njtierney / broomstick Goto Github PK

• title
• Authors@R
• Description
• License

Also I'd decrement the version of Roxygen to the CRAN version, just to be friendly to other developers.

Just a note to try and avoid using the built-in summaries and importance calculations for rpart, gbm, and so on - it would be ideal to have minimal dependencies on these packages, and to import them, as is done in broom - https://github.com/tidyverse/broom/blob/master/DESCRIPTION

So that you can specify a different dataset and get some predictions back.

I am trying to pipe rpart output into the tidyroc package, but it relies on broom::augment().

The example pipe is:

glm(am ~ disp, 
  family = binomial,
  data = mtcars
  ) %>%
  augment() %>%
  make_roc(predictor = .fitted, known_class = am) %>%
  ggplot(aes(x = fpr, y = tpr)) + 
  geom_line()

Since all the tree based tidiers have been moved to broomstick, I tried that. Unfortunately, while broomstick::tidy() works, broomstick::augment() does not.

I was able to manually generate the required data frame, but is there a reason augment() is missing or is it just not ready yet?

Hey Nicholas,

I'd like to add some tidiers for xgboost sometime this month - is this still under active development/the correct place to submit a PR?
If so, how's that sound?
Thanks!

https://www.tidymodels.org/learn/develop/broom/

Lots of good links from rstats folks: https://twitter.com/nj_tierney/status/1263687639779033088

I'm happy to take point on this.

re: tidymodels/broom#163

https://www.tidyverse.org/articles/2018/11/generics-0.0.1/

Hi there @njtierney! My coworker Hannah Frick said she had chatted with you about this package recently and you said you thought it might be time for some updates.

Would you be open to a few PRs to integrate this package and its machinery more closely with broom, bonsai, and modeltests? What are your thoughts on releasing this package to CRAN?

Happy to put in some effort here and can file a few issues to get discussion started on what I have in mind specifically, if you're up for it! No pressure from me if you're not looking to develop this package further / in this direction. :)