How to unpack the group_by() do() output from dplyr pipe - r

Question

I have the following code that performs the summary() for the iris$Petal.Width grouped by species
library(tidyverse)
dat <- iris %>%
as.tibble() %>%
select(Petal.Width, Species) %>%
group_by(Species) %>%
do(fn = summary(.$Petal.Width))
dat
#> Source: local data frame [3 x 2]
#> Groups: <by row>
#>
#> # A tibble: 3 x 2
#> Species fn
#> * <fct> <list>
#> 1 setosa <S3: summaryDefault>
#> 2 versicolor <S3: summaryDefault>
#> 3 virginica <S3: summaryDefault>
What I want to do is to unpack the fn column into the following ( I do this by hand)
Species Min. 1st Qu. Median Mean 3rd Qu. Max.
setosa 0.100 0.200 0.200 0.246 0.300 0.600
versicolor 1.000 1.200 1.300 1.326 1.500 1.800
virginica 1.400 1.800 2.000 2.026 2.300 2.500
How can I do it?
I tried this but failed dat %>% ungroup(fn)

Answer 1

Presently, explicit list columns are preferred as an idiom over do. In this case, it might look like
library(tidyverse)
iris %>%
group_by(Species) %>%
summarise(summary = list(broom::tidy(summary(Petal.Width)))) %>%
unnest()
#> # A tibble: 3 x 7
#> Species minimum q1 median mean q3 maximum
#> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 setosa 0.100 0.200 0.200 0.246 0.300 0.600
#> 2 versicolor 1.00 1.20 1.30 1.33 1.50 1.80
#> 3 virginica 1.40 1.80 2.00 2.03 2.30 2.50
If you like, this is one of those cases where the base R idiom is more concise and readable:
aggregate(Petal.Width ~ Species, iris, summary)
#> Species Petal.Width.Min. Petal.Width.1st Qu. Petal.Width.Median
#> 1 setosa 0.100 0.200 0.200
#> 2 versicolor 1.000 1.200 1.300
#> 3 virginica 1.400 1.800 2.000
#> Petal.Width.Mean Petal.Width.3rd Qu. Petal.Width.Max.
#> 1 0.246 0.300 0.600
#> 2 1.326 1.500 1.800
#> 3 2.026 2.300 2.500
However, note that if you call str on the result, it shows Petal.Width is actually a matrix column (which isn't possible in tibbles, but is in data.frames). To extract it, tack on %>% {cbind(.[1], .[[2]])} or equivalent.
skimr::skim is another option which respects dplyr grouping:
library(dplyr)
iris %>% group_by(Species) %>% skimr::skim(Petal.Width)
#> Skim summary statistics
#> n obs: 150
#> n variables: 5
#> group variables: Species
#>
#> Variable type: numeric
#> Species variable missing complete n mean sd p0 p25 p50 p75 p100
#> setosa Petal.Width 0 50 50 0.25 0.11 0.1 0.2 0.2 0.3 0.6
#> versicolor Petal.Width 0 50 50 1.33 0.2 1 1.2 1.3 1.5 1.8
#> virginica Petal.Width 0 50 50 2.03 0.27 1.4 1.8 2 2.3 2.5
#> hist
#> ▂▇▁▂▂▁▁▁
#> ▆▃▇▅▆▂▁▁
#> ▂▁▇▃▃▆▅▃
What it displays is actually a print method for underlying long data. skimr is built to keep working with dplyr methods, but at some point you may need to hack the underlying data out. The documentation explains pretty well.

Answer 2

Try
dat <- iris %>%
as.tibble() %>%
select(Petal.Width, Species) %>%
group_by(Species) %>%
do(fn = summary(.$Petal.Width) %>% as.matrix() %>% t() %>% as.data.frame())
dat %>% unnest()
# # A tibble: 3 x 7
# Species Min. `1st Qu.` Median Mean `3rd Qu.` Max.
# <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
# 1 setosa 0.100 0.200 0.200 0.246 0.300 0.600
# 2 versicolor 1.00 1.20 1.30 1.33 1.50 1.80
# 3 virginica 1.40 1.80 2.00 2.03 2.30 2.50