Using the value of a variable to select columns

Question

I have some code that calculates quantiles and saves it under a new variable. Now I am trying to modify the code so I can use it in a loop. Here is a sample dataset, the original code, and the output.

df <- data.frame(country = c("Algeria", "Egypt", "Ethiopia"), 
                 ward_1 = c(25, 75, 35),
                 ward_2 = c(30, 185, 60))

df %>%
  mutate(
    ward_1_quant = case_when(
      ward_1 >= quantile(df$ward_1, 2/3) ~"High",
      ward_1 >= quantile(df$ward_1, 1/3) ~"Med",
      TRUE~"Low"
    )
  )

df <- data.frame(country = c("Algeria", "Egypt", "Ethiopia"), 
                 ward_1 = c(25, 75, 35),
                 ward_2 = c(30, 185, 60),
                 ward_1_quant = c(Low, Med, High))

Now I want to use the value of a variable "ward_num" to specify the column. I turn that into a string called "ward_name" and try to use it in the mutate(). Here is where I got stuck:

ward_num <- 2
ward_name <- paste0("ward_",ward_num)
df %>%
  mutate(
    paste0(ward_name,"_quant") = case_when(
      ward_name >= quantile(df[ward_name], 2/3) ~"High",
      ward_name >= quantile(df[ward_name], 1/3) ~"Med",
      TRUE~"Low"
    )
  )

r returns the error:

Error: unexpected '=' in: " mutate( paste0(ward_name,"_quant") ="

I think the issue is that I want to use the string in ward_name but the code is using ward_name literally. Any suggestions?

Answer 1

We may use .data with [[. Also, the expression on the lhs of = is not permitted. Instead use := and evaluate (!!) the paste expression

library(dplyr)
df %>%
  mutate(
    !! paste0(ward_name,"_quant") := case_when(
      .data[[ward_name]] >= quantile(.data[[ward_name]], 2/3) ~"High",
      .data[[ward_name]] >= quantile(.data[[ward_name]], 1/3) ~"Med",
      TRUE~"Low"
    )
  )

-output

   country ward_1 ward_2 ward_2_quant
1  Algeria     25     30          Low
2    Egypt     75    185         High
3 Ethiopia     35     60          Med

---

Or another option is to pass this in across

df %>% 
  mutate(across(all_of(ward_name),
   ~ case_when(.x >= quantile(.x, 2/3) ~ "High", 
    .x >= quantile(.x, 1/3)~ "Med", TRUE ~ "Low"), 
      .names = "{.col}_quant"))
   country ward_1 ward_2 ward_2_quant
1  Algeria     25     30          Low
2    Egypt     75    185         High
3 Ethiopia     35     60          Med

---

Or another option is to use quantile with cut

df %>%
   mutate(across(all_of(ward_name),
   ~ cut(.x, breaks = c(-Inf, quantile(.x, probs = c(1/3, 2/3)), 
      Inf), labels = c("Low", "Med", "High")), .names = "{.col}_quant"))
     country ward_1 ward_2 ward_2_quant
1  Algeria     25     30          Low
2    Egypt     75    185         High
3 Ethiopia     35     60          Med

Answer 2

Instead of messing around with across(), !! and its friends, I think it's better to use base R here.

ward_num <- 2
ward_name <- paste0("ward_",ward_num)
ward_name_quant <- paste0(ward_name,"_quant")
df[[ward_name_quant]] <- case_when(
                         df[[ward_name]] >= quantile(df[[ward_name]], 2/3) ~ "High",
                         df[[ward_name]] >= quantile(df[[ward_name]], 1/3) ~ "Med",
                         TRUE ~ "Low"
                       )
 

Also, you can use cut, or santoku::chop_quantiles() instead of the case_when call. (Side note: case_when is often a code smell.)

df[[ward_name_quant]] <- cut(df[[ward_name]], 
                             quantile(df[[ward_name]], c(0, 1/3, 2/3, 1)), 
                             c("Low", "Med", "High"), 
                             include.lowest = TRUE
                           )

# or

library(santoku)
df[[ward_name_quant]] <- chop_quantiles(df[[ward_name]], 
                                        c(1/3, 2/3), 
                                        c("Low", "Med", "High")
                                      )