Using the bootstrap function about broom

[dgrtwo]

Unfortunately bootstrap hasn't worked with summarize in a while (it did when the function was originally written, but stopped being compatible after a dplyr update/improvement). It relies on a manipulation of the group indices that is consistent with do but not summarize.

My understanding is that bootstrapping in the future should probably be done with the lowliner package, which has some very slick solutions for this kind of problem; I just haven't investigated it yet. The main reason bootstrap is in the package is to cover the cases in the broom manuscript (and I'm certainly open to pull requests that would improve it).

In the meantime you could also work around it with do:

mtcars %>% bootstrap(10) %>% do(data.frame(mean = mean(.$mpg)))

[benmarwick]

Thanks, that's great to know! I was wondering what you plan might be for broom::bootstrap. Thanks very much for making that PR to modelr, hopefully that'll make modelr::bootstrap a bit easier to use. Great to see resampling methods becoming easy to implement in the tidyverse, thanks again!

[vnijs]

Thanks for the workaround David. I will look into the lowliner package as well. Thanks for the tip.

Vincent

—
Sent from Mailbox

On Sun, Feb 1, 2015 at 8:37 PM, David Robinson [email protected]
wrote:

Unfortunately bootstrap hasn't worked with summarize in a while (it did when the function was originally written, but stopped being compatible after a dplyr update/improvement). It relies on a manipulation of the group indices that is consistent with do but not summarize.
My understanding is that bootstrapping in the future should probably be done with the lowliner package, which has some very slick solutions for this kind of problem; I just haven't investigated it yet. The main reason bootstrap is in the package is to cover the cases in the broom manuscript (and I'm certainly open to pull requests that would improve it).
In the meantime you could also work around it with do:

mtcars %>% bootstrap(10) %>% do(data.frame(mean = mean(.$mpg)))

Reply to this email directly or view it on GitHub:
https://github.com/dgrtwo/broom/issues/25#issuecomment-72405484

[benmarwick]

I also stumbled into this recently. Fortunately we now also have a bootstrap in @hadley's modelr, as well as the excellent purrr package (formerly lowliner), and we can do something like this with the above example:

library(dplyr)
library(purrr)
library(modelr)

mtcars %>%
  modelr::bootstrap(10) %>%
  mutate(mean_mpg = map_dbl(strap, ~mean(data.frame(.)$mpg)))

# A tibble: 10 × 3
            strap   .id mean_mpg
           <list> <chr>    <dbl>
1  <S3: resample>    01 20.26875
2  <S3: resample>    02 19.89062
3  <S3: resample>    03 19.20000
4  <S3: resample>    04 19.79062
5  <S3: resample>    05 19.50625
6  <S3: resample>    06 19.19063
7  <S3: resample>    07 19.40625
8  <S3: resample>    08 21.53125
9  <S3: resample>    09 19.84375
10 <S3: resample>    10 20.97812

I was a little surprised about the need to coerce the resample object in the strap column to a data frame in order to access the mpg column. Is that intentional?

[dgrtwo]

Yes, I'm generally recommending people use modelr's bootstrap rather than broom's now. Once modelr is stable and on CRAN I will deprecate broom::bootstrap.

As for needing to do as.data.frame(): a resample object is more like a promise of a data frame than an actual one (though not technically a promise) and only creates the object when requested. It has recently occurred to me that we should really have $ or [[]] evaluate as well and I plan on submitting a modelr pull request to that effect soon!

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.