matteomics.bsky.social
Measurements want to be accurate; Experiments want to be elegant; Data wants to be beautiful and Data wants to be free #proteomics #rstats
👤 Joined on 07 January 2025
Followers
Following
Total Posts
@matteomics.bsky.socialjustjosy.bsky.social
thatqbcolliead.bsky.social
niallharbison.bsky.social
dukerocheister.bsky.social
koomabear.bsky.social
bigbuttbiggerheart.online
xxsinnerxx.bsky.social
pervertedkitty.bsky.social
holybible.bsky.social
afterdarkcoffii.bsky.social
beefy-135.bsky.social
saffypuppy.bsky.social
virg0bab3.bsky.social
Not Found
Not Found
helenaskzi3b.bsky.social
padresalsa.bsky.social
redditspaceviewes.bsky.social
Not Found
Not Found
lifeisterrible.bsky.social
aikostratus.bsky.social
deedubcub.bsky.social
realchrisbrunet.bsky.social
Not Found
Emeritus at last! I've gotten more (of my own) work done in the last few weeks than I have in the same number of years. It's expensive to retire early, but not as expensive as my physical and mental health. Cheers and huzzahs!
Show Postagain, the NYT Spelling Bee is wrong. #RStats
Show Posti even have a note where i say "oh weird all of these numbers are coming back 0.8889 i guess i fucked up the test"
Show PostThis is a good excuse to say if you prefer difference scores, you can still condition on the pre score. You can think of it as a difference-score ANCOVA, and it brings the same power advantages as the conventional ANCOVA. Here's a full blog post on the topic: solomonkurz.netlify.app/blog/2023-06...
Show Post