Modelling nested factors in Bambi+PyMC
#stats
thestippe.github.io/statistics/n...
@stippe87.bsky.social
Former theoretical physicist, now data scientist. Bayesian, PyMC addicted. Science outreach enthusiast and dataviz lover. Skeptik Mostly in Turin (IT) https://thestippe.github.io/statistics
Modelling nested factors in Bambi+PyMC
#stats
thestippe.github.io/statistics/n...
Two dangerous terrorists arrested in Trafalgar Square yesterday.
06.10.2025 10:45 β π 37 π 13 π¬ 4 π 1Il #Nobel per la Medicina Γ¨ andato a Mary E. Brunkow, Fred Ramsdell e Shimon Sakaguchi, per aver scoperto come il sistema immunitario evita di attaccare i propri tessuti, attraverso meccanismi di tolleranza immunitaria
ilpost.link/hnk0mhwW3B
External hard disk!
03.10.2025 22:35 β π 2 π 0 π¬ 1 π 0Il fotografo francese Antoni Lallican Γ¨ stato ucciso da un drone in Ucraina
03.10.2025 18:04 β π 10 π 7 π¬ 2 π 0Post muto
(da Instagram)
#MeloniNullitΓ
PubMed is in trouble
This is beyond bad for medical research
pubmed.ncbi.nlm.nih.gov
More Post-Bayes π»
03.10.2025 05:14 β π 6 π 2 π¬ 0 π 0Repeated measures are one of the most commonly misunderstood topics in statistics, and here I will show how to handle them in Python using Bambi and PyMC.
Bonus: to categorize or not to categorize continuous variables such as age?
thestippe.github.io/statistics/r...
Seriously @globalchangebio.bsky.social & @wiley.com ?! π‘π§ͺπ
29.09.2025 15:36 β π 7 π 2 π¬ 0 π 0In Germany it rains a lot, so it's quite common to take off shoes and wear slippers in the office at the uni.
25.09.2025 20:51 β π 1 π 0 π¬ 0 π 0You can't be an expert: according to popular press, experts are ALWAYS puzzled by new findings with an obvious explanation!
24.09.2025 22:39 β π 0 π 0 π¬ 0 π 0The 3rd chapter of the "post-Bayes" seminar, focused on PAC-Bayes bounds, started yesterday. I gave a very introdutory talk, which is already on Youtube.
www.youtube.com/watch?v=hT-d...
There will be 5 more talks in this chapter, see the full schedule there: postbayes.github.io/seminar/
Ma ci tengo anche a ribadire che hai ragione, con l'Ucraina non stiamo facendo abbastanza
23.09.2025 16:05 β π 1 π 0 π¬ 0 π 0D'accordo su tutto, ma credo anche che uno le cause se le debba scegliere, altrimenti ci sarebbero Palestina, Ucraina, Sud Sudan...
Mi pare piu' sensato focalizzarsi su qualcosa su cui si puo' fare la differenza e provare a cambiare le cose. E con Putin non vedo come sia applicabile.
> e le vittime civili di una popolazione non appartenente ad uno Stato da noi riconosciuto, prive di un esercito che le divenda, attaccate da chi si siede accanto a noi all'onu e sostiene di essere nostro alleato e di condividere i nostri stessi valori.
23.09.2025 15:26 β π 3 π 0 π¬ 1 π 0Personalmente vedo una grossa differenza fra le vittime civili di uno stato nostro alleato, che stiamo sostenendo attivamente (anche se non abbastanza, imho) attaccato da un nostro avversario storico >
23.09.2025 15:26 β π 3 π 0 π¬ 1 π 0Non so se non ricordino quando Salvini cantava "lavali col fuoco" o se pensano che sia in grado di distinguere Etna e Vesuvio
21.09.2025 12:14 β π 1 π 0 π¬ 0 π 0Sempre piΓΉ felice di essermene andato dal FVG.
Se vuoi a Torino ho una stanza per gli ospiti.
Une universitΓ© amΓ©ricaine lance une enquΓͺte sur un professeur dβarchΓ©ologie suspectΓ© d'Γͺtre un Β« antifa Β» www.legorafi.fr/2025/09/19/u...
19.09.2025 09:37 β π 2213 π 890 π¬ 37 π 68I'm saying this for years, and finally someone proved it!
delta.tudelft.nl/en/article/i....
The answer is always vim.
18.09.2025 18:54 β π 1 π 0 π¬ 0 π 0E anche la Disney si inchina a Trump...
www.ilpost.it/2025/09/18/j...
Another blog post on a personal experiment: finding the faster algorithm using a completely randomized design.
Why? Computational experiments are fast and cheap, they are an easy way to learn experiment design principles in a hands-on way.
#stats
thestippe.github.io/statistics/crd
Nearly one-quarter of heatwaves would have been βvirtually impossibleβ without global warming β and can be attributed to the emissions of individual energy producers
go.nature.com/3Kir39u
My preference is similar for similar reasons.
application -> theory -> here's how/when it breaks
"when it breaks" is too often underrared
13.09.2025 19:36 β π 2 π 0 π¬ 0 π 0That's Johns Hopkins Professor Marty Makary.
13.09.2025 13:10 β π 3 π 1 π¬ 0 π 0100% agree, that would be a promising combination! unfortunately most ML people (afaik) have little or no knowledge about experimental design
13.09.2025 10:44 β π 1 π 0 π¬ 1 π 0I think you can try with CustomDist. Here is an example discourse.pymc.io/t/creating-a...
13.09.2025 04:35 β π 4 π 1 π¬ 1 π 0