How can I combine hue and style groups in a Seaborn legend?

Question

I'm doing a Seaborn lineplot for longitudinal data which is grouped by "Subscale" using hue and by "Item" using style.

Here is my code (I hope this is understandable also without data):

ax = sns.lineplot(data = df, x = 'Week', y = 'Value', style = 'Item', hue = 'Subscale', palette = 'colorblind', markers = True)
plt.legend(bbox_to_anchor = (1.03, 1.02), fontsize = 10)

Whicht gets me this plot:

What I want is to combine the sublegends so that it only shows the legend for "Item" but the items are colored according to "Subscale", kind of like this:

I've failed to create this, so if any of you could help, this would be highly appreciated! Thanks :)

user_12046409 · Answer

If I understand correctly, all items of a certain type have the same subscale. So, you already have (or can create) a dictionary that maps an item type to the corresponding subscale.

Seaborn creates following labels for the legend:

'Subscale' for a subtitle
each of the subscales
'Item' for a second subtitle
each of the items

Each label corresponds to a "handle" identifying how the marker looks like.

The following code:

finds the index of 'Item' to be able to split the arrays of labels and handles
extracts the colors of the 'Subscale's
applies these colors to the item markers
only uses the items for the legend

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np

# first create some random data similar to the description
np.random.seed(123)
items = ['01', '04', '05', '06', '07', '10', '11', '13']
N = len(items)
subscale_dict = {'01': 'A', '04': 'C', '05': 'C', '06': 'C', '07': 'A', '10': 'B', '11': 'B', '13': 'A'}

df = pd.DataFrame({'Week': np.tile(np.arange(7), N),
                   'Value': np.random.rand(7 * N),
                   'Item': np.repeat(items, 7)})
df['Subscale'] = df['Item'].apply(lambda i: subscale_dict[i])
df['Subscale'] = pd.Categorical(df['Subscale'])  # creates a fixed order

# create the line plot as before
ax = sns.lineplot(data=df, x='Week', y='Value', style='Item', hue='Subscale', palette='colorblind', markers=True)

# create a dictionary mapping the subscales to their color
handles, labels = ax.get_legend_handles_labels()
index_item_title = labels.index('Item')
color_dict = {label: handle.get_color()
              for handle, label in zip(handles[1:index_item_title], labels[1:index_item_title])}

# loop through the items, assign color via the subscale of the item idem
for handle, label in zip(handles[index_item_title + 1:], labels[index_item_title + 1:]):
    handle.set_color(color_dict[subscale_dict[ label]])

# create a legend only using the items
ax.legend(handles[index_item_title + 1:], labels[index_item_title + 1:], title='Item',
          bbox_to_anchor=(1.03, 1.02), fontsize=10)

plt.tight_layout()
plt.show()

How can I combine hue and style groups in a Seaborn legend?

1 Answers