Michael Boo-tancourt
@betanalpha.bsky.social
2.7K followers 160 following 1.6K posts
Zealous modeler. Annoying statistician. Reluctant geometer. Support my writing at http://patreon.com/betanalpha. He/him.
Posts Media Videos Starter Packs
Pinned
betanalpha.bsky.social
Are you excited about the elegant philosophy of Bayesian inference, but struggling to see how it can be applied beyond the idealized examples in introductory texts and tutorials? Over the past few months I’ve released a series of demonstrative analysis that might help. 👇
betanalpha.bsky.social
I wrote about my logo and my recent experiment with some topologically-accurate swag over on patreon dot com, www.patreon.com/posts/logo-h....
betanalpha.bsky.social
On Wed Dec 10 join me to learn about what a regression model is, and what a regression model is not, while raising funds for World Central Kitchen and United Farm Workers. For details on how to register, sponsored registration possibilities, and more see betanalpha.github.io/courses/.
A slide with the text “Predicting a missing variate for a fully observed covariate, however, is not always the relevant predictive task.”, a probabilistic graphical model for predicting a missing variate given a partially observed covariate, and a corresponding equation for the posterior prediction distribution.
Reposted by Michael Boo-tancourt
betanalpha.bsky.social
“Black box modeling assumptions.”
impavid.us
In honor of spooky month, share a 4 word horror story that only someone in your profession would understand

I'll go first: Six page commercial lease.
betanalpha.bsky.social
Zeroeth-order jet, first-order jet, and second-order jet is alphabetical if you just reverse cycle the alphabet by one character…
betanalpha.bsky.social
Agreed! If you're don't find derivatives useful then you're just differentiating the wrong thing. :-p
betanalpha.bsky.social
An explicit target functional, of course, would make this all much less confusing, which nicely builds off of your initial joint modeling comment!
betanalpha.bsky.social
I think the missing gap is that people are implicitly smoothing/mean-fielding the simulation output, but because it's implicit it's not obvious that the same should be applied to derivatives.
betanalpha.bsky.social
The push back I had from my limited conversations with people doing weather/climate forecasting in particular is that the systems tend to be so chaotic that point derivatives aren't that meaningful.
Reposted by Michael Boo-tancourt
konsta.happonen.eu
Relatedly, ages ago @betanalpha.bsky.social wrote "Remember that using Bayes' Theorem doesn't make you a Bayesian. Quantifying uncertainty with probability makes you a Bayesian". I still think about this regularly.
Reposted by Michael Boo-tancourt
banalplay.bsky.social
When your custom character appears in a cut scene
The Portland frog backing down an army of jackboots
Reposted by Michael Boo-tancourt
betanalpha.bsky.social
Friendly reminder that on Dec 10 I will be offering my Bayesian regression modeling course at a steep discount in an effort to raise funds for World Central Kitchen and United Farm Workers. Details about the course and registration process can be found on my website, betanalpha.github.io/courses/.
Courses
betanalpha.github.io
betanalpha.bsky.social
Friendly reminder that on Dec 10 I will be offering my Bayesian regression modeling course at a steep discount in an effort to raise funds for World Central Kitchen and United Farm Workers. Details about the course and registration process can be found on my website, betanalpha.github.io/courses/.
Courses
betanalpha.github.io
betanalpha.bsky.social
Good news! There’s nothing predictive about them. Functions like that implement _retrodictive_ comparisons, making them PR checks which is slightly less embarrassing to say.

<steps off soapbox>
Reposted by Michael Boo-tancourt
dopestatistics.bsky.social
Telling my professor I cant turn in my homework because my expected value is now sentient
Reposted by Michael Boo-tancourt
Reposted by Michael Boo-tancourt
cscheid.net
In case it isn't clear: it's obviously the right thing because

1) one of those two options is easy to undo.
2) Given the choice, a command that is easy to undo should always be the easiest to reach for.
Reposted by Michael Boo-tancourt
bghgmg.bsky.social
gender norms admit the existence of gender metrics which can be abstracted to gender topologies which, here, we categorify into gender locales, also known as pointless genders
Reposted by Michael Boo-tancourt
Reposted by Michael Boo-tancourt
chelseaparlett.bsky.social
HMC: *has lots of divergent transitions*

Me: *fixing it*
Russel from up saying “the target distribution must be explored”
Reposted by Michael Boo-tancourt
betanalpha.bsky.social
On Wed, Dec 10 I will be offering my comprehensive introduction to regression modeling at a steep discount in an effort to raise funds for World Central Kitchen and United Farm Workers. Details about the course and registration process can be found on my website, betanalpha.github.io/courses/.
Courses
betanalpha.github.io
betanalpha.bsky.social
On Wed, Dec 10 I will be offering my comprehensive introduction to regression modeling at a steep discount in an effort to raise funds for World Central Kitchen and United Farm Workers. Details about the course and registration process can be found on my website, betanalpha.github.io/courses/.
Courses
betanalpha.github.io