thread that's the beauty of ML
from VeryComplicatedLibraryMadeBySomeoneMuchClevererThanMe import bootstrap


permalink someone just tried to explain the logic of using an IPCW Cox model
he might as well have been speaking to a cat for all the good it did him.
permalink assert dominance by making up your own model
'I understand your thinking but wouldn't the Gremblefunk-Schlossinger Matrix be more suited to the underlying distribution? I mean, you're the expert, but that would seem the better idea given the third-order skew of the residuals?'

and they'll try and save face by pretending they've heard of it

NB: this doesn't work on the 1 in 100 stats people who aren't winging it