Two sided grouped barplots with Python seaborn

Question

I'm trying to draw a two-sided graph similar to the population pyramid explained here and here. The problem is that I have categorical variables (male/female) that I want to group together:

import pandas as pd
import seaborn as sns

# data
data = {'species': ['X', 'X', 'Y', 'Y', 'Z', 'Z', 'X', 'X', 'Y', 'Y', 'Z', 'Z'],
        'sex': ['male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female'], 
        'mass (g)': [4000, 3500, 3800, 3200, 5500, 4900, 2500, 2100, 2400, 2000, 4200, 3800],
        'age': ['adult', 'adult', 'adult', 'adult', 'adult', 'adult', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile']}
df = pd.DataFrame(data)

# convert juvenile mass to negative
df.loc[df.age.eq('juvenile'), 'mass (g)'] = df['mass (g)'].mul(-1)

# plot
sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(10,5))
sns.barplot(data=df, x='mass (g)', y='species', hue='sex', ci=False, orient='horizontal', dodge=True)
ax.yaxis.tick_right()
ax.yaxis.set_label_position("right")
plt.show()

The figure below is what I'm aiming for. Different color bars are for male/female sex. Different species X, Y, Z are in separate categorical groups. The bars on the right side of the figure show mass of adults for each species.

I sketched in red the bars on the left side to show mass of juveniles for each species. How do I plot this? I can't find anything useful in the seaborn docs or on SO.

Scott Boston · Answer 1 · 2021-11-23T22:18:43.623

Try something like this:

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

# data
data = {'species': ['X', 'X', 'Y', 'Y', 'Z', 'Z', 'X', 'X', 'Y', 'Y', 'Z', 'Z'],
        'sex': ['male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female'], 
        'mass (g)': [4000, 3500, 3800, 3200, 5500, 4900, 2500, 2100, 2400, 2000, 4200, 3800],
        'age': ['adult', 'adult', 'adult', 'adult', 'adult', 'adult', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile']}
df = pd.DataFrame(data)

# convert juvenile mass to negative
df.loc[df.age.eq('juvenile'), 'mass (g)'] = df['mass (g)'].mul(-1)

# plot
sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(10,5))
df_reshape = df.set_index(['species','sex','age']).unstack(['age','sex'])['mass (g)']
df_reshape.loc[:, 'adult'].plot.barh(ax=ax)
df_reshape.loc[:, 'juvenile'].plot.barh(legend=False, ax=ax)
plt.show()

Output:

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

# data
data = {'species': ['X', 'X', 'Y', 'Y', 'Z', 'Z', 'X', 'X', 'Y', 'Y', 'Z', 'Z'],
        'sex': ['male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female'], 
        'mass (g)': [4000, 3500, 3800, 3200, 5500, 4900, 2500, 2100, 2400, 2000, 4200, 3800],
        'age': ['adult', 'adult', 'adult', 'adult', 'adult', 'adult', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile']}
df = pd.DataFrame(data)

# convert juvenile mass to negative
df.loc[df.age.eq('juvenile'), 'mass (g)'] = df['mass (g)'].mul(-1)

# plot
sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(10,5))
df_reshape = df.set_index(['species','sex','age']).unstack(['age','sex'])['mass (g)']
df_reshape.loc[:, ['adult']].plot.barh(ax=ax, edgecolor='k')
df_reshape.loc[:, ['juvenile']].plot.barh(ax=ax, label='Juvenile', color=['navy','red'], alpha=.6, edgecolor='k', hatch='/')
plt.show()

Output:

JohanC · Accepted Answer · 2021-11-24T00:32:50.240

If you mix positive and negative values, by default seaborn's barplot will average them out.

You could draw two barplots back to back and reverse the left one:

from matplotlib import pyplot as plt
import seaborn as sns
import pandas as pd

data = {'species': ['X', 'X', 'Y', 'Y', 'Z', 'Z', 'X', 'X', 'Y', 'Y', 'Z', 'Z'],
        'sex': ['male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female'],
        'mass (g)': [4000, 3500, 3800, 3200, 5500, 4900, 2500, 2100, 2400, 2000, 4200, 3800],
        'age': ['adult', 'adult', 'adult', 'adult', 'adult', 'adult', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile']}
df = pd.DataFrame(data)
df['sex'] = pd.Categorical(df['sex'])  # make hue column categorical, forcing a fixed order

sns.set_theme(style='whitegrid')
fig, (ax1, ax2) = plt.subplots(ncols=2, figsize=(10, 5), sharey=True, gridspec_kw={'wspace': 0})


# draw adult subplot at the right
sns.barplot(data=df[df['age'] == 'adult'], x='mass (g)', y='species', hue='sex',
            ci=False, orient='horizontal', dodge=True, ax=ax2)
ax2.yaxis.set_label_position('right')
ax2.tick_params(axis='y', labelright=True, right=True)
ax2.set_title('  '+'adult', loc='left')
ax2.legend_.remove()  # remove the legend; the legend will be in ax1

# draw juvenile subplot at the left
sns.barplot(data=df[df['age'] == 'juvenile'], x='mass (g)', y='species', hue='sex',
            ci=False, orient='horizontal', dodge=True, ax=ax1)

# optionally use the same scale left and right
xmax = max(ax1.get_xlim()[1], ax2.get_xlim()[1])
ax1.set_xlim(xmax=xmax)
ax2.set_xlim(xmax=xmax)

ax1.invert_xaxis()  # reverse the direction
ax1.tick_params(labelleft=False, left=False)
ax1.set_ylabel('')
ax1.set_title('juvenile'+'  ', loc='right')

plt.tight_layout()
plt.show()

An interesting feature of seaborn's barplots is that it will also do the work of averaging out the values given a dataframe with a row for each individual (and calculate a confidence interval).

On downside of this approach is that the bars on the two handside are not bound to the same scale, which can be misleading. — Quang Hoang, Nov 23 '21 at 22:52
@johanC for some reason, the code under `# optionally use the same scale left and right` doesn't have any effect (for my actual code, not the MWE in my post). ie. the bars on left and right side of graph aren't aligned properly — Medulla Oblongata, Nov 24 '21 at 00:22
yes, I basically copied your code and used my own dataset for it - all the commands are unchanged. It looks like the bars on the left graph have larger widths than on the right - can you enforce a bar width for both `ax1` and `ax2`? — Medulla Oblongata, Nov 24 '21 at 00:42
Could you [edit](https://stackoverflow.com/posts/70088196/edit) your question and add an image of that plot? Maybe there are more hue values available in one vs the other set? Maybe the species column should also explicitly be made categorical? (`df['species'] = pd.Categorical(df['species'])`) — JohanC, Nov 24 '21 at 01:09
I might post a separate question for this, thanks for your help though. — Medulla Oblongata, Nov 24 '21 at 17:28

score 1 · Answer 3 · answered Nov 23 '21 at 22:34

I just used `pivot to shape the data correctly

import pandas as pd
import seaborn as sns

# data
data = {'species': ['X', 'X', 'Y', 'Y', 'Z', 'Z', 'X', 'X', 'Y', 'Y', 'Z', 'Z'],
        'sex': ['male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female', 'male', 'female'], 
        'mass (g)': [4000, 3500, 3800, 3200, 5500, 4900, 2500, 2100, 2400, 2000, 4200, 3800],
        'age': ['adult', 'adult', 'adult', 'adult', 'adult', 'adult', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile', 'juvenile']}
df = pd.DataFrame(data)

# convert juvenile mass to negative
df.loc[df.age.eq('juvenile'), 'mass (g)'] = df['mass (g)'].mul(-1)

# pivot data
df=df.pivot(columns=['age'], index=['species', 'sex'], values=['mass (g)']).reset_index()
df = df.set_index(['species', 'sex'])['mass (g)'].reset_index()

# plot
sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(10,5))
sns.barplot(data=df, x='adult', y='species', hue='sex', ci=False, orient='horizontal', dodge=True)
sns.barplot(data=df, x='juvenile', y='species', hue='sex', ci=False, orient='horizontal', dodge=True)

Two sided grouped barplots with Python seaborn

3 Answers3

Related