How To Draw Multiple Levels Groupby Histograms In Python?

I have a dataframe like this, col1 col2 col3 A east 5 A west 7 A east 1 A east 6 B east 2 B

Solution 1:

Using seaborn:

• pandas.DataFrame.sort_values on col1 will also order the plot sequence from A to Z

Data:

import string
import numpy as np
import pandas as pd
import random

alpha_list = [random.choice(list(string.ascii_uppercase)) for _ inrange(10_000)]
coor_list = [random.choice(['east', 'west']) for _ inrange(10_000)]
rand_val = [np.random.randint(10) for _ inrange(10_000)]
df = pd.DataFrame({'col1': alpha_list, 'col2': coor_list, 'col3': rand_val})
df.sort_values(by='col1', inplace=True)

col1  col2  col3
   A  west     1
   A  east     3
   A  west     9
   A  west     5
   A  west     7
   A  east     1
   A  east     5
   A  west     2
   A  east     2
   A  west     2

Plot:

• Select the number displayed per row with col_wrap
• Building structured multi-plot grids
• seaborn.FacetGrid
• seaborn.distplot

g = sns.FacetGrid(df, col='col1', hue='col2', col_wrap=7)
g.map(sns.distplot, 'col3', hist_kws=dict(edgecolor='black'), bins=range(0, 11, 1), kde=False)
plt.xlabel('Value Range')
plt.ylabel('Frequency')
plt.legend()
plt.xticks(range(1, 11, 1))
plt.show()

• if you want tick labels on each graph, put the following code on the line before plt.show():

for ax in g.axes.flatten():
    ax.tick_params(labelbottom=True)

Solution 2:

You can pivot the table and groupby plot:

df = pd.DataFrame({'col1': np.random.choice(('A','B'), size=100),
                   'col2': np.random.choice(('east', 'west'), size=100),
                   'col3': np.random.randint(0,10, 100)})

df['idx'] = df.groupby('col1').col2.cumcount()

(df.pivot_table(index=['idx','col1'], 
                columns=['col2'], 
                values='col3')
   .groupby('col1').plot.hist(subplots=False, alpha=0.5)

)

Output: