Open hadley opened 5 years ago
A two column tidy tibble may make sense
# devtools::install_github("tidyverse/dplyr", ref = "across_simpler")
library(dplyr, warn.conflicts = FALSE)
library(tibble)
library(tidyr)
tidy_quantile <- function(x) {
enframe(quantile(x, probs = c(0, .5, 1)), name = "percentile")
}
# ---
# One column
iris %>%
group_by(Species) %>%
summarise(tidy_quantile(Sepal.Length))
#> # A tibble: 9 x 3
#> Species percentile value
#> <fct> <chr> <dbl>
#> 1 setosa 0% 4.3
#> 2 setosa 50% 5
#> 3 setosa 100% 5.8
#> 4 versicolor 0% 4.9
#> 5 versicolor 50% 5.9
#> 6 versicolor 100% 7
#> 7 virginica 0% 4.9
#> 8 virginica 50% 6.5
#> 9 virginica 100% 7.9
# Multiple columns with a return value in the most useful format
iris %>%
pivot_longer(-Species, names_to = "measure") %>%
group_by(Species, measure) %>%
summarise(tidy_quantile(value))
#> # A tibble: 36 x 4
#> # Groups: Species [3]
#> Species measure percentile value
#> <fct> <chr> <chr> <dbl>
#> 1 setosa Petal.Length 0% 1
#> 2 setosa Petal.Length 50% 1.5
#> 3 setosa Petal.Length 100% 1.9
#> 4 setosa Petal.Width 0% 0.1
#> 5 setosa Petal.Width 50% 0.2
#> 6 setosa Petal.Width 100% 0.6
#> 7 setosa Sepal.Length 0% 4.3
#> 8 setosa Sepal.Length 50% 5
#> 9 setosa Sepal.Length 100% 5.8
#> 10 setosa Sepal.Width 0% 2.3
#> # … with 26 more rows
Created on 2019-11-09 by the reprex package (v0.3.0.9000)
Could possibly have quantile_wide()
and quantile_long()
? Or have some special data frame subclass so that you could could t()
it?
Created on 2019-02-08 by the reprex package (v0.2.1.9000)
Will need to think carefully about how the columns should be named.