forcats fct_reorder() in R: Sort Factors by a Variable

The forcats fct_reorder() function reorders the levels of a factor by a summary of another variable, so plots display groups in a meaningful order instead of alphabetically.

fct_reorder(f, x)                      # order levels by median of x
fct_reorder(f, x, .fun = mean)         # order by the mean instead
fct_reorder(f, x, .desc = TRUE)        # high-to-low order
fct_reorder(f, x, na.rm = TRUE)        # drop NA before summarising
fct_reorder2(f, x, y)                  # order for line-chart legends
ggplot(df, aes(fct_reorder(g, v), v))  # sorted bars in one line

What fct_reorder() does in one sentence

fct_reorder() sorts factor levels by a number. You give it a factor and a numeric vector of the same length. It computes one summary value per level (the median by default), then reorders the levels from the smallest summary to the largest. The underlying data is untouched, only the level order changes.

This matters because factor level order controls plotting order. A bar chart, a boxplot, or a legend draws groups in level order, and the default level order is alphabetical. Alphabetical order is rarely the order a reader wants. fct_reorder() replaces it with a value-based order so the tallest bar, or the group with the highest median, lands where the eye expects it.

Syntax

fct_reorder() takes a factor, a numeric vector, and an optional summary function. Every other argument has a sensible default, so the two-argument form covers most real cases.

The full signature is:

fct_reorder(.f, .x, .fun = median, ..., .na_rm = NULL, .default = Inf, .desc = FALSE)

• .f is the factor (a character vector also works, it is coerced to a factor).
• .x is a numeric vector the same length as .f. Levels are sorted by a summary of this.
• .fun is the summary applied to .x within each level. The default is median.
• ... passes extra arguments straight to .fun, for example na.rm = TRUE.
• .na_rm controls whether levels with an NA summary are dropped.
• .default is the placeholder summary for levels that produce NA; Inf sends them last.
• .desc, when TRUE, sorts from largest to smallest instead.

fct_reorder() examples

1. Reorder factor levels by median

The two-argument call is the workhorse. Pass the factor and a numeric vector, and the levels come back sorted by group median.

Spray C has the lowest median insect count, so it becomes the first level. Spray B has the highest and moves last. The original count and spray columns are unchanged.

2. Order a ggplot boxplot

The most common use of fct_reorder() is inside aes(). Calling it directly on the mapping reorders the plot without creating an intermediate object.

The boxes now climb steadily from low median to high median, which reads far better than the alphabetical A-to-F default.

3. Reverse the order with .desc

Set .desc = TRUE to sort from high to low. This is handy when you want the largest group first, such as a ranked bar chart.

This is the exact reverse of the median order from example 1. You could also wrap the result in fct_rev(), but .desc is clearer.

4. Order by a custom summary function

Pass .fun to summarise by something other than the median. The mean, the maximum, or any function that turns a numeric vector into one number all work.

Notice that B and F swap places compared with the median order. Spray F has a higher mean but a lower median because a few large counts pull its average up.

fct_reorder() vs other forcats ordering functions

fct_reorder() is one of several level-ordering tools in forcats. Each sorts by a different rule, so the choice depends on what your chart should communicate.

Use fct_reorder() when a numeric variable defines the order. Use fct_infreq() when count itself is the order. Use fct_relevel() when one level is a baseline that must come first regardless of its value.

Common pitfalls

Pitfall 1: expecting the rows to move. fct_reorder() changes level order, not row order. The data frame stays in its original sequence.

If you need the rows themselves sorted, use dplyr::arrange() after reordering the factor.

Pitfall 2: missing values poison the summary. The default median returns NA for any group that contains an NA, and that level gets pushed to the end.

Pitfall 3: using fct_reorder() for a line chart legend. When several lines share a plot, you want the legend ordered by where the lines end, not by a single summary. That is the job of fct_reorder2(), which sorts by a second variable at the largest value of the first.

Try it yourself

Related forcats functions

After fct_reorder(), these forcats functions round out factor handling:

• fct_reorder2(): order levels for line chart legends by a second variable.
• fct_relevel(): move specific levels to the front by hand.
• fct_infreq(): order levels by frequency, most common first.
• fct_inorder(): order levels by first appearance in the data.
• fct_rev(): reverse the current level order.

The base R function reorder() does the same job as fct_reorder() and needs no extra package. Most users still prefer the forcats version because its arguments are explicit and it composes cleanly with the rest of the fct_ family. See the forcats reference for the full argument list.

FAQ

What does fct_reorder do in R?

fct_reorder() reorders the levels of a factor based on a summary of a second variable. For each level, it computes one number (the median by default), then sorts the levels from smallest to largest. It is mainly used so ggplot2 charts display groups in value order rather than alphabetical order. The data values themselves are not changed, only the order in which the factor stores its levels.

What is the difference between fct_reorder and fct_relevel?

fct_reorder() sorts levels automatically by a numeric variable, so you never name the levels yourself. fct_relevel() is manual: you list the levels in the order you want, or move a specific level to the front. Use fct_reorder() when data defines the order, and fct_relevel() when you need a fixed reference level first, such as a control group in a model.

How do I reorder a ggplot bar chart by value?

Wrap the categorical variable in fct_reorder() inside aes(). For a bar chart of a value v by group g, write aes(x = fct_reorder(g, v), y = v). The bars then climb from the smallest value to the largest. Add .desc = TRUE inside fct_reorder() if you want the tallest bar on the left instead.

Can fct_reorder handle missing values?

Yes, but you must ask for it. The default summary function median returns NA for any group that contains an NA, and that level is sent to the end. Pass na.rm = TRUE through the ... argument, as in fct_reorder(f, x, na.rm = TRUE), so the summary ignores missing values and the level is ranked by its real data.