Toots for mjskay.com@bsky.brid.gy account

Written by Matthew Kay on 2025-01-29 at 14:47

This entire thread is why my approach to typography in #rstats is "do the typography in Illustrator"

RE: https://bsky.app/profile/did:plc:myqic77tkmdhmaufgkvj43hr/post/3lguwln6dq22h

=> More informations about this toot | View the thread

Written by Matthew Kay on 2025-01-17 at 20:24

in honor of David Lynch, my partner wearing one of the greatest t-shirts ever invented

=> View attached media

=> More informations about this toot | View the thread

Written by Matthew Kay on 2025-01-12 at 04:25

lol it seems you can do C-like += in #rstats after all

+<- = +

x = 1

+(x) = 2

x

[1] 3

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-12-04 at 15:13

hear me out: if a #datavis API has ops for superimposing layers (+), combining layer properties (*), and juxtaposition (vector concat: c()), then the matrix product of a col vector of y encodings (aes(y = var1..n)) and a row vector of x encodings (aes(x = var1..n)) is a scatterplot matrix #rstats

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-24 at 04:56

The top of the visual hierarchy is the function responsible for each layer --- 11pt heavy weight (900), to pull it above the body text (400) and variables (700) --- paired with a circled label ("B") at a smaller size (~8pt) but with weight (788) punched up to better match the function name

=> View attached media

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-24 at 04:56

To give an example of what I was talking about in this #datavis #typography thread: this figure uses at least seven (just counted!) different weights of Source Sans Variable to both create visual hierarchy/emphasis and to maintain even visual texture

RE: https://bsky.app/profile/did:plc:6fs5x6cofzyz4si2xhqiqj37/post/3lbnnlerwqc2s

=> View attached media

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-24 at 04:56

Meanwhile in table headers, I set the Roman letters at a bit of an odd weight (634) to try to visually match the Greek letters from Cambria, which only has two weights and a different stroke contrast (ratio of thick to thin strokes), so I went for a kind of visual average --- maybe could be better

=> View attached media

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-20 at 07:52

This is a nice use of {distributional} and {ggdist} for figuring out some priors on the logit scale #rstats

RE: https://bsky.app/profile/did:plc:w4ciafcy2ct6hf2eypzbmszl/post/3lbd472gi622z

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-15 at 04:21

apropos of nothing (ahem #rstats bsky atm), the people who have useful opinions about the difference b/w two languages/APIs are those who have actually built real shit in those languages/APIs, and their takes are invariably very different from some jabroni who looked at the syntax once and got mad

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-12 at 22:58

The cool thing about multiple dispatch is it makes me want to turn every API into an abstract algebra

The problem with multiple dispatch is it makes me want to turn every API into an abstract algebra

[#]S7 #rstats

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-10 at 17:32

so all we needed to do to kill X was sacrifice one (1) democracy

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-11-03 at 20:50

figuring out {targets} #rstats on Northwestern's HPC, part 507:

compiling rstanarm takes so much RAM that if I try to do it on a login node the process will be killed, had to provision a compute node (!!) just to compile the package

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-10-30 at 00:33

is that nate silver

RE: https://bsky.app/profile/did:plc:5jb2734ccyccx6a3hjjuzhrr/post/3l7ov5owttx27

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-10-20 at 16:51

It's an interesting question: if #rstats did a partially-backwards-incompatible break in a major version, what would you fix? Thoughts:

• Consistent NSE (obviously)

• One object system (maybe two for reference types)

• Gradual typing

• Fix API warts like default drop = TRUE

• ??

RE: https://bsky.app/profile/did:plc:aljpcrrf7qt6lhlnumjum6oz/post/3l6xfcthnyx2f

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-09-08 at 11:19

I hereby publicly commit --- when I do finally write a function for cumulative fuzzy bar charts --- to break from the #rstats tradition of cumsum() and to not call it geom_cumbar() or geom_cumstack()

even though I could

RE: https://bsky.app/profile/did:plc:6fs5x6cofzyz4si2xhqiqj37/post/3l3mm7h3zgm2t

=> More informations about this toot | View the thread

Written by Matthew Kay on 2024-09-07 at 14:25

I work in #datavis, best known for uncertainty vis: see my work on election forecast vis (forecasts.cs.northwestern.edu), this talk (youtu.be/E1kSnWvqCw0), or my lab (mucollective.northwestern.edu)

I like writing #rstats 📦s: ggdist (mjskay.github.io/ggdist), tidybayes, ggblend, posterior::rvar...

RE: https://bsky.app/profile/did:plc:dgqjm6bwa4hldo6t2hiq7dkp/post/3l3igchyf5g2m

=> View attached media | View attached media

=> More informations about this toot | View the thread

=> This profile without reblog | Go to mjskay.com@bsky.brid.gy account