Mark Rieke

some enterprising nerd with more time than me could compare the CRPS from the forecast publications for a better comparison. Even better still would be to see if the outlets (rip 538) could share the actual draws & look at pr(obs | model)

I don't think this metric is necessarily any better than brier scores, but it is ironic that the SB model would've lost if used to inform bets on polymarket, given Silver's affiliation
dl.acm.org/doi/10.1145/...

as a workout, Fran has the highest density of suffering / time spent moving

sometimes they put the yee haw right into the programming language

olympic bobsled should add podracing noises send tweet

Also, Yglesias lecturing people on uncertainty when it comes to interpreting polling data is laughable

Someone touches grass, realized they don't like computer science anymore

Or, as written up in SIGBOIVK last year:

The stupidity, the criminal vandalism, the wanton destruction of information involved in dichotomisation
statmodeling.stat.columbia.edu/2026/02/14/t...

I have concluded that the gamma distribution has too many parameterizations and simply needs to chill, thank you

after my sufficient gamma crashout last night, I've done some simulations & realized:
- the pdf below works for a sufficient gamma
- the original pdf I shared ALSO works for a sufficient statistic under a different parameterization
- weve already done this at work under a DIFFERENT parameterization

we have a slightly different formulation at work that gets around this --- I'll ping you on slack when I get online !

ooh! no, wasn't aware --- aside from the super simple sufficient forms (binomial & counts of poisson obs), I haven't done too much w/sufficiency (I've implemented a sufficient normal like one time remember it being a struggle lol)

oh yeah that's why my gut says that sum(x) ~ Gamma(n * alpha, theta) isn't a sufficient stat.

Link here shows what I get if I just repeatedly multiply the gamma pdf given new elements of x. Need to double check but it feels like this might actually give sufficiency?

bsky.app/profile/mark...

if I just sit down and bake out what a repeated product of the gamma pdf is I end up with this --- needs to be checked with simulation, but certainly passes the gut check

I haven't yet simulated, but IIRC my thinking was that by summarizing (x1, x2, ... xn) to sum(x), you lose some information in that there are many combinations of alpha/theta that can reasonably be fit to a single observation of sum(x)

someone also has pointed out that x ~ Gamma(alpha, theta) is equivalent to sum(x) ~ Gamma(n * alpha, theta). At one point I had convinced myself that there was a loss of individual information here, but now I can't quite see why I came to that conclusion --- gotta simulate some stuff to see !

oops xprod and xsum should index n here (xprod[n] & xsum[n])

spitballing a function and maybe this works? would need to test it out against a non-sufficient gamma

moreso scary in the sense that I've got ptsd from thinking the sufficient normal was straightforward then banging my head against a wall until I got a working solution lol

(the sufficient normal was also my crash course in "sometimes the centered parameterization samples better")

pinging the #bayesian / #rstats / #pydata hivemind --- has anyone implemented a sufficient formulation of a gamma distribution/willing to share code (ideally in stan or pymc)? supposedly this equation is a density function for the sufficient gamma, but ngl it's scary

point estimate probability says 100% chance of bayesian takeover 😈

it is a plot, we will get all of you, eventually, to embrace bayesian inference, resistance is futile, etc. etc.

butterfly meme ME ANYTHING IS THIS A GAUSSIAN PROCESS

New gaussian process slides going well

I have reached the "writing yaml specifications" stage of my career

why is jazz so fuckin' good

"Is my sample large enough for Bayesian statistics?" is a weird question coming from someone who only ever uses frequentist stats that are only approximately correct for large samples.

"production notebook" is an oxymoron and should be shunned from pleasant society

this series is mostly horny nonsense (positive review), but this scene is legitimately incredible and moving

nothing so captivating as watching someone exert complete mastery over a skill with ease

coulda left it at "please don't email me"