r-statistics.co by Selva Prabhakaran

About

dplyr case_when(): Replace Nested if_else with Clean Conditional Logic

case_when() evaluates conditions top to bottom and returns the value for the first TRUE match — like SQL's CASE WHEN. It replaces ugly nested ifelse() chains with clean, readable code.

The Problem: Nested ifelse()

# Hard to read, easy to break scores <- c(95, 82, 67, 55, 73) grades <- ifelse(scores >= 90, "A", ifelse(scores >= 80, "B", ifelse(scores >= 70, "C", ifelse(scores >= 60, "D", "F")))) cat("Nested ifelse:", grades, "\n")


  






case_when: The Clean Way





  


    
library(dplyr)

df <- data.frame(name = c("Alice","Bob","Carol","David","Eve"),
                 score = c(95, 82, 67, 55, 73))

df |>
  mutate(grade = case_when(
    score >= 90 ~ "A",
    score >= 80 ~ "B",
    score >= 70 ~ "C",
    score >= 60 ~ "D",
    TRUE         ~ "F"
  ))


    


      

      

    


    


  






Conditions are evaluated top to bottom. The first TRUE match wins. TRUE at the end is the catch-all default — like else in if/else.



Multiple Columns in Conditions





  


    
library(dplyr)

mtcars |>
  mutate(type = case_when(
    mpg > 25 & hp < 100  ~ "Efficient & Light",
    mpg > 25             ~ "Efficient & Powerful",
    hp > 200             ~ "Muscle Car",
    TRUE                 ~ "Standard"
  )) |>
  count(type, sort = TRUE)


    


      

      

    


    


  






Handling NA





  


    
library(dplyr)

df <- data.frame(x = c(1, NA, 3, NA, 5))

# case_when treats NA comparisons as FALSE (no match)
# Put is.na() check FIRST to handle NAs explicitly
df |>
  mutate(label = case_when(
    is.na(x) ~ "Missing",
    x > 3    ~ "High",
    TRUE     ~ "Low"
  ))


    


      

      

    


    


  






case_when vs ifelse vs if_else







Feature
ifelse()
dplyr::if_else()
case_when()






Conditions
1
1
Many




Type safety
No
Yes (strict)
Yes




Readability (nested)
Poor
OK
Excellent




NA handling
Returns NA type
Strict
Explicit




Best for
Simple binary
Type-safe binary
Multi-level







Practical Examples





  


    
library(dplyr)

# Categorize by multiple thresholds
mtcars |>
  mutate(
    mpg_class = case_when(
      mpg > 30 ~ "Excellent",
      mpg > 25 ~ "Good",
      mpg > 20 ~ "Average",
      mpg > 15 ~ "Below Average",
      TRUE     ~ "Poor"
    ),
    size = case_when(
      wt > 4   ~ "Heavy",
      wt > 3   ~ "Medium",
      TRUE     ~ "Light"
    )
  ) |>
  count(mpg_class, size) |>
  head(10)


    


      

      

    


    


  






Practice Exercises



Exercise 1: Categorize Cars



Create a "decade" column based on mpg: >30 "Future", 20-30 "Modern", 15-20 "Standard", <15 "Vintage".





  


    
library(dplyr)



    


      

      

    


    


  






Click to reveal solution

```r


  

    
library(dplyr)

mtcars |>
  mutate(decade = case_when(
    mpg > 30 ~ "Future",
    mpg > 20 ~ "Modern",
    mpg > 15 ~ "Standard",
    TRUE     ~ "Vintage"
  )) |>
  count(decade, sort = TRUE)

    

      
      
    

    

  







FAQ



What does TRUE ~ "default" mean?



TRUE is the catch-all — it matches any row not caught by earlier conditions. Like else in if/else. Always put it last. If you omit it, unmatched rows get NA.



Can I return different types in different conditions?



No. All right-hand-side values must be the same type. case_when(x > 0 ~ "yes", TRUE ~ 0) fails because "yes" (character) and 0 (numeric) conflict. Use case_when(x > 0 ~ 1, TRUE ~ 0) or convert.



Is case_when vectorized?



Yes. It operates on entire vectors at once — no row-by-row loop. This makes it fast, even on millions of rows.



What's Next?




        


        
      


      

        

        
© 2016- Selva Prabhakaran.
          This work is licensed under the Creative Commons License.