Half violin plot in matplotlib

Question

Recently matplotlib has added a native support for violin plot. What I want to do is half-violin plot. I guess that it can be done changing the body the function is returning. Do you know how to plot half violin plot as in the example but using the new function from matplotlib?

Bedmate

seaborn does paired half violins, might be easier to start there: stanford.edu/~mwaskom/software/seaborn-dev/generated/… – Primm

The Stanford link in the comments appears to be dead. – Newsmonger

@PhilB Seaborn now has its doc pages at pydata: seaborn.pydata.org/generated/seaborn.violinplot.html – Usquebaugh

Answer 1

Here simplify modifying the patches, clipping the values on the right or left

data1 = (np.random.normal(0, 1, size=10000), np.random.normal(0, 2, size=10000))
data2 = (np.random.normal(1, 1, size=10000), np.random.normal(1, 2, size=10000))

fig, ax = plt.subplots(figsize=(18, 7))

v1 = ax.violinplot(data1, points=100, positions=np.arange(0, len(data1)),
               showmeans=False, showextrema=False, showmedians=False)
for b in v1['bodies']:
    # get the center
    m = np.mean(b.get_paths()[0].vertices[:, 0])
    # modify the paths to not go further right than the center
    b.get_paths()[0].vertices[:, 0] = np.clip(b.get_paths()[0].vertices[:, 0], -np.inf, m)
    b.set_color('r')

v2 = ax.violinplot(data2, points=100, positions=np.arange(0, len(data2)), 
               showmeans=False, showextrema=False, showmedians=False)

for b in v2['bodies']:
    # get the center
    m = np.mean(b.get_paths()[0].vertices[:, 0])
    # modify the paths to not go further left than the center
    b.get_paths()[0].vertices[:, 0] = np.clip(b.get_paths()[0].vertices[:, 0], m, np.inf)
    b.set_color('b')

ax.legend([v1['bodies'][0],v2['bodies'][0]],['data1', 'data2'])

Bedmate

with some comments, it may help... TY – Terra

For those using horizontal violin plots vert=False, to make this work, just swap out .vertices[:, 0] with .vertices[:, 1]. – Flogging

Answer 2

New in matplotlib 3.9

ax.violinplot now includes a side param to automatically plot only one half.

Now it's as simple as setting side="low" and side="high" respectively:

v1 = ax.violinplot(data1, showmeans=False, showextrema=False, showmedians=False,
                   side="low")  # `side` param requires matplotlib 3.9+
v2 = ax.violinplot(data2, showmeans=False, showextrema=False, showmedians=False,
                   side="high") # `side` param requires matplotlib 3.9+

Richardricharda

Answer 3

As of 3.9 you can now use the side parameter to set this:

# Fake data with reproducible random state.
np.random.seed(19680801)
data = np.random.normal(0, 8, size=100)

fig, ax = plt.subplots()

ax.violinplot(data, [0], showmeans=True, showextrema=True)
ax.violinplot(data, [1], showmeans=True, showextrema=True, side='low')
ax.violinplot(data, [2], showmeans=True, showextrema=True, side='high')

ax.set_title('Violin Sides Example')
ax.set_xticks([0, 1, 2], ['Default', 'side="low"', 'side="high"'])
ax.set_yticklabels([])

Lubricious