Seaborn workaround for hue barplot

Question

I have the following DataFrame on a Jupyter notebook which plots using seaborn a barplot:

data = {'day_index': [0, 1, 2, 3, 4, 5, 6],
        'avg_duration': [708.852242, 676.7021900000001, 684.572677, 708.92534, 781.767476, 1626.575057, 1729.155673],
        'trips': [114586, 120936, 118882, 117868, 108036, 43740, 37508]}

df = pd.DataFrame(data)

daysOfWeek = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']

plt.figure(figsize=(16,10));
sns.set_style('ticks')
ax = sns.barplot(data=df, \
                 x='day_index', \
                 y='avg_duration', \
                 hue='trips', \
                 palette=sns.color_palette("Reds_d", n_colors=7, desat=1))

ax.set_xlabel("Week Days", fontsize=18, alpha=0.8)
ax.set_ylabel("Duration (seconds)", fontsize=18, alpha=0.8)
ax.set_title("Week's average Trip Duration", fontsize=24)
ax.set_xticklabels(daysOfWeek, fontsize=16)
ax.legend(fontsize=15)
sns.despine()
plt.show()

Plot A:

As it can be seen the bars do not match the x_ticklabels and are very thin.
This is all fixed if I remove the hue='trips' part, it's a known seaborn issue. Although It's very important to show the amount of trips in the visualization so: is there a way around seaborn (maybe with matplotlib directly) to add a hue attribute?

Note that you can use `sns.barplot(..., hue='trips', dodge=False)` to have bars with normal widths. By default, `dodge=True` to prevent multiple bars with the same x from overlapping. — JohanC, Jan 05 '23 at 20:21

score 5 · Answer 1 · edited May 23 '17 at 12:17

The hue argument probably only makes sense to introduce a new dimension to the plot, not to show another quantity on the same dimension.

It's probably best to plot the bars without the hue argument (it's quite misleading to call it hue actually) and simply colorize the bars according to the values in the "trips" column.

This is shown also in this question: Seaborn Barplot - Displaying Values.

The code here would look like:

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

di = np.arange(0,7)
avg  = np.array([708.852242,676.702190,684.572677,708.925340,781.767476,
                 1626.575057,1729.155673])
trips = np.array([114586,120936,118882,117868,108036,43740,37508])
df = pd.DataFrame(np.c_[di, avg, trips], columns=["day_index","avg_duration", "trips"])

daysOfWeek = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', \
'Friday', 'Saturday', 'Sunday']

plt.figure(figsize=(10,7));
sns.set_style('ticks')
v  = df.trips.values
colors=plt.cm.viridis((v-v.min())/(v.max()-v.min()))
ax = sns.barplot(data=df, x='day_index',   y='avg_duration', palette=colors)

for index, row in df.iterrows():
    ax.text(row.day_index,row.avg_duration, row.trips, color='black', ha="center")

ax.set_xlabel("Week Days", fontsize=16, alpha=0.8)
ax.set_ylabel("Duration (seconds)", fontsize=16, alpha=0.8)
ax.set_title("Week's average Trip Duration", fontsize=18)
ax.set_xticklabels(daysOfWeek, fontsize=14)
ax.legend(fontsize=15)
sns.despine()
plt.show()

MaxU - stand with Ukraine · Accepted Answer · 2017-04-17T11:55:37.700

I think you don't need to specify hue parameter in this case:

In [136]: ax = sns.barplot(data=dfGroupedAgg, \
     ...:                  x='day_index', \
     ...:                  y='avg_duration', \
     ...:                  palette=sns.color_palette("Reds_d", n_colors=7, desat=1))
     ...:

you can add amount of trips as annotations:

def autolabel(rects, labels=None, height_factor=1.05):
    for i, rect in enumerate(rects):
        height = rect.get_height()
        if labels is not None:
            try:
                label = labels[i]
            except (TypeError, KeyError):
                label = ' '
        else:
            label = '%d' % int(height)
        ax.text(rect.get_x() + rect.get_width()/2., height_factor*height,
                '{}'.format(label),
                ha='center', va='bottom')

autolabel(ax.patches, labels=df.trips, height_factor=1.02)

Trenton McKinney · Answer 3 · 2022-07-15T22:48:05.913

Build the legend from a color map

Remove hue. As already noted, the bars will not be centered when using this parameter, because they are placed according to the number of hue levels, and there are 7 levels in this case.
Using the palette parameter instead of hue, places the bars directly over the ticks.
This option requires "manually" associating 'trips' with the colors and creating the legend.
- patches uses Patch to create each item in the legend. (e.g. the rectangle, associated with color and name).

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from matplotlib.patches import Patch

daysOfWeek = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']

# specify the colors
colors = sns.color_palette('Reds_d', n_colors=len(df))

# create the plot
plt.figure(figsize=(16,10))
ax = sns.barplot(data=df, x='day_index', y='avg_duration', palette=colors)

# plot cosmetics
ax.set_xlabel("Week Days", fontsize=18, alpha=0.8)
ax.set_ylabel("Average Duration (seconds)", fontsize=18, alpha=0.8)
ax.set_title("Week's average Trip Duration", fontsize=24)
ax.set_xticklabels(daysOfWeek, fontsize=16)
sns.despine()

# setup the legend

# map names to colors
cmap = dict(zip(df.trips, colors))

# create the rectangles for the legend
patches = [Patch(color=v, label=k) for k, v in cmap.items()]

# add the legend
ax.legend(title='Number of Trips', handles=patches, bbox_to_anchor=(1.04, 0.5), loc='center left', borderaxespad=0, fontsize=15)

Alternatively, add bar labels with matplotlib.pyplot.bar_label, as shown with How to add value labels on a bar chart. This option is available from matplotlib v3.4.0.
- labels=df.trips can be used to pass custom labels directly from the dataframe column.

plt.figure(figsize=(16,10))
ax = sns.barplot(data=df, x='day_index', y='avg_duration', palette=colors)

# plot cosmetics
ax.set_xlabel("Week Days", fontsize=18, alpha=0.8)
ax.set_ylabel("Average Duration (seconds)", fontsize=18, alpha=0.8)
ax.set_title("Week's average Trip Duration", fontsize=24)
ax.set_xticklabels(daysOfWeek, fontsize=16)
sns.despine()

# add bar labels
_ = ax.bar_label(ax.containers[0], labels=df.trips, padding=1)

# add bar labels with customized text in a list comprehension
_ = ax.bar_label(ax.containers[0], labels=[f'Trips: {v}' for v in df.trips], padding=1)

score 1 · Answer 4 · answered Dec 02 '20 at 09:56

Here is the solution

ax = sns.barplot(data=df, \
                 x='day_index', \
                 y='avg_duration', \
                 hue='trips', \
                 dodge=False, \
                 palette=sns.color_palette("Reds_d", n_colors=7, desat=1))

Seaborn workaround for hue barplot

4 Answers4

Build the legend from a color map

Linked

Related