CODEWITHSUNDEEP
rdplyrtibble
Emman
Emman

Reputation: 4201

Using case_when, how to mutate a new list-column that nests a vector within?

I'm trying to use dplyr's case_when() to mutate a new column based on conditions in other columns. However, I want the new column to be nesting a vector.

Example

Consider the following toy data. Based on it, I want to summarize the geographical territory of the UK.

library(tibble)

set.seed(1)
my_mat <- matrix(sample(c(TRUE, FALSE), size = 40, replace = TRUE), nrow = 10, ncol = 4) 
colnames(my_mat) <- c("England", "Wales", "Scotland", "Northern_Ireland")
my_df <- as_tibble(my_mat)

> my_df

## # A tibble: 10 x 4
##    England Wales Scotland Northern_Ireland
##    <lgl>   <lgl> <lgl>    <lgl>           
##  1 TRUE    TRUE  TRUE     FALSE           
##  2 FALSE   TRUE  TRUE     FALSE           
##  3 TRUE    TRUE  TRUE     TRUE            
##  4 TRUE    TRUE  TRUE     FALSE           
##  5 FALSE   TRUE  TRUE     TRUE            
##  6 TRUE    FALSE TRUE     TRUE            
##  7 TRUE    FALSE FALSE    FALSE           
##  8 TRUE    FALSE TRUE     TRUE            
##  9 FALSE   FALSE TRUE     FALSE           
## 10 FALSE   TRUE  FALSE    FALSE

I want to mutate a new collective_geo_territory column.

  1. if both England, Scotland, Wales, and Northern_Ireland are TRUE, then we say this is United_Kingdom.
  2. otherwise, if only England, Scotland, and Wales are TRUE, then we say this is Great_Britain
  3. any other combination would simply return a vector with the names of countries that are TRUE.

My attempt

So far, I know how to address conditions (1) and (2) detailed above, using the following code

library(dplyr)

my_df %>%
  mutate(collective_geo_territory = case_when(England == TRUE & Wales == TRUE & Scotland == TRUE & Northern_Ireland == TRUE ~ "United_Kingdom",
                                              England == TRUE & Wales == TRUE & Scotland == TRUE ~ "Great_Britain"))

Desired Output

However, I want to achieve an output with collective_geo_territory column that looks like the following:

## # A tibble: 10 x 5
##      England Wales Scotland Northern_Ireland collective_geo_territory
##      <lgl>   <lgl> <lgl>    <lgl>            <list>                   
##   1  TRUE    TRUE  TRUE     FALSE            <chr [1]>   # c("Great_Britain")           
##   2  FALSE   TRUE  TRUE     FALSE            <chr [2]>   # c("Wales", "Scotland")                      
##   3  TRUE    TRUE  TRUE     TRUE             <chr [1]>   # c("United_Kingdom")        
##   4  TRUE    TRUE  TRUE     FALSE            <chr [1]>   # c("Great_Britain")
##   5  FALSE   TRUE  TRUE     TRUE             <chr [3]>   # c("Wales", "Scotland", "Northern_Ireland")
##   6  TRUE    FALSE TRUE     TRUE             <chr [3]>   # c("England", "Scotland", "Northern_Ireland")
##   7  TRUE    FALSE FALSE    FALSE            <chr [1]>   # c("England") 
##   8  TRUE    FALSE TRUE     TRUE             <chr [3]>   # c("England", "Scotland", "Northern_Ireland")
##   9  FALSE   FALSE TRUE     FALSE            <chr [1]>   # c("Scotland") 
##   10 FALSE   TRUE  FALSE    FALSE            <chr [1]>   # c("Wales")

Upvotes: 0

Views: 660

Answers (1)

Emily Riederer
Emily Riederer

Reputation: 106

Here's one approach:

library(purrr) # used for pmap

my_df %>%
  mutate(collective_geo_territory = case_when(
    England & Wales & Scotland & Northern_Ireland ~ list("United_Kingdom"),
    England & Wales & Scotland ~ list("Great_Britain"),
    TRUE ~ pmap(my_df, ~names(my_df)[c(...)]))
    )

Essentially, the last line works as follows:

  1. The left-hand side can simply be TRUE because case_when() terminates on the first relevant TRUE. So, we will only reach this line if conditions 1 and 2 have failed.
  2. The right-hand side essentially says iterate over the rows of my dataset (pmap) and apply the follow function: get the names of the columns in my dataset (names) and subset them ([]) only to those where the values are true (contained in c())

A few additional notes:

  1. Note that I also had to wrap the right-hand slide of the first two conditions (e.g. "United_Kingdom") in a list() because case_when() requires consistent types for the resulting vector
  2. I changed the redundant England == TRUE (and same for other countries) simply to England. Since these columns already contain logical values, there's no need to recheck their values, and this makes the code a bit more readable.

Upvotes: 3

Related Questions