Referencing a nested column using purrr::map-CodePudding

I'm trying to bootstrap some data that are nested by group:

library(tidyverse)
library(tidymodels)    
mtcars %>%
  group_by(cyl) %>%
  nest() %>%
  mutate(boots = map(data, ~bootstraps(.x,times=1000, apparent = TRUE)))

I then need to access and process the data in the boots column in order to calculate the mean mpg, per group. I have tried the following:

calc_mpg_mean <- function(split){
  dat <- analysis(split) %>% pull(mpg)
  return(tibble(
    term = "mean",
    estimate = mean(dat),
    std.err = sd(dat)/sqrt(length(dat))))
}

mtcars %>%
  group_by(cyl) %>%
  nest() %>%
  mutate(boots = map(data, ~bootstraps(.x,times=1000, apparent = TRUE))) %>%
  mutate(mean_mpg = map(boots,calc_mpg_mean))

But this doesn't work because I get:

Error in `mutate()`:
! Problem while computing `mean_mpg = map(boots, calc_mean)`.
ℹ The error occurred in group 1: cyl = 4.
Caused by error in `analysis()`:
! `x` should be an `rsplit` object

Presumably I am not accessing the nested column boots correctly. What am I doing wrong?

CodePudding user response：

I don't have much (any) experience with tidymodels, but it looks like you just need to access your split object a few levels deeper.

library(tidyverse)
library(tidymodels)    

calc_mpg_mean <- function(split, index){
  dat <- analysis(split$splits[[index]]) %>% pull(mpg)
  return(tibble(
    term = "mean",
    estimate = mean(dat),
    std.err = sd(dat)/sqrt(length(dat))))
}

mtcars %>%
  group_by(cyl) %>%
  nest() %>%
  mutate(boots = map(data, ~bootstraps(.x,times=1000, apparent = TRUE))) %>%
  mutate(mean_mpg = imap(boots, ~calc_mpg_mean(.x, .y)))


#> # A tibble: 3 × 4
#> # Groups:   cyl [3]
#>     cyl data               boots                    mean_mpg        
#>   <dbl> <list>             <list>                   <list>          
#> 1     6 <tibble [7 × 10]>  <bootstraps [1,001 × 2]> <tibble [1 × 3]>
#> 2     4 <tibble [11 × 10]> <bootstraps [1,001 × 2]> <tibble [1 × 3]>
#> 3     8 <tibble [14 × 10]> <bootstraps [1,001 × 2]> <tibble [1 × 3]>