Thread by @drob

Let's do this

ONE LIKE = ONE INSULT OF A DISTRIBUTION

https://twitter.com/ChelseaParlett/status/1286032700281720832

NEGATIVE BINOMIAL: You are neither negative nor binomial. Get a better name

POISSON: You are the frictionless spherical cow of distributions

F DISTRIBUTION: Your inventor was a real asshole

NORMAL: In the real world, height is normally distributed. Just about everything else is LOG-NORMAL.

BERNOULLI: Your support is literally two numbers

DIRAC: You make Bernoulli’s support look good

CAUCHY: You literally don't have a mean, variance, skewness, or kurtosis? What are you even doing here? Get bent

BINOMIAL: "Sum of Bernoullis" is not a personality. Do better

POWER LAW: Are you an engineer who hasn't heard of the log-normal?

Are you in a meeting where you want to say "this is often small but can sometimes be big?"

Just say "power law!"

EXPONENTIAL: I hear you're memoryless?

Bet you'll remember this:

You stink

GAMMA: Even CAUCHY has a median, ffs

BETA: You're good

MIXTURE OF NORMALS: Nobody likes a fence-sitter. Pick a mode and STICK TO IT

BETA-BINOMIAL: The only thing worse than a mixture is a continuous mixture. What the hell man

STUDENT'S T DISTRIBUTION: If you're so great, why did your inventor go all Alan Smithee on you. Burn

WISHART: "The GAMMA is so useful, we should generalize it to multiple dimensions" said no one ever

INVERSE-WISHART: When the WISHART isn't esoteric enough for you, take a long hard look at your life

GEOMETRIC: You literally have an off-by-one error in your definition and now we're all paying the price

I'm not even tired

CHI-SQUARED: Ffs how many eugenicists can one field have

CONWAY-MAXWELL-POISSON: It took THREE PEOPLE to come up with this?

UNIFORM: You're the Diet LaCroix of distributions

HALF-NORMAL: Ooh, you took the normal and cut it in half, you sure earned your tenure today

FOLDED NORMAL DISTRIBUTION: Somebody saw the HALF-NORMAL and was like "I can get at least three more papers out of this"

WEIBULL: More like WEIBULLSHIT, am I right?

Yes, yes, WEIBULLSHIT, this is the highbrow statistical content my followers come for

HYPEREXPONENTIAL: Your name contains two synonyms for "fast", you have zero chill

LOG-CAUCHY: Your mean is infinite.

Your variance is infinite.

Your median is about 3.

What the actual fuck

CATEGORICAL: "Pick a card, any card" does not need a 3000-word Wikipedia entry

BENFORD: So you tell me 1 appears more often in leading digits than you'd expect by chance

Take a guess what leading digit I'd rate you out of 10

LOG-LOGISTIC: Statisticians are literally the "Have you tried logarithms?" guy from XKCD

PARETO: You're the LOG-NORMAL for jocks

FRÉCHET: You're the LOG-NORMAL for nerds

BURR: You're the LOG-NORMAL for nerdy economists

LÉVY: You're the LOG-NORMAL for jock-ey chemists

TRUNCATED NORMAL: You make the folks who invented the FOLDED NORMAL look busy

LOMAX: Stop trying to make the PARETO happen! It's not going to happen!

SKELLAM: If the POISSON is a frictionless spherical cow, the SKELLAM is a frictionless spherical cow in a vacuum at sea level

GENERALIZED NORMAL: You know that sophomore philosophy major who shows up in a Reddit thread saying "Well [your field] is really just a special case of philosophy"?

That's you

EVERYONE TELLING ME TO DO THE DIRICHLET NEXT: No

CANTOR: Even the log-Cauchy had a probability density. How do you exist

ERLANG: “Special” case of the GAMMA my ass

I’ve cleaned stuff out of my lint tray that’s more special than you

Applied statistician: [sobbing] You can’t just put “log-“ in front of a distribution and pretend it’s a new distribution!

Mathematical statisician: [points at LAPLACE] LOG-LAPLACE

MAXWELL-BOLTZMANN: I'll just keep my mouth shut, I can't afford this beef

This is the Suge Knight of distributions

SKEW-NORMAL: Got a brand new stats degree

Analyzing your first dataset

It’s not NORMAL

Maybe the solution is to add some skew?

(It isn’t)

(The solution is LOG-NORMAL)

ZERO-INFLATED NEGATIVE BINOMIAL: It's 3 AM, you've got a presentation in the morning, and your count data is really, really overdispersed

If ZINB starts looking good to you, go to sleep

HYPERGEOMETRIC: If I sample you you can bet I’ll be doing some replacement

Boom, roasted

RADEMACHER: Somebody looked at the BERNOULLI and was like "whoa too complicated"

DIRICHLET-MULTINOMIAL Everything about you is extra

DAGUM: You have TWO shape parameters and a scale parameter. Nobody's going to fit three parameters. Who the hell do you think you are

JOHNSON SU: You have four parameters. First of all, how dare you

HYPERBOLIC SECANT: From the name you might think this would be an interesting distribution. You'd be wrong, though. It's boring af

LOG-NORMAL: WE STAN A LEGEND

RAYLEIGH: Hey, uh, where'd you get your moment generating function, at the, uh, idiot store?

I give up, I'm out /END

@mbeisen ^ how did I get away with this