beeswarm (version 0.4.0)

beeswarm: Bee swarm plot


Create a bee swarm plot. A bee swarm plot is a one-dimensional scatter plot similar to stripchart, but with various methods to separate coincident points such that each point is visible. Also, beeswarm introduces additional features unavailable in stripchart, such as the ability to control the color and plotting character of each point.


beeswarm(x, …)

# S3 method for formula beeswarm(formula, data = NULL, subset, na.action = NULL, pwpch = NULL, pwcol = NULL, pwbg = NULL, pwcex = NULL, dlab, glab, …)

# S3 method for default beeswarm(x, method = c("swarm", "compactswarm", "center", "hex", "square"), vertical = TRUE, horizontal = !vertical, cex = 1, spacing = 1, breaks = NULL, labels, at = NULL, corral = c("none", "gutter", "wrap", "random", "omit"), corralWidth, side = 0L, priority = c("ascending", "descending", "density", "random", "none"), fast = TRUE, pch = par("pch"), col = par("col"), bg = NA, pwpch = NULL, pwcol = NULL, pwbg = NULL, pwcex = NULL, do.plot = TRUE, add = FALSE, axes = TRUE, log = FALSE, xlim = NULL, ylim = NULL, dlim = NULL, glim = NULL, xlab = NULL, ylab = NULL, dlab = "", glab = "", …)



A formula, such as y ~ grp, where y is a numeric vector of data values to be split into groups according to the grouping variable grp (usually a factor).


A data.frame (or list) from which the variables in formula should be taken.


An optional vector specifying a subset of observations to be used.


A function which indicates what should happen when the data contain NAs. The default is to quietly ignore missing values in either the response or the group.


A numeric vector, or a data frame or list of numeric vectors, each of which is plotted as an individual swarm.


Method for arranging points (see Details).

vertical, horizontal

Orientation of the plot. horizontal takes precedence if both are specified.


Size of points relative to the default given by par("cex"). Unlike other plotting functions, this must be a single value. (But see also the pwcex argument)


Relative spacing between points.


Breakpoints (optional). If NULL, breakpoints are chosen automatically. If NA, bins are not used (similar to stripchart with method = "stack").


Labels for each group. Recycled if necessary. By default, these are inferred from the data.


Numeric vector giving the locations where the swarms should be drawn; defaults to 1:n where n is the number of groups.


Method to adjust points that would be placed outside their own group region (see Details).


Width of the "corral" in user coordinates. If missing, a sensible value will be chosen.


Direction to perform jittering: 0: both directions; 1: to the right or upwards; -1: to the left or downwards.


Order used to perform point layout when method is "swarm" or "compactswarm"; ignored otherwise (see Details).


Use compiled version of algorithm? This option is ignored for all methods except "swarm" and "compactswarm".

pch, col, bg

Plotting characters and colors, specified by group. Recycled if necessary (see Details).

pwpch, pwcol, pwbg, pwcex

“Point-wise” plotting characteristics, specified for each data point (see Details).


Draw a plot?


Add to an existing plot?


Draw axes and box?


Use a logarithmic scale on the data axis?

xlim, ylim

Limits of the plot.

dlim, glim

An alternative way to specify limits (see Details).

xlab, ylab

Axis labels.

dlab, glab

An alternative way to specify axis labels (see Details).

Further arguments passed to plot.


A data frame with plotting information, invisibly.


Several methods for placing the points are available; each method uses a different algorithm to avoid overlapping points.

The default method, swarm, places points in increasing order. If a point would overlap an existing point, it is shifted sideways (along the group axis) by a minimal amount sufficient to avoid overlap. breaks is ignored.

The other three methods first discretize the values along the data axis, in order to create more efficient packing: square places the points on a square grid, whereas hex uses a hexagonal grid. center uses a square grid to produce a symmetric swarm. By default, the number of breakpoints for discretization is determined by a combination of the available plotting area and the plotting character size. The discretization of the data can be explicitly controlled using breaks. If breaks is set to NA, the data will not be grouped into intervals; this may be a sensible option if the data is already discrete.

In contrast to most other plotting functions, changing the size of the graphics device will often change the position of the points.

The plotting characters and colors can be controlled in two ways. First, the arguments pch, col and bg can specify plotting characters and colors in the same way as stripchart and boxplot: in short, the arguments apply to each group as a whole (and are recycled if necessary).

Alternatively, the “point-wise” characteristics of each individual data point can be controlled using pwpch, pwcol, and pwbg, which override pch, col and bg if these are also specified. Likewise, pwcex controls the size of each point relative to the default (which may be adjusted by cex). Notably, the point layout algorithm is applied without considering the point-wise arguments; thus setting pwcex larger than 1 will usually result in partially overlapping points. These arguments can be specified as a list or vector. If supplied using the formula method, the arguments can be specified as part of the formula interface; i.e. they are affected by data and subset.

The dlab and glab labels may be used instead of xlab and ylab if those are not specified. dlab applies to the continuous data axis (the Y axis unless horizontal is TRUE); glab to the group axis. Likewise, dlim and glim can be used to specify limits of the axes instead of xlim or ylim.

This function is intended to be mostly compatible with calls to stripchart or boxplot. Thus, code that works with these functions should work with beeswarm with minimal modification.

By default, swarms from different groups are not prevented from overlapping. Thus, large data sets, or data sets with uneven distributions, may produce somewhat unpleasing beeswarms. If this is a problem, consider reducing cex. Another approach is to control runaway points (those that would be plotted outside a region allotted to each group) with the corral argument: The default, "none", does not control runaway points. "gutter" collects runaway points along the boundary between groups. "wrap" implements periodic boundaries. "random" places runaway points randomly in the region. "omit" omits runaway points. See Examples below.

When using the "swarm" method, priority controls the order in which the points are placed; this generally has a noticeable effect on the resulting appearance. "ascending" gives the "traditional" beeswarm plot in which the points are placed in an ascending order. "descending" is the opposite. "density" prioritizes points with higher local density. "random" places points in a random order. "none" places points in the order provided.

Whereas the "swarm" method places points in a predetermined order, the "compactswarm" method uses a greedy strategy to determine which point will be placed next. This often leads to a more tightly-packed layout. The strategy is very simple: on each iteration, a point that can be placed as close as possible to the non-data axis is chosen and placed. If there are two or more equally good points, priority is used to break ties.

See Also

stripchart, boxplot


Run this code
  ## One of the examples from 'stripchart'
  beeswarm(decrease ~ treatment,
    data = OrchardSprays, log = TRUE, 
    pch = 16, col = rainbow(8))

  ## One of the examples from 'boxplot', with a beeswarm overlay
   boxplot(len ~ dose, data = ToothGrowth,
              main = "Guinea Pigs' Tooth Growth",
              xlab = "Vitamin C dose mg",
              ylab = "Tooth length")  
   beeswarm(len ~ dose, data = ToothGrowth, col = 2, add = TRUE)

  ## Compare the 5 methods
  op <- par(mfrow = c(2,3))
  for (m in c("swarm", "compactswarm", "center", "hex", "square")) {
    beeswarm(len ~ dose, data = ToothGrowth, method = m, main = m)

  ## Demonstrate the use of 'pwcol'
  beeswarm(time_survival ~ ER, data = breast,
    pch = 16, pwcol = 1 + as.numeric(event_survival),
    xlab = "", ylab = "Follow-up time (months)",
    labels = c("ER neg", "ER pos"))
  legend("topright", legend = c("Yes", "No"),
    title = "Censored", pch = 16, col = 1:2)

  ## The list interface
  distributions <- list(runif = runif(200, min = -3, max = 3), 
                        rnorm = rnorm(200),
                        rlnorm = rlnorm(200, sdlog = 0.5))
  beeswarm(distributions, col = 2:4)

  ## Demonstrate 'pwcol' with the list interface 
  myCol <- lapply(distributions, function(x) cut(x, breaks = quantile(x), labels = FALSE))
  beeswarm(distributions, pch = 16, pwcol = myCol)
  legend("bottomright", legend = 1:4, pch = 16, col = 1:4, title = "Quartile")

  ## Demonstrate the 'corral' methods
  par(mfrow = c(2,3))
  beeswarm(distributions, col = 2:4, 
    main = 'corral = "none" (default)')
  beeswarm(distributions, col = 2:4, corral = "gutter", 
    main = 'corral = "gutter"')
  beeswarm(distributions, col = 2:4, corral = "wrap", 
    main = 'corral = "wrap"')
  beeswarm(distributions, col = 2:4, corral = "random", 
    main = 'corral = "random"')
  beeswarm(distributions, col = 2:4, corral = "omit", 
    main = 'corral = "omit"')  
  ## Demonstrate 'side' and 'priority'
  par(mfrow = c(2,3))
  beeswarm(distributions, col = 2:4, 
    main = 'Default')
  beeswarm(distributions, col = 2:4, side = -1, 
    main = 'side = -1')
  beeswarm(distributions, col = 2:4, side = 1, 
    main = 'side = 1')
  beeswarm(distributions, col = 2:4, priority = "descending", 
    main = 'priority = "descending"')
  beeswarm(distributions, col = 2:4, priority = "random", 
    main = 'priority = "random"')  
  beeswarm(distributions, col = 2:4, priority = "density", 
    main = 'priority = "density"')  

  ## Demonstrate 'side' and 'priority' for compact method
  par(mfrow = c(2,3))
  beeswarm(distributions, col = 2:4, method = "compactswarm",
    main = 'Default')
  beeswarm(distributions, col = 2:4, method = "compactswarm", side = -1, 
    main = 'side = -1')
  beeswarm(distributions, col = 2:4, method = "compactswarm", side = 1, 
    main = 'side = 1')
  beeswarm(distributions, col = 2:4, method = "compactswarm",
    priority = "descending",  main = 'priority = "descending"')
  beeswarm(distributions, col = 2:4, method = "compactswarm",
    priority = "random",  main = 'priority = "random"')  
  beeswarm(distributions, col = 2:4, method = "compactswarm",
    priority = "density",  main = 'priority = "density"')  

  ## Demonstrate pwcol, pwpch, pwbg, and pwcex
  beeswarm(mpg ~ cyl, data = mtcars, cex = 3, 
    pwcol = gear, pwbg = am + 1, pwpch = gear + 18, pwcex = hp / 335)
# }

Run the code above in your browser using DataCamp Workspace