matplotlib boxplot doesn't align with overlaid scatterplot

Question

I have a plot where I'm trying to overlay a scatter series on a boxplot series... here is a simple example of the problem so that you can re-create it.

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

names = ['a','b','c','d','e','f']

df = pd.DataFrame(np.random.rand(6,6), columns=names)
display(df)

plt.boxplot(df, labels=names)
plt.show()

plt.scatter(names, df.head(1))
plt.show()

plt.boxplot(df, labels=names)
plt.scatter(names, df.head(1))
plt.show()

Results:

So you see that when both the boxplot and scatter are added to the same figure the labels no longer align correctly. How can I fix this alignment?

Answer 1

• Tested in python 3.8.11, pandas 1.3.2, matplotlib 3.4.3, seaborn 0.11.2
• Notice the xticklabel locations are misaligned.
• As per matplotlib.pyplot.boxplot, position defaults to range(1, N+1)

fig, (ax1, ax2, ax3) = plt.subplots(3, 1, figsize=(6, 8))
ax1.boxplot(df, labels=names)
print(ax1.get_xticks())
ax2.scatter(names, df.head(1))
print(ax2.get_xticks())

ax3.boxplot(df, labels=names)
ax3.scatter(names, df.head(1))
[out]:
[1 2 3 4 5 6]
[0, 1, 2, 3, 4, 5]

• A correct solution, given the existing code, is to set the positions parameter
• This also requires converting the dataframe to long form with pandas.DataFrame.melt, for the scatterplot.

plt.boxplot(df, labels=names, positions=range(len(df.columns)))
plt.scatter(data=df.melt(), x='variable', y='value')

• Alternatively, plot directly with pandas.DataFrame.plot

ax = df.plot(kind='box', positions=range(len(df.columns)))
df.melt().plot(kind='scatter', x='variable', y='value', ax=ax)

• The simplest solution is to use seaborn, which is a high-level API for matplotlib. This plots the given dataframe directly, without any transformations
  • How can box plot be overlaid on top of swarm plot in Seaborn?
  • seaborn.boxplot and seaborn.swarmplot, which is a better option than seaborn.scatterplot
• .boxplot positions begin at 0

import seaborn as sns

sns.boxplot(data=df, boxprops={'facecolor':'None'})
print(plt.xticks())
sns.swarmplot(data=df)
print(plt.xticks())

[out]:
(array([0, 1, 2, 3, 4, 5]), [Text(0, 0, 'a'), Text(1, 0, 'b'), Text(2, 0, 'c'), Text(3, 0, 'd'), Text(4, 0, 'e'), Text(5, 0, 'f')])
(array([0, 1, 2, 3, 4, 5]), [Text(0, 0, 'a'), Text(1, 0, 'b'), Text(2, 0, 'c'), Text(3, 0, 'd'), Text(4, 0, 'e'), Text(5, 0, 'f')])

Answer 2

The shape of your data passed to scatter is incorrect, and the index needs to be shifted as boxplot indices start from 1.

Here is a working solution:

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

names = ['a','b','c','d','e','f']

df = pd.DataFrame(np.random.rand(6,6), columns=names)
display(df)

plt.boxplot(df, labels=names)
plt.show()

df2 = df.copy()
df2.columns = range(1, len(df2.columns)+1)
df2 = df2.unstack().reset_index(level=0)

plt.scatter(*df2.values.T)
plt.show()

plt.boxplot(df, labels=names)
plt.scatter(*df2.values.T)
plt.show()

What needs to be fed to scatter:

[1., 1., 1., 1., 1., 1., 2., 2., 2., 2., 2., 2., 3., 3., 3., 3., 3.,
3., 4., 4., 4., 4., 4., 4., 5., 5., 5., 5., 5., 5., 6., 6., 6., 6.,
6., 6.]

and

[0.40196123, 0.24802927, 0.3152357 , 0.76739054, 0.06847434,
0.30173798, 0.30350823, 0.36324843, 0.35928768, 0.69194326,
0.61586685, 0.74039024, 0.56971769, 0.37970524, 0.67049866,
0.82031296, 0.15917911, 0.58760412, 0.02660848, 0.67014213,
0.46027398, 0.39161172, 0.01191327, 0.48197714, 0.79882359,
0.84189102, 0.10873758, 0.56858526, 0.48169526, 0.08860958,
0.43500853, 0.26380151, 0.14911615, 0.47846386, 0.25583401,
0.04949216]