Customizing a seaborn countplot with additional categorical variable

Question

I am following the guidance from there to create a bar chart similar to the last example:

https://seaborn.pydata.org/generated/seaborn.countplot.html

however, I want to show only the survived =1 plot.

My data is very similar, and my chart looks like this (so I want to show only the right part of my chart with left =yes).

also, what values does the order parameter take? I wish I could sort them by count. It does not take desc or descending.

JohanC · Accepted Answer · 2022-01-21T14:14:24.640

1

You can use data=df[df['left']=='yes'] to just plot the right part. To order by count, you can use ... .value_counts().index for the order= parameter.

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

prob = np.random.rand(7) + 0.1
prob /= prob.sum()
df = pd.DataFrame({'department': np.random.choice(['helium', 'neon', 'argon', 'krypton', 'xenon', 'radon', 'oganesson'],
                                                  1000, p=prob),
                   'left': np.random.choice(['yes', 'no'], 1000)})
sns.set_style('white')
filter = df['left'] == 'yes'
g = sns.catplot(data=df[filter], kind='count', y='department', palette='mako_r',
            order=df[filter]['department'].value_counts(ascending=True).index)
for ax in g.axes.flat:
    ax.bar_label(ax.containers[0], fontsize=12)
    ax.margins(x=0.1)
plt.tight_layout()
plt.show()

edited Jan 21 '22 at 14:14

answered Jan 21 '22 at 12:42

JohanC

71,591
8
33
66

cheers. very nice plot! – Bluetail Jan 21 '22 at 13:49
ok, I will. do you know how to add values to the end of the bars? – Bluetail Jan 21 '22 at 14:02
See second answer [here](https://stackoverflow.com/questions/49044131/how-to-add-data-labels-to-seaborn-countplot-factorplot/49052124) – JohanC Jan 21 '22 at 14:09
thanks. I was trying ax = grid.axes[0, 0] ax.bar_label(ax.containers[0]) and it did not work. – Bluetail Jan 21 '22 at 14:10
I added it to the example code. – JohanC Jan 21 '22 at 14:14
you're a star. I have been trying like this, it did not work. for p, label in zip(g.patches, df[filter]['department'].value_counts().index): g.annotate(label, (p.get_y()+0.375, p.get_width()+0.15)) – Bluetail Jan 21 '22 at 14:31
I know! your example did not render labels with me - I am using Azure Databricks. their notebooks must be different from Jupyter. – Bluetail Jan 21 '22 at 14:34
all done! updated ML runtime to 9.1 from 7.3 and it works :) – Bluetail Jan 21 '22 at 14:56

Customizing a seaborn countplot with additional categorical variable

1 Answers1