Boxplot with groupby two features

Question

Say I have a df consisting of 3 columns ['Y/N','Test Before','Test After'], where 'Y/N' is boolean and 'Test Before,' 'Test After' is float.

eg,

'Y/N'    'Test Before'    'Test After'
True         75               100
False        75                50
True         50                60
False        50                40
  ...

I can use df.boxplot(column='Test Before/Test After', by 'Y/N') to create separate boxplots of 'column name' each grouped by 'Y/N.'

However, I would like to create separate boxplots of boolean 'Y/N,' grouped by 'Test Before,' 'Test After,' eg:

Boxplot A contains boxplots of df['Y/N'] == True for x values df['Test Before'] and df['Test After'].

Boxplot B contains boxplots of df['Y/N'] == False for x values df['Test Before'] and df['Test After'].

Answer 1

While your desired result is a bit unclear, consider melting your data before plotting to have one indicator column for Test Stage and one value column for Test Value. Then plot with seaborn's boxplot with Y/N as the legend (hue) series or with seaborn's FacetGrid with separate graphs for each Y/N distinct value. Below runs on seeded, random data for demonstration:

Data

import pandas as pd
import numpy as np

import seaborn as sns
import matplotlib.pyplot as plt

np.random.seed(11012018)
df = pd.DataFrame({'Y/N': np.random.choice([True, False], 50),
                   'Test Before': [np.random.uniform()*10 for _ in range(50)],
                   'Test After': [np.random.uniform()*10 for _ in range(50)]},
                  columns = ['Y/N', 'Test Before', 'Test After'])

# MELT DATA (WIDE TO LONG)
melt_df = df.melt(id_vars="Y/N", value_name="Test_Value", var_name="Test_Stage")
print(melt_df.head())
#      Y/N   Test_Stage  Test_Value
# 0  False  Test Before    7.573898
# 1   True  Test Before    3.487735
# 2  False  Test Before    1.506599
# 3  False  Test Before    9.833866
# 4   True  Test Before    1.340375
# 5  False  Test Before    3.678094
# 6   True  Test Before    3.407419
# 7  False  Test Before    0.427210
# 8  False  Test Before    6.988953
# 9  False  Test Before    2.912770

Plot

fig, ax = plt.subplots(figsize=(10,4))
sns.boxplot(data=melt_df, x='Test_Stage',  y='Test_Value', hue='Y/N', ax=ax)

plt.legend(loc='upper right')

g = sns.FacetGrid(melt_df, col="Y/N", height=4, aspect=6/4)
g.map(sns.boxplot, data=melt_df, x='Test_Stage',  y='Test_Value')

Additional Data

Should there be more than just Test_Before and Test_After values, melt scales to it with order specifying box plots arrangement:

np.random.seed(11012018)
df['Test Middle'] = [np.random.uniform()*10 for _ in range(50)]
melt_df = df.melt(id_vars="Y/N", value_name="Test_Value", var_name="Test_Stage")

fig, ax = plt.subplots(figsize=(10,4))
sns.boxplot(data=melt_df, x='Test_Stage',  y='Test_Value', hue='Y/N', ax=ax,
            order=['Test Before', 'Test Middle', 'Test After'])

g = sns.FacetGrid(melt_df, col="Y/N", height=4, aspect=6/4)
g.map(sns.boxplot, data=melt_df, x='Test_Stage',  y='Test_Value',
      order=['Test Before', 'Test Middle', 'Test After'])