dplyr group_by() & summarise() Exercises: 10 Aggregation Problems

Q: What is the difference between summarise() with group_by() and mutate() with group_by()?

summarise() collapses each group to one row. mutate() keeps every original row but runs the calculation per group. Use summarise() for group-level outputs (means, counts). Use mutate() to add a per-group value as a new column on the original data.

Q: Should I use count() or summarise(n = n())?

Use count() for plain frequencies — it is shorter. Use summarise(n = n(), ...) when you need counts alongside other summaries like means or sums. They produce the same count column either way.

Q: Can I group by an expression, not just a column?

Yes. group_by(cyl, hp_class = hp > 150) creates a new grouping variable on the fly from any expression. The computed variable appears as a column in the output and behaves like any other grouping variable.

Ten runnable exercises to master dplyr::group_by() and summarise() — from a single-column count to grouped percentages, across(), NA handling, and per-group ranking. Each problem has a worked solution you can run and tweak in your browser.

Introduction

Reading about group_by() and summarise() is quick. Using them fluently on messy data — with missing values, multiple grouping columns, and questions like "what share of the total does each group hold?" — takes practice. These ten exercises give you that practice.

The problems grow from simple to genuine. The first three ask for counts and means. The middle three bring in multi-column grouping, across(), and NA handling. The last four mix summarise() with filtering, percentages, the .groups argument, and per-group slicing. Every exercise states the task, hands you a starter block, and hides the worked solution behind a click-to-reveal.

Try writing your own answer first. Run it. Compare with the solution. Read the short explanation under the solution to check your reasoning. If you are new to these verbs, skim the parent tutorial on dplyr group_by() + summarise() before starting.

All code on this page runs in one shared R session, so variables you create in one block are available in the next. Use distinct names like my_result in your exercise code so you do not overwrite tutorial variables from earlier blocks.

Quick Reference

Skim this cheat sheet before you start. It lists the functions you will use in the ten exercises.

Task	Function	Example
Split into groups	`group_by()`	`group_by(df, cyl)`
Count rows per group	`n()`	`summarise(n = n())`
Count shortcut	`count()`	`count(df, cyl)`
Mean with NA safe	`mean(x, na.rm = TRUE)`	`summarise(m = mean(x, na.rm = TRUE))`
Apply to many columns	`across()`	`summarise(across(where(is.numeric), mean))`
Drop grouping after summarise	`.groups = "drop"`	`summarise(..., .groups = "drop")`
Remove grouping explicitly	`ungroup()`	`ungroup(df)`
Keep top-k per group	`slice_max()`	`slice_max(df, mass, n = 2)`
Group share as percent	custom formula	`100 * x / sum(x)`

Tip

Load dplyr once. Every later block reuses it. The first code block loads dplyr and previews the two datasets. Because all blocks share a single R session, later exercises can call group_by() and summarise() without reloading.

# Load dplyr and preview the datasets we will use library(dplyr) # mtcars: 32 rows, 11 numeric columns, 1973-74 car models head(mtcars, 3) #> mpg cyl disp hp drat wt qsec vs am gear carb #> Mazda RX4 21.0 6 160 110 3.90 2.620 16.46 0 1 4 4 #> Mazda RX4 Wag 21.0 6 160 110 3.90 2.875 17.02 0 1 4 4 #> Datsun 710 22.8 4 108 93 3.85 2.320 18.61 1 1 4 1 # starwars: 87 rows, several NA values (great for na.rm practice) starwars |> select(name, height, mass, species, homeworld) |> head(3) #> # A tibble: 3 × 5 #> name height mass species homeworld #> <chr> <int> <dbl> <chr> <chr> #> 1 Luke Skywalker 172 77 Human Tatooine #> 2 C-3PO 167 75 Droid Tatooine #> 3 R2-D2 96 32 Droid Naboo

Both datasets are ready. mtcars has clean numeric columns, so aggregation is straightforward. starwars has NA values in height, mass, and species — that is the point. You will use it for na.rm practice.

Easy (1-3): Counts and Simple Means

Start here if you have barely used group_by() before. Each exercise uses one or two verbs only.

Exercise 1: Count cars per cylinder

Count how many cars in mtcars have 4, 6, and 8 cylinders. Save the result to my_counts. The output should have two columns: cyl and n.

# Exercise 1: count rows per cyl group # Hint: group_by(cyl) then summarise(n = n()) # Write your code below:

Click to reveal solution

my_counts <- mtcars |> group_by(cyl) |> summarise(n = n(), .groups = "drop") print(my_counts) #> # A tibble: 3 × 2 #> cyl n #> <dbl> <int> #> 1 4 11 #> 2 6 7 #> 3 8 14

Explanation: group_by(cyl) splits the data into three pieces — one per unique cylinder value. n() counts the rows inside each piece. .groups = "drop" drops the grouping after summarising, so my_counts is a plain tibble. Without it you would see a friendly message from dplyr explaining which groups remain.

Exercise 2: Average mpg per cylinder

Compute the average mpg for each cyl group in mtcars. Round to one decimal. Save to my_mpg. The column should be named avg_mpg.

# Exercise 2: mean mpg per cyl # Hint: use summarise(avg_mpg = round(mean(mpg), 1)) # Write your code below:

Click to reveal solution

my_mpg <- mtcars |> group_by(cyl) |> summarise(avg_mpg = round(mean(mpg), 1), .groups = "drop") print(my_mpg) #> # A tibble: 3 × 2 #> cyl avg_mpg #> <dbl> <dbl> #> 1 4 26.7 #> 2 6 19.7 #> 3 8 15.1

Explanation: Four-cylinder cars average 26.7 mpg while eight-cylinder cars average 15.1 mpg. mean(mpg) runs once per group because group_by(cyl) already split the rows. round(..., 1) formats the output. No na.rm here — mtcars has no NA values.

Exercise 3: Use count() as a shortcut

Repeat Exercise 1, but use count() instead of group_by() + summarise(n = n()). Count cars per gear value. Save to my_gears.

# Exercise 3: use count() shortcut # Hint: count(mtcars, gear) does the same thing in one call # Write your code below:

Click to reveal solution

my_gears <- mtcars |> count(gear) print(my_gears) #> gear n #> 1 3 15 #> 2 4 12 #> 3 5 5

Explanation: count() is a shortcut. It does the exact same thing as group_by(gear) |> summarise(n = n(), .groups = "drop") — but shorter. Most dplyr users reach for count() when they only need a row count per group. Add sort = TRUE to order the result from largest to smallest.

[TIP]

Prefer count() for plain frequencies, summarise() for everything else. count() is read-at-a-glance for "how many per group". Once you add other summaries (means, sums), switch back to group_by() + summarise().

Medium (4-6): Multi-Column Groups, across(), and NA Handling

These three exercises mix two ideas at once. Take them slowly.

Exercise 4: Group by two columns

Group mtcars by both cyl and am (automatic/manual). Compute the count and mean mpg per combination. Save to my_combo. Include a .groups = "drop" to keep the output ungrouped.

# Exercise 4: group by two columns # Hint: group_by(cyl, am) then summarise with n() and mean() # Write your code below:

Click to reveal solution

my_combo <- mtcars |> group_by(cyl, am) |> summarise( n = n(), avg_mpg = round(mean(mpg), 1), .groups = "drop" ) print(my_combo) #> # A tibble: 6 × 4 #> cyl am n avg_mpg #> <dbl> <dbl> <int> <dbl> #> 1 4 0 3 22.9 #> 2 4 1 8 28.1 #> 3 6 0 4 19.1 #> 4 6 1 3 20.6 #> 5 8 0 12 15.0 #> 6 8 1 2 15.4

Explanation: group_by(cyl, am) creates six groups — every unique (cyl, am) combination with at least one row. Manual 4-cylinder cars average 28.1 mpg; automatic 8-cylinder cars average 15.0 mpg. The .groups = "drop" flag removes the grouping, so the result behaves like a flat tibble for any downstream work.

Exercise 5: Summarise many columns with across()

Use the iris dataset. Group by Species and compute the mean of every numeric column in one call. Save to my_iris. The result should have four columns: Species plus the four numeric means.

# Exercise 5: summarise all numeric columns per species # Hint: summarise(across(where(is.numeric), mean)) # Write your code below:

Click to reveal solution

my_iris <- iris |> group_by(Species) |> summarise(across(where(is.numeric), mean), .groups = "drop") print(my_iris) #> # A tibble: 3 × 5 #> Species Sepal.Length Sepal.Width Petal.Length Petal.Width #> <fct> <dbl> <dbl> <dbl> <dbl> #> 1 setosa 5.01 3.43 1.46 0.246 #> 2 versicolor 5.94 2.77 4.26 1.33 #> 3 virginica 6.59 2.97 5.55 2.03

Explanation: across(where(is.numeric), mean) says: for every column that is numeric, apply mean(). This is the modern replacement for the old summarise_if() and summarise_at() helpers. If you needed two summaries per column, use a named list: across(where(is.numeric), list(mean = mean, sd = sd)).

[KEY INSIGHT]

across() scales with your data, not your typing. Hand-writing summarise(mean_a = mean(a), mean_b = mean(b), mean_c = mean(c), ...) breaks once you have twenty columns. With across() the code is the same for four columns or four hundred.

Exercise 6: Handle NA values with starwars

Use starwars. Group by species and compute mean height and mean mass. Drop any NA inputs from the means. Save to my_species. Sort the result by mean_height descending and keep only the top 5 rows.

# Exercise 6: mean height and mass per species, NA-safe # Hint: mean(x, na.rm = TRUE); then arrange(desc(mean_height)) and head(5) # Write your code below:

Click to reveal solution

my_species <- starwars |> group_by(species) |> summarise( mean_height = mean(height, na.rm = TRUE), mean_mass = mean(mass, na.rm = TRUE), .groups = "drop" ) |> arrange(desc(mean_height)) |> head(5) print(my_species) #> # A tibble: 5 × 3 #> species mean_height mean_mass #> <chr> <dbl> <dbl> #> 1 Quermian 264 NaN #> 2 Kaminoan 221 88 #> 3 Wookiee 231 124 #> 4 Yoda's species 66 17 #> 5 Kaleesh 216 159

Explanation: na.rm = TRUE tells mean() to ignore missing values before averaging. Without it, any group containing one NA would return NA for that group. Notice the NaN in mean_mass for Quermian — that group had zero non-NA mass values, so the mean of an empty set is undefined. Real analysis almost always needs na.rm = TRUE.

[WARNING]

Forgetting na.rm silently poisons every group with missing data. The result still runs. It is just full of NA values you did not expect. Always check for NA first with summary(df) before summarising.

Hard (7-10): Filtering Groups, Percentages, .groups, and Ranking

The final four exercises combine two or more dplyr ideas. Take your time.

Exercise 7: Keep only groups with at least N rows

From starwars, compute the mean height per species, but keep only species with at least 2 characters in the dataset. Use na.rm = TRUE. Save to my_big_species.

# Exercise 7: filter groups by size # Hint: summarise n = n() AND mean_height together, then filter(n >= 2) # Write your code below:

Click to reveal solution

my_big_species <- starwars |> group_by(species) |> summarise( n = n(), mean_height = mean(height, na.rm = TRUE), .groups = "drop" ) |> filter(n >= 2) |> arrange(desc(n)) print(my_big_species) #> # A tibble: 9 × 3 #> species n mean_height #> <chr> <int> <dbl> #> 1 Human 35 177. #> 2 Droid 6 140 #> 3 Gungan 3 209. #> 4 Wookiee 2 231 #> ...

Explanation: The summarise() call produces one row per species with both n and mean_height. Then filter(n >= 2) keeps only rows where the group had at least two characters. This is the standard pattern for "ignore small or noisy groups" in analysis. Because .groups = "drop" was used, filter() works on a plain tibble.

Exercise 8: Group share as a percentage

For mtcars, compute each gear group's share of total mpg as a percentage. The output should have three columns: gear, sum_mpg, and pct_of_total. Save to my_share. The pct_of_total column should sum to 100.

# Exercise 8: group share as percent of total # Hint: first summarise sum_mpg per gear, then mutate pct = 100 * sum_mpg / sum(sum_mpg) # Write your code below:

Click to reveal solution

my_share <- mtcars |> group_by(gear) |> summarise(sum_mpg = sum(mpg), .groups = "drop") |> mutate(pct_of_total = round(100 * sum_mpg / sum(sum_mpg), 1)) print(my_share) #> # A tibble: 3 × 3 #> gear sum_mpg pct_of_total #> <dbl> <dbl> <dbl> #> 1 3 294. 45.9 #> 2 4 294. 45.9 #> 3 5 52.3 8.2

Explanation: The trick is .groups = "drop". After dropping, the mutate() call sees one row per gear and computes sum(sum_mpg) across all three rows — that is the grand total. Without dropping, mutate() would run inside each group and divide by itself, giving 100% for every row. Always ungroup before computing shares.

[KEY INSIGHT]

Shares need a two-step pipeline: summarise first, then mutate with an ungrouped sum. The first step aggregates rows to groups; the second step compares each group to the global total. Keeping the grouping active during mutate() is a top-3 source of wrong percentage results.

Exercise 9: Compare .groups = "drop" vs .groups = "keep"

Run two near-identical pipelines on mtcars: group by cyl and am, then summarise n = n(). In the first, use .groups = "drop". In the second, use .groups = "keep". After each, call group_vars() to see which grouping remains. Save the results to my_drop and my_keep.

# Exercise 9: contrast .groups values # Hint: run the same pipeline twice with different .groups; then call group_vars(result) # Write your code below:

Click to reveal solution

my_drop <- mtcars |> group_by(cyl, am) |> summarise(n = n(), .groups = "drop") my_keep <- mtcars |> group_by(cyl, am) |> summarise(n = n(), .groups = "keep") group_vars(my_drop) #> character(0) group_vars(my_keep) #> [1] "cyl" "am"

Explanation: group_vars() shows the active grouping columns. "drop" removes all grouping after summarise, so my_drop is ungrouped (empty character vector). "keep" holds on to every grouping variable — both cyl and am. The other options are "drop_last" (removes only the rightmost grouping — this is dplyr's default when you do not specify) and "rowwise" (rarely needed). Use "drop" as a safe default unless you know you need the grouping later.

[NOTE]

dplyr's default .groups behaviour prints a message; "drop" silences it. When you leave .groups out, dplyr picks "drop_last" and prints a note: `summarise() has grouped output by 'cyl'. That message is helpful once, annoying forever. Set .groups` explicitly to silence it.

Exercise 10: Top-k per group using slice_max

From starwars, find the two heaviest characters per homeworld. Only consider rows where mass and homeworld are not NA. Save to my_top2. Sort the result by homeworld, then by mass descending within each homeworld.

# Exercise 10: top-2 per group # Hint: filter NA rows, group_by(homeworld), slice_max(mass, n = 2) # Write your code below:

Click to reveal solution

my_top2 <- starwars |> filter(!is.na(mass), !is.na(homeworld)) |> group_by(homeworld) |> slice_max(mass, n = 2, with_ties = FALSE) |> arrange(homeworld, desc(mass)) |> select(name, homeworld, mass) print(my_top2, n = 8) #> # A tibble: ... × 3 #> # Groups: homeworld [...] #> name homeworld mass #> <chr> <chr> <dbl> #> 1 Jabba Desilijic Tiure Nal Hutta 1358 #> 2 IG-88 null 140 #> 3 Bossk Trandosha 113 #> 4 Dexter Jettster Ojom 102 #> ...

Explanation: slice_max(mass, n = 2) keeps the top 2 rows per group based on mass. Because group_by(homeworld) is still active, "top 2" means top 2 per homeworld, not top 2 overall. with_ties = FALSE stops slice_max from keeping extra rows when two characters have the same mass. Always filter out NA values before slicing — slice_max() treats NA as infinity-like by default and you will get surprises.

[TIP]

slice_max is the modern replacement for top_n(). The old top_n() still works but is superseded. Use slice_max() and slice_min() going forward — they have cleaner behaviour around ties and NAs.

Common Mistakes and How to Fix Them

Five mistakes that trip up almost everyone learning these verbs.

Mistake 1: Forgetting na.rm with missing data

Missing values poison the whole group summary.

The mean of a vector with even one NA is NA. Unless you tell R to ignore NAs, every group containing a missing value returns NA.

# Wrong: silent NA pollution starwars |> group_by(species) |> summarise(m = mean(height), .groups = "drop") |> head(3) #> # A tibble: 3 × 2 #> species m #> <chr> <dbl> #> 1 Aleena 79 #> 2 Besalisk 198 #> 3 Cerean 198 #> ...

# Correct: na.rm = TRUE starwars |> group_by(species) |> summarise(m = mean(height, na.rm = TRUE), .groups = "drop") |> head(3)

Always check summary(your_data) first. If any column has NAs, plan for na.rm = TRUE on every summariser that touches it.

Mistake 2: Leaving groups attached after summarise

A grouped tibble behaves differently in downstream verbs. Percentages, joins, and filters all change.

# Wrong: still grouped, so mutate computes per-group, not overall bad <- mtcars |> group_by(cyl) |> summarise(total_hp = sum(hp)) bad |> mutate(pct = 100 * total_hp / sum(total_hp)) #> per-group sum divides by itself → every row 100%

# Correct: drop groups, then mutate sees the whole tibble good <- mtcars |> group_by(cyl) |> summarise(total_hp = sum(hp), .groups = "drop") good |> mutate(pct = 100 * total_hp / sum(total_hp))

Use .groups = "drop" or call ungroup() explicitly when the grouped step is finished.

Mistake 3: Using mean() on non-numeric columns

across(everything(), mean) crashes if any column is a character or factor.

# Wrong: iris has a factor column iris |> group_by(Species) |> summarise(across(everything(), mean)) #> Error: `across()` argument is not numeric

# Correct: restrict to numeric columns iris |> group_by(Species) |> summarise(across(where(is.numeric), mean), .groups = "drop")

Use where(is.numeric) inside across() whenever you are not 100% sure every column is numeric.

Mistake 4: Not specifying .groups and being surprised by messages

dplyr prints a helpful message when you leave .groups off, but readers often mistake it for an error.

# Noisy (prints message but works) mtcars |> group_by(cyl, am) |> summarise(n = n()) #> `summarise()` has grouped output by 'cyl'. You can override using the `.groups` argument.

# Quiet: state your intent mtcars |> group_by(cyl, am) |> summarise(n = n(), .groups = "drop")

Always set .groups explicitly. It documents your intent and silences the message.

Summary

Ten verbs and patterns to keep in mind.

Pattern	Use when
`group_by() + summarise(n = n())`	Counting rows per group
`count()`	Shortcut when you only need counts
`summarise(across(where(is.numeric), mean))`	Many columns at once
`mean(x, na.rm = TRUE)`	Data has missing values
`summarise(..., .groups = "drop")`	Default to drop grouping
`filter(n >= k)` after summarise	Exclude tiny or noisy groups
Two-step share: summarise then mutate	Percentages of a total
`group_by() + slice_max(x, n = k)`	Top-k rows per group
`ungroup()` before join/mutate	Avoid silent per-group behaviour

FAQ

Why does summarise() print a .groups message?

dplyr tells you which grouping it kept after summarising. By default it drops the last grouping variable. Set .groups = "drop", "keep", "drop_last", or "rowwise" to make your choice explicit and silence the message.

What is the difference between summarise() with group_by() and mutate() with group_by()?

summarise() collapses each group to one row. mutate() keeps every original row but runs the calculation per group. Use summarise() for group-level outputs (means, counts). Use mutate() to add a per-group value as a new column on the original data.

Should I use count() or summarise(n = n())?

Use count() for plain frequencies — it is shorter. Use summarise(n = n(), ...) when you need counts alongside other summaries like means or sums. They produce the same count column either way.

Can I group by an expression, not just a column?

Yes. group_by(cyl, hp_class = hp > 150) creates a new grouping variable on the fly from any expression. The computed variable appears as a column in the output and behaves like any other grouping variable.

References

dplyr — summarise() reference. tidyverse.org
dplyr — group_by() reference. tidyverse.org
dplyr — Grouped data vignette. tidyverse.org
Wickham H. & Grolemund G. — R for Data Science, 2nd edition, Chapter 4 (Data transformation). r4ds.hadley.nz
dplyr — across() reference. tidyverse.org
dplyr — slice_max() reference. tidyverse.org

What's Next?

dplyr group_by() + summarise(): Aggregate Data by Group (10 Examples) — the parent tutorial behind these exercises.
dplyr filter() & select() Exercises: 12 Practice Problems — companion exercise set for row filtering and column picking.
dplyr Exercises — broader dplyr practice spanning the full verb family.

r‑statistics.co by Selva Prabhakaran

dplyr group_by() & summarise() Exercises: 10 Aggregation Problems

Introduction

Quick Reference

Easy (1-3): Counts and Simple Means

Exercise 1: Count cars per cylinder

Exercise 2: Average mpg per cylinder

Exercise 3: Use count() as a shortcut

Medium (4-6): Multi-Column Groups, across(), and NA Handling

Exercise 4: Group by two columns

Exercise 5: Summarise many columns with across()

Exercise 6: Handle NA values with starwars

Hard (7-10): Filtering Groups, Percentages, .groups, and Ranking

Exercise 7: Keep only groups with at least N rows

Exercise 8: Group share as a percentage

Exercise 9: Compare .groups = "drop" vs .groups = "keep"

Exercise 10: Top-k per group using slice_max

Common Mistakes and How to Fix Them

Mistake 1: Forgetting na.rm with missing data

Mistake 2: Leaving groups attached after summarise

Mistake 3: Using mean() on non-numeric columns

Mistake 4: Not specifying .groups and being surprised by messages

Summary

FAQ

Why does summarise() print a .groups message?

What is the difference between summarise() with group_by() and mutate() with group_by()?

Should I use count() or summarise(n = n())?

Can I group by an expression, not just a column?

References

What's Next?

On this page