Bayesian Updating be like
Frei nach Sepp Herberger, nach dem Updaten ist vor dem Updaten.
November 3, 2026 at 7:14 AM
Bayesian Updating be like
My newest #Rstats package, {bartisan}, was accepted by CRAN today! {bartisan} implements Bayesian Additive Regression Trees (BART), a machine learning method that has the benefits of Bayesian posterior inference and automatic regularization.
#statssky #episky #causalinference #ML
#statssky #episky #causalinference #ML
bartisan: Generalized Bayesian Additive Regression Trees
Fits Bayesian additive regression trees (BART) models for likelihoods outside the conditionally conjugate Gaussian case, using the Laplace-approximation reversible-jump sampler of Linero (2025) <<a...
cran.r-project.org
October 28, 2026 at 6:22 PM
My newest #Rstats package, {bartisan}, was accepted by CRAN today! {bartisan} implements Bayesian Additive Regression Trees (BART), a machine learning method that has the benefits of Bayesian posterior inference and automatic regularization.
#statssky #episky #causalinference #ML
#statssky #episky #causalinference #ML
Bayesian analysis only gets you so far.
It’s like stearing a ship by looking only at the wake.
It’s like stearing a ship by looking only at the wake.
November 2, 2026 at 3:51 PM
Bayesian analysis only gets you so far.
It’s like stearing a ship by looking only at the wake.
It’s like stearing a ship by looking only at the wake.
more like Bayesian prions
October 30, 2026 at 7:35 PM
more like Bayesian prions
Putting up a demonstrative showing the Bayesian priors
November 2, 2026 at 4:06 PM
Putting up a demonstrative showing the Bayesian priors
*clears throat*
Equal to or greater than 95% identical at the barcoding gene (i.e. ITS1 through the D1/D2 variable domains of the rDNA LSU) and show very high (>97%) Bayesian posterior probability and/or bootstrap values in a phylogenetic tree.
Equal to or greater than 95% identical at the barcoding gene (i.e. ITS1 through the D1/D2 variable domains of the rDNA LSU) and show very high (>97%) Bayesian posterior probability and/or bootstrap values in a phylogenetic tree.
November 2, 2026 at 8:55 PM
*clears throat*
Equal to or greater than 95% identical at the barcoding gene (i.e. ITS1 through the D1/D2 variable domains of the rDNA LSU) and show very high (>97%) Bayesian posterior probability and/or bootstrap values in a phylogenetic tree.
Equal to or greater than 95% identical at the barcoding gene (i.e. ITS1 through the D1/D2 variable domains of the rDNA LSU) and show very high (>97%) Bayesian posterior probability and/or bootstrap values in a phylogenetic tree.
I'd just like to call everyone's attention that unless you are computing Bayesian credible intervals, which I don't think it's the case here, it doesn't make sense to say that parts of a realization of a confidence interval are more likely than others.
November 3, 2026 at 2:42 AM
I'd just like to call everyone's attention that unless you are computing Bayesian credible intervals, which I don't think it's the case here, it doesn't make sense to say that parts of a realization of a confidence interval are more likely than others.
Nice priors. Bayesian?
October 30, 2026 at 5:05 PM
Nice priors. Bayesian?
as a less terrible justification for doing this: confidence intervals from maximum likelihood estimation are roughly equivalent to approximate Bayesian estimation using a flat prior and Laplace approximation credible intervals
November 2, 2026 at 1:19 PM
as a less terrible justification for doing this: confidence intervals from maximum likelihood estimation are roughly equivalent to approximate Bayesian estimation using a flat prior and Laplace approximation credible intervals
But then you are being Bayesian, Mark!
November 3, 2026 at 3:40 AM
But then you are being Bayesian, Mark!
I'm a data editor for Ecology Letters now!
stoked to provide this service to the field. my grad school self who barely survived the Bayesian Modeling for Ecologists short course at SESYNC would never believe it
stoked to provide this service to the field. my grad school self who barely survived the Bayesian Modeling for Ecologists short course at SESYNC would never believe it
October 30, 2026 at 3:02 PM
I'm a data editor for Ecology Letters now!
stoked to provide this service to the field. my grad school self who barely survived the Bayesian Modeling for Ecologists short course at SESYNC would never believe it
stoked to provide this service to the field. my grad school self who barely survived the Bayesian Modeling for Ecologists short course at SESYNC would never believe it
(frequentist inference is not synonymous with null-hypothesis significance testing, and Bayesian decision theory is A Whole Thing!)
November 3, 2026 at 5:27 AM
(frequentist inference is not synonymous with null-hypothesis significance testing, and Bayesian decision theory is A Whole Thing!)
yeah my Bayesian priors have become too updated to "it can no good happen, only bad can happen"
I hope I'm wrong but
I hope I'm wrong but
Democrats are more likely than not to win both the House and the Senate tomorrow, according to our forecasts at
@fiftyplusone.news. But remember:
@fiftyplusone.news. But remember:
November 2, 2026 at 7:03 PM
yeah my Bayesian priors have become too updated to "it can no good happen, only bad can happen"
I hope I'm wrong but
I hope I'm wrong but
METHODS: We used a Bayesian productivity-adjusted life-year framework with archived Bayesian age-period-cohort projections. Data were derived from Global Burden of Disease Study 2023 estimates. (🧵 2/6)
November 2, 2026 at 10:41 AM
METHODS: We used a Bayesian productivity-adjusted life-year framework with archived Bayesian age-period-cohort projections. Data were derived from Global Burden of Disease Study 2023 estimates. (🧵 2/6)
The most fancy Bayesian model in the universe is not going to turn messy underlying data into truth.
November 1, 2026 at 8:10 PM
The most fancy Bayesian model in the universe is not going to turn messy underlying data into truth.
Oh I don't think Fisher would like being told he's anything like a Bayesian...........
November 3, 2026 at 7:57 AM
Oh I don't think Fisher would like being told he's anything like a Bayesian...........
giving a talk this week, "The Twist: Capitalist Temporality,
Fascism, & the new Bayesian, Intersectional Left, which can be summarized by two slides.
1/
Fascism, & the new Bayesian, Intersectional Left, which can be summarized by two slides.
1/
November 1, 2026 at 9:35 AM
giving a talk this week, "The Twist: Capitalist Temporality,
Fascism, & the new Bayesian, Intersectional Left, which can be summarized by two slides.
1/
Fascism, & the new Bayesian, Intersectional Left, which can be summarized by two slides.
1/
My Bayesian prior is so tight that regardless of what happens tomorrow, I already know the answer 🫡
November 2, 2026 at 11:44 PM
My Bayesian prior is so tight that regardless of what happens tomorrow, I already know the answer 🫡
It helps that there are now several online Bayesian probability calculators. But, I don't see it taking root before they move on. Medicine would seem a prime discipline for Bayesian thinking, especially in emerging disease epidemiol and pandemic preparedness. Also in health equity.
November 2, 2026 at 12:12 AM
It helps that there are now several online Bayesian probability calculators. But, I don't see it taking root before they move on. Medicine would seem a prime discipline for Bayesian thinking, especially in emerging disease epidemiol and pandemic preparedness. Also in health equity.
my spicy opinion is that for all practical purposes frequentist statistics is just a subset of bayesian statistics 🌶️
November 3, 2026 at 3:43 AM
my spicy opinion is that for all practical purposes frequentist statistics is just a subset of bayesian statistics 🌶️
If you are interested in how to model and quantify metacognition within a flexible hierarchical Bayesian framework, @kevingoneill.github.io has developed a fantastic tutorial paper to complement our new R package hmetad. Do check it out!
this halloween, why not read up on the scariest topic of all (hierarchical Bayesian modeling)? 🦇🎃💀
I've got a new tutorial with @smfleming.bsky.social demonstrating the use of our R package hmetad!
📄: osf.io/preprints/ps...
🔗: metacoglab.github.io/hmetad_tutor...
📦: metacoglab.github.io/hmetad/
I've got a new tutorial with @smfleming.bsky.social demonstrating the use of our R package hmetad!
📄: osf.io/preprints/ps...
🔗: metacoglab.github.io/hmetad_tutor...
📦: metacoglab.github.io/hmetad/
Quantifying metacognition with the hmetad package in R
A longstanding goal in metacognition research is the quantification of confidence-related parameters independently from task performance. While there is now a proliferation of generative models of con...
osf.io
October 30, 2026 at 1:01 PM
If you are interested in how to model and quantify metacognition within a flexible hierarchical Bayesian framework, @kevingoneill.github.io has developed a fantastic tutorial paper to complement our new R package hmetad. Do check it out!
Is the puzzle bayesian?
November 2, 2026 at 8:00 PM
Is the puzzle bayesian?
Been saying this. I assumed polls were skewed, and now I know that they were intentionally skewed. There's only one rational belief this Bayesian can have.
October 31, 2026 at 8:43 PM
Been saying this. I assumed polls were skewed, and now I know that they were intentionally skewed. There's only one rational belief this Bayesian can have.