Create a sampler function useful for repeated reproducible sampling

Create a sampler function useful for repeated reproducible sampling

sampler(times, size = 1, replace = FALSE, group = NULL,
  seed = NULL, key = ".draw", row = ".row", id = ".id",
  original_id = ".original_id")

Arguments

times	Number of independent sampling draws to perform.
size	Sample size for each random sample.
replace	Bool indicating whether sampling should occur with or without replacement.
group	An optional expression setting up the grouping to use for bootstrapping. If not provided, any grouping present in the original dataset will be used.
seed	Random seed to use.
key	Name (as character) of the column that will hold an integer running from 1 to times indicating the bootstrap replicate.
row	Name (as character) of the column that will hold an integer counting rows in the final bootstrapped dataset. Useful for animations with gganimate.
id	Name (as character) of the column that will hold an integer running from 1 to n for each bootstrap, where n is the number of observations in each group.
original_id	Name (as character) of the column that indicates the row from which the sampled row originates.

See also

bootstrapper()

Examples

spl <- sampler(3, 2)
spl(data.frame(letter = letters[1:4]))
#> # A tibble: 6 x 5
#> # Groups:   .draw [3]
#>   .draw   .id .original_id letter  .row
#>   <int> <int>        <int> <fct>  <int>
#> 1     1     1            3 c          1
#> 2     1     2            4 d          2
#> 3     2     1            1 a          3
#> 4     2     2            4 d          4
#> 5     3     1            1 a          5
#> 6     3     2            2 b          6

library(ggplot2)
library(dplyr)
ggplot(iris, aes(Sepal.Length, Sepal.Width)) +
  geom_point(aes(color = Species), alpha = 0.3) +
  geom_point(
    data = sampler(1, 5, group = Species),
    aes(fill = Species),
    color = "black", shape = 21
  ) +
  theme_bw()

# it is important to set grouping correctly for sampling
set.seed(1234)

df <- data.frame(
  type = c(rep("A", 100), rep("B", 10), rep("C", 3)),
  y = rnorm(113)
)

# incorrect: sampling ungrouped dataset leads to missing data
# in some categories
ggplot(df, aes(type, y)) +
  geom_pointrange(
    data = sampler(6, 3, replace = TRUE, seed = 559),
    stat = "summary"
  ) +
  facet_wrap(~.draw)
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> Warning: Removed 4 rows containing missing values (geom_pointrange).

# correct: sampling within groups
ggplot(df, aes(type, y)) +
  geom_pointrange(
    data = sampler(6, 3, replace = TRUE, group = type, seed = 559),
    stat = "summary"
  ) +
  facet_wrap(~.draw)
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()

# also correct: use grouped data frame
ggplot(group_by(df, type), aes(type, y)) +
  geom_pointrange(
    data = sampler(6, 3, replace = TRUE, seed = 559),
    stat = "summary"
  ) +
  facet_wrap(~.draw)
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()
#> No summary function supplied, defaulting to `mean_se()

# NOT RUN {
library(gganimate)

p <- ggplot(iris, aes(Sepal.Length, Species, color = Species)) +
  geom_point(color = "grey50", alpha = 0.3, size = 2) +
  geom_point(data = sampler(20, 1, group = Species), size = 4) +
  scale_color_brewer(type = "qual", palette = 2, guide = "none") +
  theme_bw()

p + facet_wrap(~.draw)
p + transition_states(.draw, 1, 2)
# }