dplyr slice_max() in R: Top N Rows by Column Value

The slice_max() function in dplyr returns the rows with the LARGEST values of a specified column, optionally per group. It is the modern replacement for the deprecated top_n().

slice_max(df, mpg, n = 5)                   # 5 highest by mpg
slice_max(df, mpg, prop = 0.1)              # top 10%
slice_max(df, mpg, n = 3, by = cyl)         # top 3 per cyl
slice_max(df, mpg, n = 3, with_ties = FALSE)# strict 3 (no ties)
slice_max(df, mpg, n = -3)                  # all but top 3
arrange(df, desc(mpg)) |> slice_head(n = 5) # equivalent older form

What slice_max() does in one sentence

slice_max(.data, order_by, n) sorts by order_by descending and returns the top n rows. On a grouped tibble (or with by = g), it returns the top n per group.

This is the cleanest way to answer "top N by metric" questions. It supersedes the older top_n() (deprecated in dplyr 1.0).

Syntax

slice_max(.data, order_by, n = NULL, prop = NULL, by = NULL, with_ties = TRUE, na_rm = FALSE).

Five common patterns

1. Top n by a column

By default, ties are included (you may get more rows than n).

2. Top n per group

by = cyl scopes to each cyl group; one row per group.

3. Top fraction (prop)

10% of 32 = 3 rows (or more with ties).

4. Strict top n (no ties)

with_ties = FALSE returns exactly n rows; arbitrary among ties.

5. Latest record per group (timestamp)

The canonical "latest per group" idiom in modern dplyr.

slice_max() vs top_n() vs arrange + slice_head vs slice_min

Four ways to grab "top n by column" in R, with different ergonomics.

When to use which:

• slice_max for top-by-metric.
• slice_min for bottom-by-metric.
• arrange + slice_head only if you also need the full sorted intermediate state.
• Avoid top_n in new code.

A practical slice_max workflow

The "top N per group" pattern is the most common slice_max use case in real pipelines.

Common variations:

• Top 1 per group → slice_max(metric, n = 1, by = g) ("best of each")
• Top 5 by metric overall → slice_max(metric, n = 5) ("leaderboard")
• Highest-priced item per category → slice_max(price, n = 1, by = category)
• Most recent record per user → slice_max(timestamp, n = 1, by = user)

The pattern is so common it deserves its own one-liner. slice_max(df, col, n, by) IS that one-liner.

Common pitfalls

Pitfall 1: ties expand the result. slice_max(mpg, n = 3) returns 4 rows if two are tied at rank 3. Use with_ties = FALSE for strict n.

Pitfall 2: NAs sort to the bottom of descending order. slice_max excludes NAs in order_by only if na_rm = TRUE. Default behavior may surprise.

Try it yourself

Related slice functions

After mastering slice_max, look at:

• slice_min(): bottom n by column (mirror of slice_max)
• slice_head() / slice_tail(): first/last n by row order
• slice_sample(): random n rows
• slice(): specific row indexes
• arrange(): sort the entire frame
• top_n(): deprecated; do not use

For "lowest n by metric", slice_min(col, n) is the direct counterpart.

FAQ

What is the difference between slice_max and top_n in dplyr?

top_n() is deprecated since dplyr 1.0. slice_max() is the replacement: clearer name, supports prop, by, and with_ties arguments.

How does slice_max handle ties?

By default with_ties = TRUE includes all tied rows, which may return more than n rows. Set with_ties = FALSE for exactly n rows.

How do I get the top n per group?

Pass by = group_col (dplyr 1.1+): slice_max(df, col, n = 3, by = g). Or df |> group_by(g) |> slice_max(col, n = 3) |> ungroup().

What is the difference between slice_max and slice_head?

slice_max(col, n) sorts by COLUMN VALUE descending and takes top n. slice_head(n) takes the FIRST n rows IN CURRENT ORDER, ignoring values. Use slice_max when ranking is the criterion.

How does slice_max handle NA values?

NAs go to the bottom of descending order, so they are not picked unless n exceeds the count of non-NAs. Set na_rm = TRUE to exclude NAs explicitly.