How can I extract bootstrap generated datasets into individual dataframes?

Question

I am having a bit of trouble understanding Bootstrapping and what/how I can manipulate the bootstrap created dataset.

This is all in R

My original dataset is structured like this:

Time Animal_ID Treatment Volume
0     1          Control   1
2     1          Control   2
4     1          Control   6
0     2          T1        6
2     2          T1        6
4     2          T1        8
0     3          T2        21
2     3          T2        7
4     3          T2        8
0     4          T3        4
2     4          T3        3
4     4          T3        2

I am going to compute a Z-score,mean, and median statistic on the bootstrap generated data sets which is something I can do in the function. However, What I really need as a proof of concepts is a way to extract my bootstrap generated data sets into individual data frames or write them as excel files.

To clarify, I want to showcase a simple example of the statistic I am performing so if I create 5 bootstrap data sets I would need a way to explore them like they were unique individual data sets. I would need to show that this works in concept when looking at R=5 and then I would be able to confidently expand this to R = 100 no extraction required.

Does anyone know how to accomplish this? Or if there are other alternative?

Answer 1

One way is to use the infer package.

library(tidyverse)
library(infer)

df <- read_table(pipe("pbpaste"))

five_bootstraps <- df %>%
  generate(reps = 5, type = "bootstrap")

five_bootstraps

# # A tibble: 60 x 5
# # Groups:   replicate [5]
#     replicate  Time Animal_ID Treatment Volume
#         <int> <int>     <int> <chr>      <int>
#   1         1     4         3 T2             8
#   2         1     2         1 Control        2
#   3         1     0         3 T2            21
#   4         1     0         3 T2            21
#   5         1     2         2 T1             6
#   6         1     0         3 T2            21
#   7         1     0         4 T3             4
#   8         1     2         3 T2             7
#   9         1     0         2 T1             6
#  10         1     2         4 T3             3
# # ... with 50 more rows

So each "bootstrap" replicate is labeled in the replicate column. If you'd like to write each replicate to a separate .csv file, you could do:

five_bootstraps %>%
  split(.$replicate) %>%
  walk2(.x = names(.), .y = ., ~ write_csv(.y, paste0(.x, ".csv")))