R function: using argument as a character variable and column name-CodePudding

I have some data in this format:

#> # A tibble: 3 × 5
#>    item  cost  blue  pink black
#>   <int> <int> <int> <int> <int>
#> 1     1     4     1     0     1
#> 2     2    10     1     0     1
#> 3     3     3     0     1     1

The output I would like is the relative frequency of the color columns. Items can have more than one color, so the relative frequencies don't have to sum to 1.

#>    color  rel_freq  
#>   <int> <int> 
#> 1  blue  0.66  
#> 2  pink  0.33  
#> 2  black 1.00

I can do it manually for one color, say, blue:

library(tidyverse)
df <- tibble::tribble(
  ~item, ~cost, ~blue, ~pink, ~black,
     1L,    4L,    1L,    0L,     1L,
     2L,   10L,    1L,    0L,     1L,
     3L,    3L,    0L,    1L,     1L
  )

df %>% 
  group_by(blue) %>% 
  summarise(count = n()) %>% 
  mutate(rel_freq = (count/sum(count)*100) ) %>%
  filter(blue==1) %>% 
  mutate(color = deparse(substitute(blue))) %>% 
  select(-blue, -count) %>% 
  select(color, everything())

which gives

#>   color rel_freq
#>   <chr>    <dbl>
#> 1 blue      66.7

but when I put it into a function, I can't figure out how to pass in an argument so that it can be treated as a column (using "curly-curly" notation) and as a character variable (that's what I'm stuck on).

calc_rel_freq <- function(input_color){
df %>% 
  group_by({{input_color}}) %>% 
  summarise(count = n()) %>% 
  mutate(rel_freq = (count/sum(count)*100) ) %>%
  filter({{input_color}}==1) %>% 
  mutate(color = deparse(substitute({{input_color}}))) %>% # This is where I'm stuck.
  select(-{{input_color}}, -count) %>% 
  select(color, everything())
}
calc_rel_freq(blue)

My final goal is to be able to use a function like this:

input_colors <- c("blue", "pink", "black")
map(input_colors, calc_relative_freq)

Code for data input below:

library(tidyverse)
df <- tibble::tribble(
  ~item, ~cost, ~blue, ~pink, ~black,
     1L,    4L,    1L,    0L,     1L,
     2L,   10L,    1L,    0L,     1L,
     3L,    3L,    0L,    1L,     1L
  )

df

CodePudding user response：

Since your desired output is the relative frequencies, you can do that more directly with

df %>% 
  select(-cost) %>% 
  pivot_longer(blue:black) %>% 
  group_by(name) %>% 
  summarize(rel_freq=mean(value))
#   name  rel_freq
#   <chr>    <dbl>
# 1 black    1    
# 2 blue     0.667
# 3 pink     0.333

If you really wanted just one, you can just filter() at the end.

CodePudding user response：

We can use dplyr with across.

library(dplyr)
df %>% summarise(across(blue:black, mean))

# A tibble: 1 × 3
   blue  pink black
  <dbl> <dbl> <dbl>
1 0.667 0.333     1

across can also work with a vector of names of selected columns with the all_of selection helper:

library(dplyr)

input_colors <- c("blue", "pink", "black")

df %>% summarise(across(all_of(input_colors), mean))

If we really want to use non-standard evaluation to use character elements as column selections inside a dplyr function, we can convert to symbol (sym) and evaluate (!!), with the rlang package:

library(dplyr)
library(purrr)
library(rlang)

map_dfc(input_colors, ~df %>% summarise(across(!!(sym(.x)), mean)))

# A tibble: 1 × 3
   blue  pink black
  <dbl> <dbl> <dbl>
1 0.667 0.333     1