Lol they make them dig their own graves and crawl in coffins to symbolize the stakes of getting voted out in this game that is just mafia/werewolf

Is it collider "bias" though? Seems like the predictive result is the actual interesting part they're selling (ie among a certain range of child performance don't worry too much where you fall within it), and the impact of colliders counts!

Does one of the answers change if you change the scale of the data? I'd be surprised if floating point errors show up for the data you've shown on screen but that's one possible explanation.

Very odd! I don't know but curious or find out

i think the only annual watch i do is Elf and if it were up to me i would not watch it again. i have already seen Elf.

if you're counting childhood viewing i have a few. eliminating Disney maybe leaves me with... a land before time? Maybe Totoro but I'm not sure. That might be it.

if i think about it a while i can name maybe four movies i've watched *twice* by my own choice. six times??

I am on parental leave which almost certainly was a contributing factor of me missing this news

@shokru.bsky.social has a recent paper that notices the same issue and suggests an alternative variance decomposition as well as going way more in-depth on the issue of comparison www.paris-december.eu/sites/defaul...

probably better to try to think of what meaningful effect size variation is in a given context which is obviously not as satisfyingly objective :-/

you could find a meaningful way to interpret this metric but it is not as clean or useful as i originally intended it!

which doesn't really work as intended in terms of comparing across different contexts or data cleaning choices, since it's sensitive to things like the sample size of the relevant data set or noise in the true model, even for constant levels of researcher variation

Six