lets_plot.geom_violin(mapping=None, *, data=None, stat=None, position=None, show_legend=None, sampling=None, tooltips=None, orientation=None, show_half=None, draw_quantiles=None, scale=None, trim=None, tails_cutoff=None, kernel=None, bw=None, adjust=None, n=None, fs_max=None, **other_args)

A violin plot is a mirrored density plot with an additional grouping as for a boxplot.


Set of aesthetic mappings created by aes() function. Aesthetic mappings describe the way that variables in the data are mapped to plot “aesthetics”.

datadict or DataFrame or polars.DataFrame

The data to be displayed in this layer. If None, the default, the data is inherited from the plot data as specified in the call to ggplot.

statstr, default=’ydensity’

The statistical transformation to use on the data for this layer, as a string.

positionstr or FeatureSpec

Position adjustment, either as a string (‘identity’, ‘stack’, ‘dodge’, …), or the result of a call to a position adjustment function.

show_legendbool, default=True

False - do not show legend for this layer.


Result of the call to the sampling_xxx() function. To prevent any sampling for this layer pass value “none” (string “none”).


Result of the call to the layer_tooltips() function. Specify appearance, style and content.

orientationstr, default=’x’

Specify the axis that the layer’s stat and geom should run along. Possible values: ‘x’, ‘y’.

show_halffloat, default=0

If -1, only half of each violin is drawn. If 1, another half is drawn. If 0, violins look as usual.

draw_quantileslist of float

Draw horizontal lines at the given quantiles of the density estimate.

scale{‘area’, ‘count’, ‘width’}, default=’area’

If ‘area’, all violins have the same area. If ‘count’, areas are scaled proportionally to the number of observations. If ‘width’, all violins have the same maximum width.

trimbool, default=True

Trim the tails of the violins to the range of the data.

tails_cutofffloat, default=3.0

Extend domain of each violin on tails_cutoff * bw if trim=False.

kernelstr, default=’gaussian’

The kernel we use to calculate the density function. Choose among ‘gaussian’, ‘cosine’, ‘optcosine’, ‘rectangular’ (or ‘uniform’), ‘triangular’, ‘biweight’ (or ‘quartic’), ‘epanechikov’ (or ‘parabolic’).

bwstr or float

The method (or exact value) of bandwidth. Either a string (choose among ‘nrd0’ and ‘nrd’), or a float.


Adjust the value of bandwidth by multiplying it. Change how smooth the frequency curve is.

nint, default=512

The number of sampled points for plotting the function.

fs_maxint, default=500

Maximum size of data to use density computation with ‘full scan’. For bigger data, less accurate but more efficient density computation is applied.


Other arguments passed on to the layer. These are often aesthetics settings used to set an aesthetic to a fixed value, like color=’red’, fill=’blue’, size=3 or shape=21. They may also be parameters to the paired geom/stat.


Geom object specification.


Computed variables:

  • ..violinwidth.. : density scaled for the violin plot, according to area, counts or to a constant maximum width (mapped by default).

  • ..density.. : density estimate.

  • ..count.. : density * number of points.

  • ..scaled.. : density estimate, scaled to maximum of 1.

geom_violin() understands the following aesthetics mappings:

  • x : x-axis coordinates.

  • y : y-axis coordinates.

  • alpha : transparency level of a layer. Accept values between 0 and 1.

  • color (colour) : color of the geometry lines. Can be continuous or discrete. For continuous value this will be a color gradient between two colors.

  • fill : color of geometry filling.

  • size : lines width.

  • linetype : type of the line of border. Codes and names: 0 = ‘blank’, 1 = ‘solid’, 2 = ‘dashed’, 3 = ‘dotted’, 4 = ‘dotdash’, 5 = ‘longdash’, 6 = ‘twodash’.

  • weight : used by ‘ydensity’ stat to compute weighted density.


1import numpy as np
2from lets_plot import *
4n = 100
6x = np.random.choice(['a', 'b', 'c'], size=n)
7y = np.random.normal(size=n)
8ggplot({'x': x, 'y': y}, aes(x='x', y='y')) + \
9    geom_violin()

1import numpy as np
2from lets_plot import *
4n = 100
6x = np.random.choice(['a', 'b', 'b', 'c'], size=n)
7y = np.random.normal(size=n)
8ggplot({'x': x, 'y': y}, aes('x', 'y')) + \
9    geom_violin(scale='count', draw_quantiles=[.25, .5, .75])

 1import numpy as np
 2from lets_plot import *
 4n = 3
 6x = ['a'] * n + ['b'] * n + ['c'] * n
 7y = 3 * list(range(n))
 8vw = np.random.uniform(size=3*n)
 9ggplot({'x': x, 'y': y, 'vw': vw}, aes('x', 'y')) + \
10    geom_violin(aes(violinwidth='vw', fill='x'), stat='identity')

 1import numpy as np
 2import pandas as pd
 3from lets_plot import *
 5n, m = 100, 5
 7df = pd.DataFrame({'x%s' % i: np.random.normal(size=n) \
 8                   for i in range(1, m + 1)})
 9ggplot(df.melt(), aes('variable', 'value')) + \
10    geom_violin(aes(color='variable', fill='variable'), \
11                size=2, alpha=.5, scale='width') + \
12    geom_boxplot(aes(fill='variable'), width=.2)

 1import numpy as np
 2from lets_plot import *
 4n = 100
 6x = np.random.choice(["a", "b", "c", "d"], size=n)
 7y1 = np.random.normal(size=n)
 8y2 = np.random.normal(size=n)
 9ggplot({'x': x, 'y1': y1, 'y2': y2}) + \
10    geom_violin(aes('x', 'y1'), show_half=-1, \
11                trim=False, fill='#ffffb2') + \
12    geom_violin(aes('x', 'y2'), show_half=1, \
13                trim=False, fill='#74c476')