Drawing a violin plot using Python and matplotlib

A violin plot combines the Kernel Density Estimation Plot and the Box plot. It displays the whole distribution along with the probability density function, median and mode information.The python matplotlib example draws violin plots with different bandwidth values applied.

Charts Updated 5 Aug 2019 2 min read

Overview

  • A violin plot combines two aspects of a distribution in a single visualization:
    • The features of a Box Plot: Median, Interquartile Distance
    • The Probability Density Function
  • In a violin plot, the Probability Density Function-PDF of the distribution is tilted side wards and placed on both the sides of the box plot.
  • A violin plot clearly displays the multiple modes present in a multi-modal data.
  • Violin plots display the whole distribution. They do not display outliers separately as in case of Box plots.

Drawing a violin plot using Python and Matplotlib

  • To create a violin plot, import the matplotlib.pyplot module and call the method violinplot() function by passing the data as sequences.
  • The violin plot can be customized to display mean and median values.
  • Using Matplotlib both vertical and horizontal violin plots can be created through the parameter vert. By default the value of the vert parameter is True, which creates a vertical violin plot. The width of the violin can be controlled by the widths parameter.

Example

python
# Example python program to draw a violin plot using matplotlib

import matplotlib.pyplot as plot

import pandas as pd

import numpy as np

# Customized method to draw violin plot in a matplotlib subplot

def drawViolinPlot(axis, xlabel, xticks, xticklabels, ylabel, bandwidth, title):

    # Customize x-axis

    axis.set_xlabel(xlabel);

    axis.set_xticks(xticks);

    axis.set_xticklabels(xticklabels);

    # Customize y-axis

    axis.set_ylabel(ylabel);

    # Draw a violin plot for the distributions - Default bandwidth

    axis.violinplot(sequences, showmeans=True, showmedians=True, bw_method=bandwidth);

    axis.set_title(title);

# Distributions as python lists

triangularSeries = [0, 1, 3, 6, 10, 15, 21, 28, 36, 45];

pentagonalSeries = [1, 5, 12, 22, 35, 51, 70, 92, 117, 145];

# A tuple of distributions

sequences = (triangularSeries, pentagonalSeries);

figure, axis = plot.subplots(2, 1);

plot.subplots_adjust(hspace=1);

bandwidth = None;

drawViolinPlot(axis[0],

              "Number Series",

              np.arange(len(sequences)+1),

              ('', 'Triangular Series', 'Pentagonal Series'),

              "Specific number in a series", bandwidth,

              "Violin Plot - Default Bandwidth(i.e., method=scott)");

bandwidth = 0.3;

drawViolinPlot(axis[1],

              "Number Series",

              np.arange(len(sequences)+1),

              ('', 'Triangular Series', 'Pentagonal Series'),

              "Specific number in a series", bandwidth,

              "Violin Plot - Bandwidth=0.3(i.e., a scalar value)");

# Display violin plot

plot.show();

Output

Violin plot with bandwidth calculated using scott method

Violin plot with a scalar bandwidth