ggplot2 geom_density() in R: Density Plots With Examples

The geom_density() function in ggplot2 draws a smooth kernel density estimate of a continuous variable. It is the smoothed version of a histogram and the cleanest way to compare distribution SHAPES across groups.

ggplot(df, aes(x)) + geom_density()                           # basic curve
ggplot(df, aes(x, fill = group)) + geom_density(alpha = 0.4)  # multiple, transparent
ggplot(df, aes(x, color = group)) + geom_density()            # color outlines only
ggplot(df, aes(x)) + geom_density(adjust = 0.5)               # tighter (more detail)
ggplot(df, aes(x)) + geom_density(adjust = 2)                 # smoother
ggplot(df, aes(x)) + geom_density(fill = "steelblue", alpha = 0.5)
ggplot(df, aes(x, fill = group)) + geom_density(position = "stack")  # stacked

What geom_density() does in one sentence

geom_density() draws a smooth curve estimating where data values cluster. Tall regions of the curve correspond to data-dense ranges; short regions to sparse ranges. The total area under the curve equals 1 by default (probability density).

Compared to geom_histogram(), density is smoother (no bin choice) and easier to overlay across groups, but it can hide local features and depends on a bandwidth choice.

Syntax

geom_density() requires aes(x). The y axis is computed (density by default).

The full signature:

geom_density(mapping = NULL, data = NULL, stat = "density", position = "identity",
             ..., na.rm = FALSE, orientation = NA, show.legend = NA, inherit.aes = TRUE)

Six common patterns

1. Basic density curve

The simplest case: density curve with semi-transparent fill.

2. Compare densities across groups

fill = drv colors the curves. alpha = 0.4 makes them transparent so overlapping regions are visible.

3. Outlines only (no fill)

For very busy plots with many groups, color-coded LINES (no fill) are cleaner than filled areas.

4. Tighter or smoother bandwidth

adjust = 0.5 reveals more local structure. Useful when the data may be multimodal. adjust = 2 would smooth it heavily.

5. Density + histogram together

aes(y = after_stat(density)) puts the histogram on the SAME y scale as the density curve so they overlay cleanly. Without this, they would be on different scales (count vs density).

6. Stacked densities

position = "stack" stacks the densities so the total area at each x equals the marginal density. Less common than overlay but useful for showing decomposition.

geom_density() vs geom_histogram() vs geom_violin()

Three views of distribution; choose by audience and N.

When to use which:

• Use geom_density() for shape comparison across 2 to 5 groups.
• Use geom_histogram() when bin counts are part of the story.
• Use geom_violin() when comparing 6+ groups along a categorical x.

Common pitfalls

Pitfall 1: density implies continuous data. If your variable is integer counts (0, 1, 2, ...), the density curve "smears" between values and creates artifacts. Use geom_bar() instead.

Pitfall 2: bandwidth choice changes interpretation. Default adjust = 1 may oversmooth. Always check what adjust = 0.5 and adjust = 2 look like; pick whichever reveals real structure without obvious noise.

Pitfall 3: bounded data (positive only, or 0 to 1) gets distorted at edges. Density curves can extend below 0 even when data cannot. Solutions: trim the curve manually, transform the data first (log, logit), or use a bounded density estimator (bkde from KernSmooth).

Try it yourself

Related ggplot2 functions

After mastering geom_density(), look at:

• geom_histogram(): discrete bin counts; pair with density for full picture
• geom_violin(): mirrored density for grouped comparison along a categorical x
• geom_density_2d(): 2D contour density for two continuous variables
• geom_freqpoly(): line version of histogram (alternative to density)
• stat_ecdf(): empirical cumulative distribution function
• ggridges::geom_density_ridges(): stacked densities for many groups

For density estimation with bounded support (e.g., values >= 0), the bkde function from KernSmooth plus a geom_line() works.

FAQ

How do I plot density curves for multiple groups in ggplot2?

Map a categorical variable to fill or color: aes(x = value, fill = group) then geom_density(alpha = 0.4). The alpha makes overlapping curves visible. Use color = group for line-only versions.

What is the difference between geom_density and geom_histogram?

geom_density() is a smooth curve (kernel density estimate). geom_histogram() is binned counts. Density is smoother and easier to compare across groups; histogram preserves bin counts and is more honest about the data's discreteness.

How do I overlay a density curve on a histogram in ggplot2?

Set y to density on the histogram first: geom_histogram(aes(y = after_stat(density))) + geom_density(). Without rescaling y to density, the curve and histogram are on different scales (count vs density) and overlay incorrectly.

What does the adjust argument do in geom_density?

adjust is a bandwidth multiplier. adjust = 1 (default) uses the standard bandwidth. adjust = 0.5 halves it (tighter, more detail). adjust = 2 doubles it (smoother). Try several values when distribution shape matters.

How do I add a vertical line at the mean to a density plot?

Add geom_vline(xintercept = mean(your_data), linetype = "dashed"). Inside a ggplot pipe with grouped data, use stat_summary(fun = mean, geom = "vline") or compute group means separately.