Matplotlib / Seaborn Countplot with different Categories in one Plot

Question

I have two series with different lengths and amount of variables and want to plot how often each variable (Name) occurs per series. I want a grey countplot for series 1 and a red countplot for series 2, and I want them to be shown on top of each other. However, since series 2 is missing 'Nancy' it is also cutting series 1 count of 'Nancy'. How do i get a full overlay of the two series inkluding a bar for Nancy?

import matplotlib.pyplot as plt
import seaborn as sns

ser1 = pd.Series( ['tom','tom','bob','bob','nancy'])
ser2 = pd.Series( ['tom','bob'])

fig = plt.figure()
sns.countplot(x=ser1, color='grey')
sns.countplot(x=ser2, color='red')
plt.show()

Edit: Changing to the following will cause problems again. How do I make Matplotlib recognize that the two series have the same categorical values that are being counted?

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

ser1 = pd.Series( ['tom','tom','bob','bob','nancy','zulu'])
ser2 = pd.Series( ['tom','nancy'])

ser1 = ser1.astype('category')
ser2 = ser2.astype('category')

fig = plt.figure()
ax = sns.countplot(x=ser2, color='red', zorder=2)
sns.countplot(x=ser1, color='grey')

plt.show()

ImportanceOfBeingErnest · Accepted Answer · 2018-03-11 13:57:48Z

5

You may store the setting for the first plot and restore them after having plotted the second plot.

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

ser1 = pd.Series( ['tom','tom','bob','bob','nancy','zulu'])
ser2 = pd.Series( ['tom','bob'])

fig = plt.figure()
ax = sns.countplot(x=ser1, color='grey')
ticks = ax.get_xticks()
ticklabels = ax.get_xticklabels()
lim = ax.get_xlim()

sns.countplot(x=ser2, color='red')
ax.set_xlim(lim)
ax.set_xticks(ticks)
ax.set_xticklabels(ticklabels)
plt.show()

The other option could be to plot the second plot first but set the zorder to a higher value, such that those bars appear in front of the later plot.

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

ser1 = pd.Series( ['tom','tom','bob','bob','nancy','zulu'])
ser2 = pd.Series( ['tom','bob'])

fig = plt.figure()
ax = sns.countplot(x=ser2, color='red', zorder=2)
sns.countplot(x=ser1, color='grey')

plt.show()

In the more general case you need to use the order arument.

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

ser1 = pd.Series( ['tom','tom','bob','bob','nancy', 'nancy' ,'zulu'])
ser2 = pd.Series( ['tom','nancy'])
order = ser1.append(ser2).unique()

fig = plt.figure()
ax = sns.countplot(x=ser2, color='red', order=order, zorder=2)
sns.countplot(x=ser1, color='grey', order=order)

plt.show()

In case you would rather use matplotlib's categoricals to create the plot, this would look as follows:

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

ser1 = pd.Series( ['tom','tom','bob','bob','nancy', 'nancy' ,'zulu'])
ser2 = pd.Series( ['tom','nancy'])

u1, counts1 = np.unique(ser1.values, return_counts=True)
u2, counts2 = np.unique(ser2.values, return_counts=True)

fig, ax = plt.subplots()
ax.bar(u1,counts1, color='grey')
ax.bar(u2,counts2, color='red')

plt.show()

edited Mar 11, 2018 at 13:57

answered Mar 10, 2018 at 0:00

ImportanceOfBeingErnest

342k61 gold badges739 silver badges771 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Rolo Over a year ago

Thanks for the fast solution! I'm using the zorder trick now.

Rolo Over a year ago

Unfortunatley, matplotlib does not recognize the categories. If i change my series to: ser1 = pd.Series( ['tom','tom','bob','bob','nancy','zulu']) ser2 = pd.Series( ['tom','nancy']) The results will be incorrect. Even when the series dtype('category')

ImportanceOfBeingErnest Over a year ago

I updated the answer with zulu, it works as expected.

Rolo Over a year ago

If you change ser2 to ser2 = pd.Series( ['tom','nancy']) the categories won't match anymore.

ImportanceOfBeingErnest Over a year ago

I updated the answer. Note that you are using seaborn countplot here, so this does not use matplotlib categorical axes but numerical axes that seaborn creates internally. I also added a solution with matplotlib for comparisson.

Collectives™ on Stack Overflow

Matplotlib / Seaborn Countplot with different Categories in one Plot

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related