I find it WILDLY FRUSTRATING implementing contrast coding in Julia.
If I use ContrastCoding(), I can specify my own contrast matrices (yay!) but I can't label them. So the regression output just reuses my actual factor levels to label an actual model term that means something like, say, 'mean of levels A and B vs. mean of levels C and D'. Or whatever. To interpret my model, I must make physical notes on a piece of paper about what each term means.
Judging from Dave Kleinschmidt's very useful tour of it, I think possibly it was simply built for someone who learned to specify contrast matrices differently from me.
I infer this from the bit where Dave remarks that the hypothesis matrix is transformed into a contrast matrix in a way that he wouldn't be able to derive off the top of his head. To me, the contrast matrix is great! It's the hypothesis matrix that I have to puzzle over.
@ergative Ah so... yeah a bit of API opacity is common in julia. People are working on that quite hard though, julia community is great at that. If the use-case makes sense (it does!) and there isn't an issue, please do open one.
"The interesting story is not whether psychohistory is correct or how it works. The interesting story is told around what people do when they assume "psychohistory" (prevailing philosophy) is correct, and what happens with people who don't subscribe to that "truth."
Determinism is so much fun to play with, even if there is really no way to determine whether determinism has any actual merits.
The recent series Devs is almost kind of a reboot/prequel almost - sort of, "what if psychohistory was developed as a routine on a really big and fast quantum computer?"
I've had a pair of students who've taken every class I've offered all year and started out really struggling. Like, a lot. Like, getting Cs and Ds in their writing assignments. But they worked super hard, came to office hours, showed me lots of drafts, and their MA dissertation proposals are super good! Well-structured, well-written, showing evidence of really strong learning in both writing and technical experimental design skills. Proper A level work. I'm so proud!
Now that Thing of Thing and Thing is going away, it seems the next titling trend is (often plural) Proximate Demonstrative Determiners Adjective Noun
These Violent Delights
These Broken Stars
These Hollow Vows
These Tangled Vines
These Silent Woods
These Toxic Things
These Impossible Things
This Tender Land
This Woven Kingdom
This Savage Song
These Shallow Graves
These Rebel Waves
These Fleeting Shadows
These Deadly Prophecies
These Monstrous Ties
These Burning Stars
Hi, folks! I've got a set of nano-reviews up at Nerds of a Feather!
Live Long and Evolve: A non-fiction book by an evolutionary biologist about what life on other planets might look like, charmingly interwoven with relevant Star Trek lore.
The Extractionist: a very Cyber futurist heisty type book, which I found well constructed but somehow dull
THe Frame-Up_Magical art thieves. Perfectly fine, but not special.
My sister, who is a few weeks away from being a registered RN told the following story about a very Good Boy.
A patient with schizophrenia suffered from periodic hallucinations of people who weren't there. He got himself a service dog, and one of the service dog's chores was to go up and greet people on command.
So: if the dog went up to greet a person, the person was real. And if the dog did not greet a person when commanded, the person was a hallucination.
Hey, Mastodon, I'm halfway through book four of Libba Bray's 'Diviners' quartet, and it is really good! Just super-solid, well-plotted, well-paced 1920s magical flappers who have to save the world from ghosts. The ensemble cast is effortlessly diverse, and the mood does a brilliant job of capturing that breathless optimism of 1920s America without losing sight of all the darkness lurking underneath it all. A tiny bit purple at times, but super-good.