dplyr::across()

This article showcases an example of summarising a dataframe using a combination of the dplyr::summarise and dplyr::across functions.

Within the summarise function, it is possible to summarise numerous columns that adhere to any condition, using clean and succinct syntax with the help of the across function.

Let’s begin by loading the required packages…

# "tidyverse" is a collection of packages that make up the tidy-universe
library(tidyverse)

# Packages included in tidyverse are:
tidyverse::tidyverse_packages(include_self = FALSE)

##  [1] "broom"      "cli"        "crayon"     "dbplyr"     "dplyr"     
##  [6] "forcats"    "ggplot2"    "haven"      "hms"        "httr"      
## [11] "jsonlite"   "lubridate"  "magrittr"   "modelr"     "pillar"    
## [16] "purrr"      "readr"      "readxl"     "reprex"     "rlang"     
## [21] "rstudioapi" "rvest"      "stringr"    "tibble"     "tidyr"     
## [26] "xml2"

Most of the functions used in this article belong to the dplyr package.

The example dataset used in this article is the ggplot2::diamonds dataframe, it contains information of around 54,000 diamonds.

Numeric columns in this dataframe are:

ggplot2::diamonds %>% select(where(is.numeric))

For each numeric column, the min, max and mean can be calculated for each group in the cut column with help of the across function. Summarisation is as follows:

# With the diamonds dataframe:
(df <- ggplot2::diamonds %>% 
  
  # for each category in the "cut" column
  dplyr::group_by(cut) %>% 
  
  # calculate summaries for...
  dplyr::summarise(
    
    # ...each column that:
    dplyr::across(
      
      # is numeric
      .cols = where(is.numeric)
      
      # returning the max, min and mean value
      , .fns = c(min, max, mean)
      
      # and name columns accordingly
      , .names = "{.fn}_{.col}"
    )
  )
)

As can be seen in the above output, column names do not reflect the summarisation function used. Instead, column names follow the logic 1_<attribute>, 2_<attribute>, 3_<attribute> where 1, 2 and 3 represent the min, max and mean, respectively.

Here is a workaround to fix column names:

colnames(df) <- colnames(df) %>% 
  stringr::str_replace_all(
    pattern = c(
      "1" = "min"
      , "2" = "max"
      , "3" = "mean"
    )
  )

df

Voilà! The min, max and mean for each cut group have been calculated for each numeric column, complete with sensible column names, within the diamonds dataframe.

As shown in this article, the across function allows one to carryout summarisation on many columns at once. This is very handy when wanting to summarise a large dataframe.

It is also possible for columns in the across function to be referenced by a naming convention. For example, if one wishes to summarise all columns in a dataset that end in a particular word, it can be achieved using the following:

dplyr::across(
  
  # ends with "word"
  .cols = dplyr::ends_with("word")
  
  # returning the mean
  , .fns = mean
)

In addition to across complimenting summarise so well, it can also be used in conjunction with dplyr::mutate and dplyr::select.

Why not try out the across function today !?

:D