Matthew Kay
@mjskay.com
Assoc Prof Computer Science and Communication Studies at Northwestern. Infovis, HCI. Author of tidybayes & ggdist R pkgs. he/him. 🏳️🌈 https://mjskay.com/
Co-director https://mucollective.northwestern.edu
Co-founder https://journalovi.org
Co-director https://mucollective.northwestern.edu
Co-founder https://journalovi.org
Reposted by Matthew Kay
Gee, just fixing something without soliciting bad-faith input from every NIMBY in a three-mile radius. Didn’t know you could do that.
January 6, 2026 at 8:01 PM
Gee, just fixing something without soliciting bad-faith input from every NIMBY in a three-mile radius. Didn’t know you could do that.
pretty sure this is more "oh no"
January 6, 2026 at 4:51 PM
pretty sure this is more "oh no"
to think, we could have had twink pirates
January 6, 2026 at 4:29 PM
to think, we could have had twink pirates
Oh yeah that makes sense. Statefullness is interesting, kind of the opposite of how some tidyverse stuff backs off repeated warnings. Have you thought about interactive prompts? Like "hey you should deal with this eventually! (1) no (2) remind me in 10 runs (3) after 1 hour" etc
January 6, 2026 at 8:14 AM
Oh yeah that makes sense. Statefullness is interesting, kind of the opposite of how some tidyverse stuff backs off repeated warnings. Have you thought about interactive prompts? Like "hey you should deal with this eventually! (1) no (2) remind me in 10 runs (3) after 1 hour" etc
Love this!
Why default action to "warn" instead of "error"? I'd assumed it would be "error" based on the motivating example
Why default action to "warn" instead of "error"? I'd assumed it would be "error" based on the motivating example
January 6, 2026 at 7:26 AM
Love this!
Why default action to "warn" instead of "error"? I'd assumed it would be "error" based on the motivating example
Why default action to "warn" instead of "error"? I'd assumed it would be "error" based on the motivating example
lol I also have 5 internet points, which as we all know is the average of 10 and 0
January 5, 2026 at 10:08 PM
lol I also have 5 internet points, which as we all know is the average of 10 and 0
That would be a fun extension, then you could apply Dempster-Shafer theory to the results
Dempster–Shafer theory - Wikipedia
en.wikipedia.org
January 5, 2026 at 8:04 PM
That would be a fun extension, then you could apply Dempster-Shafer theory to the results
Putting data with model vis often helps judge the quality of the model and the meaningfulness of the phenomena it captures relative to variation in the data. I would only split them if it is necessary for clarity, and I think careful design is better at solving clarity issues than many people think.
January 2, 2026 at 12:40 AM
Putting data with model vis often helps judge the quality of the model and the meaningfulness of the phenomena it captures relative to variation in the data. I would only split them if it is necessary for clarity, and I think careful design is better at solving clarity issues than many people think.
I agree one-sided is often simpler than two-sided versions of these plots and is what I default to
Splitting data and inferential plots is sometimes helpful and sometimes not IMO. I usually lean towards combining them so that inferential uncertainty is placed in the context of the data.
Splitting data and inferential plots is sometimes helpful and sometimes not IMO. I usually lean towards combining them so that inferential uncertainty is placed in the context of the data.
January 2, 2026 at 12:37 AM
I agree one-sided is often simpler than two-sided versions of these plots and is what I default to
Splitting data and inferential plots is sometimes helpful and sometimes not IMO. I usually lean towards combining them so that inferential uncertainty is placed in the context of the data.
Splitting data and inferential plots is sometimes helpful and sometimes not IMO. I usually lean towards combining them so that inferential uncertainty is placed in the context of the data.
every Chicago "tavern style" cracker crust abomination is strictly worse than the exact same pizza with a NY style crust and cut
plus if you want thin crust pizza in Chicago there are good places that do NY style
plus if you want thin crust pizza in Chicago there are good places that do NY style
December 28, 2025 at 5:23 PM
every Chicago "tavern style" cracker crust abomination is strictly worse than the exact same pizza with a NY style crust and cut
plus if you want thin crust pizza in Chicago there are good places that do NY style
plus if you want thin crust pizza in Chicago there are good places that do NY style
I believe "height" is on there
December 27, 2025 at 6:07 AM
I believe "height" is on there
This is the decision tree for millennial Canadians, in my experience some branches (like height) have been pruned for younger generations. Metrification is gradual.
December 26, 2025 at 8:32 PM
This is the decision tree for millennial Canadians, in my experience some branches (like height) have been pruned for younger generations. Metrification is gradual.
Reposted by Matthew Kay
Twitter really let people go feral out there, ready to smash a stranger over the head with a rock over a swollen can of beans.
Elon spent $44 billion to turn Twitter into a goddamn Cormac McCarthy novel
Elon spent $44 billion to turn Twitter into a goddamn Cormac McCarthy novel
October 19, 2024 at 5:54 AM
Twitter really let people go feral out there, ready to smash a stranger over the head with a rock over a swollen can of beans.
Elon spent $44 billion to turn Twitter into a goddamn Cormac McCarthy novel
Elon spent $44 billion to turn Twitter into a goddamn Cormac McCarthy novel