Matthew Kay
Matthew Kay
Cool! Definitely interested in doing it 😊. I will take a stab in advance of the next iteration of ggdist, probably this summer.
FWIW {testthat} imports {withr}, so adding {withr} to Suggests in {scales} wouldn't add a dependency that's not already downstream.
> I think so. For your data structure it seems like a more efficient proxy from the point of view of vctrs would be an array where the first dimension...
In case it is helpful to others encountering this problem, I used a simple caching approach in https://github.com/stan-dev/posterior/commit/c5036ffdc73fdb05da7199096fb1848c5586e6d3 to improve runtimes on algorithms that call `vec_slice()` repeatedly. There are still...
Wandered onto this from twitter. Love this table. One suggestion might be to also add the corresponding model notation to it for those trying to translate to/from mathematical model notation...
oh yeah, please do, thanks!
Yes, this is because posterior_predict() for those models does not return chain/iteration information. In some cases this could be retrofitted onto the resulting object, though I'm not sure it can...
> In the case of ndraws instead of letting prepare_predictions() handle the conversion of ndraws in `draw_ids, see [here](https://github.com/paul-buerkner/brms/blob/1ed0113eaa052db9d3e308660ab0760c72457712/R/prepare_predictions.R#L29) this could be done by a similar call in the tidybayes...
Yeah, this is a result of the binning algorithm starting from the left edge of the data. Basically, the idea is to: 1. start at the smallest value, say x...
oh yeah, please do, thanks!