How to set a different linestyle for each hue group in a kdeplot / displot

Question

• How can each hue group of a seaborn.kdeplot, or seaborn.displot with kind='kde' be given a different linestyle?
  • Both axes-level and figure-level options will accept a str for linestyle/ls, which applies to all hue groups.

import seaborn as sns
import matplotlib.pyplot as plt

# load sample data
iris = sns.load_dataset("iris")

# convert data to long form
im = iris.melt(id_vars='species')

# axes-level plot works with 1 linestyle
fig = plt.figure(figsize=(6, 5))
p1 = sns.kdeplot(data=im, x='value', hue='variable', fill=True, ls='-.')

# figure-level plot works with 1 linestyle
p2 = sns.displot(kind='kde', data=im, x='value', hue='variable', fill=True, ls='-.')

• kdeplot

• displot

Reviewed Questions

• How to set the line style for each kdeplot in a jointgrid doesn't deal with hue groups.
• How to automatically alternate or cycle linestyles in seaborn regplot? doesn't deal with hue groups and iterates through each unique group.
• Dotted Seaborn distplot doesn't deal with hue groups and iterates through each unique group.
• change line style in seaborn facet grid hue_kws isn't a valid option.

Logotype

Answer 1

• With fill=True the object to update is in .collections
• With fill=False the object to update is in .lines
• Updating the legend is fairly simple:
  • handles = p.legend_.legendHandles[::-1] extracts and reverses the legend handles. They're reversed to update because they're in the opposite order in which the plot linestyle is updated
  • Note that figure-level plots extract the legend with ._legend, with the axes-level plots use .legend_.
• Tested in python 3.8.12, matplotlib 3.4.3, seaborn 0.11.2

kdeplot: axes-level

• Extract and iterate through .collections or .lines from the axes object and use .set_linestyle

fill=True

fig = plt.figure(figsize=(6, 5))
p = sns.kdeplot(data=im, x='value', hue='variable', fill=True)

lss = [':', '--', '-.', '-']

handles = p.legend_.legendHandles[::-1]

for line, ls, handle in zip(p.collections, lss, handles):
    line.set_linestyle(ls)
    handle.set_ls(ls)

fill=False

fig = plt.figure(figsize=(6, 5))
p = sns.kdeplot(data=im, x='value', hue='variable')

lss = [':', '--', '-.', '-']

handles = p.legend_.legendHandles[::-1]

for line, ls, handle in zip(p.lines, lss, handles):
    line.set_linestyle(ls)
    handle.set_ls(ls)

displot: figure-level

• Similar to the axes-level plot, but each axes must be iterated through
• The legend handles could be updated in for line, ls, handle in zip(ax.collections, lss, handles), but that applies the update for each subplot. Therefore, a separate loop is created to update the legend handles only once.

fill=True

g = sns.displot(kind='kde', data=im, col='variable', x='value', hue='species', fill=True, common_norm=False, facet_kws={'sharey': False})

axes = g.axes.flat

lss = [':', '--', '-.']

for ax in axes:
    for line, ls in zip(ax.collections, lss):
        line.set_linestyle(ls)
        
handles = g._legend.legendHandles[::-1]
for handle, ls in zip(handles, lss):
    handle.set_ls(ls)

fill=False

g = sns.displot(kind='kde', data=im, col='variable', x='value', hue='species', common_norm=False, facet_kws={'sharey': False})

axes = g.axes.flat

lss = [':', '--', '-.']

for ax in axes:
    for line, ls in zip(ax.lines, lss):
        line.set_linestyle(ls)
        
handles = g._legend.legendHandles[::-1]
for handle, ls in zip(handles, lss):
    handle.set_ls(ls)

Logotype