then I learned about #Bayes (and #Laplace) and realized: "Oh shit! It’s **all** gray!"

It feels like you **know** some things to be true because have assigned them such high probabilities. So high, they seem certain. Sorry. It’s not actually 1. And always remember: probability is what you **know**; reality is outside of that (just like "is your blue the same as my blue?"). Yes! Your model is good enough to navigate the world and make good decisions; but absolutely don’t confuse that with having no room left to learn.

I know I said this in a weird way, but keep growing.

🍽️ Can better nutrition science come from better statistics?

In the latest episode, Alex Andorra chats with Christoph Bamberg about using a Bayesian mindset to make psychology & nutrition research more transparent and actionable

🎧 https://learnbayesstats.com/episode/143-transforming-nutrition-science-bayesian-methods-christoph-bamberg

#bayes #nutrition

There’s a strong urge to believe what you wish instead of what you can prove. Computer rumors are a great example. Many rumors have no basis other than being a feature someone wants. They call it "wish casting".

We want the world to be black and white. Some given statement is either true or false. But it’s not. Gödel #Godel describes at least three states: true, false, and unprovable (e.g., the statement "This statement is false". Can’t be true or false; it’s unprovable. Maybe there’s a better name.)

But it’s worse than that.

In science, a theory isn’t true … it’s just the best explanation we have so far. The whole endeavor of science is to keep finding better explanations. To make good decisions you don’t need the absolute best explanation, just one good enough to guide you to beneficial choices. (I said "prove" before, but to be more accurate I should be talking not about what you can prove, but about what you can’t disprove.)

#Bayes (really #Laplace) says a given notion isn’t true, it’s actually true-with-some-probability. Each new thing you observe impacts that #Probability. This is the actual math behind the #ScientificMethod. And it’s the truth of the world. Your beliefs must adapt to your observations, constantly, forever.

If you have unshakable faith in some set of "facts", you’re probably doing it wrong. Even when you’re right, you could be righter.

Of course, if you don’t adjust your beliefs with new input, if you don’t test, if you have "facts" instead of "very probable theories". If you believe things because of how strongly the person who convinced you believed instead of what they could actually show you. If you believe simply because that’s what your parents taught you. Then, well, you **might** be right (even a stopped clock is right twice a day). But at best you’re not going to make good decisions for yourself, and at worst you’re going to try to tell others what to do based on an inaccurate understanding.

It’s messy; and that’s just how it is.

Der Satz von #Bayes ist die mathematische Formel für intelligentes Zweifeln. Er zeigt, wie man Überzeugungen rational aktualisiert.
https://www.youtube.com/watch?v=iElYEW1v0IU
https://homepage.univie.ac.at/franz.embacher/Lehre/aussermathAnw/Bayes.html
https://de.wikipedia.org/wiki/Satz_von_Bayes

Were you aware of a curious information-theoretic property of 89% credibility intervals? Find out in this preprint: <https://doi.org/10.5281/zenodo.17072199> :)

#bayes #bayesian #statistics

Busy week, I have not much posted from #RSS2025. This will follow in the next days.

For now, I enjoyed chatting with Robert Grant and am reading his co-authored book "Bayesian #MetaAnalysis" which he kindly gave me a copy of 🙇
http://www.robertgrantstats.co.uk/bma-book.html

I am always looking combinations of unusual and cross-cutting themes for teaching/ training, so I'll do a parallel read with #MixtureModels in this area: https://link.springer.com/chapter/10.1007/978-3-540-68651-4_7

#SysReview #SystematicReview #Bayes

📚#RichardCarrier has just announced a new book: The Obsolete Paradigm of a Historical Jesus (2025).

As a scholarly sequel to On the #Historicity of #Jesus (2014), it argues the assumption of a historical Jesus is obsolete. Carrier applies #Bayes to case studies (Two Swords, Rom 1:3, Gal 4:4), re-evaluates #Docetism & revises Rank-Raglan odds (≈25% historicity). Looking forward to reading it (will be available end of Nov 2025).

🌍 https://www.richardcarrier.info/archives/38118

#histodons #Christianity #Philosophy