r-statistics.co by Selva Prabhakaran

About

dplyr group_by() + summarise(): Aggregate Data by Group (10 Examples)

group_by() splits data into groups. summarise() collapses each group to a single summary row. Together they replace complex tapply() and aggregate() calls with readable, chainable code.

"What's the average mpg per cylinder count?" is a grouped summary. Without dplyr you'd write tapply(mtcars$mpg, mtcars$cyl, mean). With dplyr: mtcars |> group_by(cyl) |> summarise(avg = mean(mpg)). Same answer, clearer intent.

Example 1: Basic Grouped Summary

library(dplyr) mtcars |> group_by(cyl) |> summarise( count = n(), avg_mpg = round(mean(mpg), 1), avg_hp = round(mean(hp), 0), .groups = "drop" )


  






Example 2: Multiple Grouping Variables





  


    
library(dplyr)

mtcars |>
  group_by(cyl, am) |>
  summarise(
    n       = n(),
    avg_mpg = round(mean(mpg), 1),
    .groups = "drop"
  )


    


      

      

    


    


  






Example 3: Many Summary Functions





  


    
library(dplyr)

mtcars |>
  group_by(cyl) |>
  summarise(
    n       = n(),
    mean    = round(mean(mpg), 1),
    sd      = round(sd(mpg), 1),
    min     = min(mpg),
    max     = max(mpg),
    median  = median(mpg),
    .groups = "drop"
  )


    


      

      

    


    


  






Example 4: across() for Multiple Columns



Apply the same summary to several columns at once.





  


    
library(dplyr)

mtcars |>
  group_by(cyl) |>
  summarise(
    across(c(mpg, hp, wt), ~ round(mean(.x), 1)),
    .groups = "drop"
  )


    


      

      

    


    


  






Example 5: Multiple Functions per Column





  


    
library(dplyr)

mtcars |>
  group_by(cyl) |>
  summarise(
    across(c(mpg, hp),
      list(mean = ~ round(mean(.x), 1), sd = ~ round(sd(.x), 1)),
      .names = "{.col}_{.fn}"
    ),
    .groups = "drop"
  )


    


      

      

    


    


  






Example 6: count() and Proportions



count() is a shortcut for group_by() + summarise(n = n()).





  


    
library(dplyr)

mtcars |>
  count(cyl, am, sort = TRUE) |>
  mutate(pct = round(n / sum(n) * 100, 1))


    


      

      

    


    


  






Example 7: Grouped mutate (Window Functions)



group_by() + mutate() keeps all rows but computes values within each group.





  


    
library(dplyr)

mtcars |>
  group_by(cyl) |>
  mutate(
    mpg_rank     = rank(-mpg),
    pct_of_avg   = round(mpg / mean(mpg) * 100, 1)
  ) |>
  select(mpg, cyl, mpg_rank, pct_of_avg) |>
  filter(mpg_rank <= 3) |>
  arrange(cyl, mpg_rank) |>
  ungroup()


    


      

      

    


    


  






Example 8: Handling NAs





  


    
library(dplyr)

df <- data.frame(
  group = c("A","A","A","B","B","B"),
  value = c(10, NA, 12, 20, 30, NA)
)

df |>
  group_by(group) |>
  summarise(
    mean_val = mean(value, na.rm = TRUE),
    n_total  = n(),
    n_valid  = sum(!is.na(value)),
    .groups  = "drop"
  )


    


      

      

    


    


  






Example 9: .groups Argument





  


    
library(dplyr)

# .groups controls grouping state after summarise
# "drop" = fully ungrouped (safest, most common)
# "drop_last" = drops the last grouping level (default)
# "keep" = keeps all groups

result <- mtcars |>
  group_by(cyl, am) |>
  summarise(avg = mean(mpg), .groups = "drop")

cat("Is grouped after .groups='drop'?", is.grouped_df(result), "\n")
print(result)


    


      

      

    


    


  






Example 10: Cumulative and Rolling Within Groups





  


    
library(dplyr)

mtcars |>
  arrange(cyl, mpg) |>
  group_by(cyl) |>
  mutate(
    cumul_hp    = cumsum(hp),
    running_avg = round(cummean(mpg), 1)
  ) |>
  select(cyl, mpg, hp, cumul_hp, running_avg) |>
  slice_head(n = 3) |>
  ungroup()


    


      

      

    


    


  






summarise vs mutate After group_by








summarise()
mutate()






Output rows
1 per group
Same as input




Use case
Aggregate stats
Window functions




Example
mean(x) → one number per group
rank(x) → rank within group







Practice Exercises



Exercise 1: Sales Report



Summarise this data by region: total revenue, average quarterly revenue, and best quarter.





  


    
library(dplyr)

sales <- data.frame(
  region  = rep(c("East", "West", "South"), each = 4),
  quarter = rep(c("Q1","Q2","Q3","Q4"), 3),
  revenue = c(100,120,110,150, 200,180,220,250, 80,90,85,100)
)



    


      

      

    


    


  







Click to reveal solution

```r


  

    
library(dplyr)

sales <- data.frame(
  region  = rep(c("East", "West", "South"), each = 4),
  quarter = rep(c("Q1","Q2","Q3","Q4"), 3),
  revenue = c(100,120,110,150, 200,180,220,250, 80,90,85,100)
)

sales |>
  group_by(region) |>
  summarise(
    total    = sum(revenue),
    avg_qtr  = round(mean(revenue), 1),
    best_qtr = quarter[which.max(revenue)],
    best_rev = max(revenue),
    .groups  = "drop"
  ) |>
  arrange(desc(total))

    

      
      
    

    

  








Exercise 2: Percentage Within Group



Show each car's mpg as a percentage of its cylinder group's average.





  


    
library(dplyr)

# Add pct_of_avg column, show top 5 most efficient relative to group



    


      

      

    


    


  







Click to reveal solution

```r


  

    
library(dplyr)

mtcars |>
  group_by(cyl) |>
  mutate(pct_of_avg = round(mpg / mean(mpg) * 100, 1)) |>
  ungroup() |>
  select(mpg, cyl, pct_of_avg) |>
  arrange(desc(pct_of_avg)) |>
  head(5)

    

      
      
    

    

  




**Explanation:** `group_by() + mutate()` computes `mean(mpg)` per group but keeps all rows. Each car gets its group's mean used in the percentage calculation.





Exercise 3: Per-Species Summary



For each iris Species, compute count, mean Sepal.Length, and the coefficient of variation (sd/mean * 100).





  


    
library(dplyr)



    


      

      

    


    


  







Click to reveal solution

```r


  

    
library(dplyr)

iris |>
  group_by(Species) |>
  summarise(
    n      = n(),
    mean   = round(mean(Sepal.Length), 2),
    cv_pct = round(sd(Sepal.Length) / mean(Sepal.Length) * 100, 1),
    .groups = "drop"
  )

    

      
      
    

    

  




**Explanation:** Coefficient of variation (CV) = sd/mean × 100. It measures relative variability, making it comparable across groups with different means.





Summary







Function
Purpose






group_by(col)
Split data into groups




summarise(stat = fn(col))
One row per group




n()
Count rows in group




across(cols, fn)
Apply to multiple columns




count(col)
Shortcut: group + count




ungroup()
Remove grouping




.groups = "drop"
Ungroup after summarise







FAQ



What's the difference between summarise and summarize?



Nothing — they're aliases. Use whichever spelling you prefer.



Why do I get a ".groups" warning?



dplyr warns when you don't specify .groups after multi-column group_by(). Add .groups = "drop" for ungrouped output (most common) or "keep" to retain grouping.



Can I use custom functions in summarise?



Yes. Any function that takes a vector and returns a single value works: summarise(result = my_function(column)).



How do I summarise all numeric columns at once?



summarise(across(where(is.numeric), mean)) applies mean to every numeric column.



What's Next?




        


        
      


      

        

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