Gallery: Distributions

Distributions describe how values spread out. plotit’s distribution marks reveal shape, tails, and relationships — from histograms and densities to ECDFs, QQ plots, and bivariate binning.

Histograms

Histogram with 30 bins

mark_histogram() bins a continuous variable and counts observations per bin. The bins argument controls the resolution.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_histogram(bins = 30)

Histograms by group

Filling by a categorical variable and lowering alpha lets several histograms overlap on one panel. A small subset of mpg classes keeps the comparison readable.

mpg2 <- ggplot2::mpg[ggplot2::mpg$class %in% c("compact", "suv", "pickup", "subcompact"), ]
mpg2 |>
  plotit(encode(x = hwy, fill = class)) |>
  mark_histogram(bins = 20, alpha = 0.5)

Densities

Density with custom bandwidth

mark_density() estimates a smooth kernel density. The bandwidth bw (passed through ...) trades smoothness against detail — smaller values follow the data more closely.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_density(bw = 0.15, alpha = 0.6)

Density by group

Coloured density curves overlay several groups for direct shape comparison.

iris |>
  plotit(encode(x = Sepal.Length, colour = Species)) |>
  mark_density()

Histogram and density overlay

Rescale the histogram with after_stat(density) so the bars and the density curve share the same y scale.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_histogram(mapping = encode(y = ggplot2::after_stat(density)), bins = 30, alpha = 0.5) |>
  mark_density(bw = 0.15)

Cumulative Distributions

Empirical CDF

mark_ecdf() draws the empirical cumulative distribution function as a step with no binning to choose — every point is represented exactly.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_ecdf()

Quantile-Quantile Plots

QQ plot with normal reference

mark_qq() plots sample quantiles against the theoretical quantiles of a distribution; mark_qq_line() adds the fitted reference line. Deviations from the line signal departures from normality.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_qq() |>
  mark_qq_line()

QQ plot with exponential reference

Pass distribution to check against any theoretical distribution — here "exp" for data simulated from an exponential. Both the points and the line must use the same distribution.

set.seed(42)
rex <- data.frame(value = rexp(200))
rex |>
  plotit(encode(x = value)) |>
  mark_qq(distribution = "exp") |>
  mark_qq_line(distribution = "exp")

Bivariate Distributions

2D density contours

mark_density_2d() draws contour lines of a two-dimensional kernel density — here the classic faithful eruptions/waiting relationship.

faithful |>
  plotit(encode(x = eruptions, y = waiting)) |>
  mark_density_2d()

Filled 2D density

With filled = TRUE the contour bands are filled, giving a topographic view of where the density is highest.

faithful |>
  plotit(encode(x = eruptions, y = waiting)) |>
  mark_density_2d(filled = TRUE, bins = 10)

Bin 2D

mark_bin2d() tiles the plane into a rectangular grid of counts — a clean way to visualise overplotting in dense data.

set.seed(1)
dmid <- ggplot2::diamonds[ggplot2::diamonds$carat < 2, ]
dmid |>
  plotit(encode(x = carat, y = price)) |>
  mark_bin2d(bins = 20)

Hexbin

mark_hex() bins into hexagons instead of squares, which avoids the visual bias of axis-aligned rectangles.

dmid |>
  plotit(encode(x = carat, y = price)) |>
  mark_hex(bins = 30)

Distribution Details

Density with rug

mark_rug() adds a tick per observation along the axis, recovering the exact data positions that a smooth density can hide.

faithful |>
  plotit(encode(x = eruptions)) |>
  mark_density() |>
  mark_rug(sides = "b", color = "grey30")

Faceted histograms

split_wrap() repeats the histogram for each group in its own panel, avoiding the overlap of stacked or dodged bins.

ggplot2::mpg |>
  plotit(encode(x = hwy)) |>
  mark_histogram(bins = 20) |>
  split_wrap(drv, ncol = 3)

Mean with standard error

Summarise groups into means with error bars for the standard error of the mean. mark_errorbar() takes ymin/ymax alongside y; the point shows the mean.

se <- function(x) sd(x) / sqrt(length(x))
grpm <- data.frame(
  Species = levels(iris$Species),
  mean = tapply(iris$Sepal.Length, iris$Species, mean),
  sem = tapply(iris$Sepal.Length, iris$Species, se)
)
grpm |>
  plotit(encode(x = Species, y = mean, ymin = mean - sem, ymax = mean + sem)) |>
  mark_point(size = 3) |>
  mark_errorbar(width = 0.2)