How to manually scale count of Seaborn Violinplot

Question

I'm using Seaborn in Python to make split violinplots. The trouble is that one of my two hue's has a much smaller count that makes it almost imperceptible when side-by-side the other hue.

My question is how can I add a multiplier to scale the count on one side of the violin splits?

Here is the documentation of the kind of thing I'm doing:

And here is how it currently affecting me:

In the end I'd like it to be clearly perceptible that the right-side of the violins has a significantly smaller overall count -while still portraying its distribution.

you can change the `scale` parameter to `area` or even `width` but I'm afraid you can't have the effect of `count` and `area` at the same time — Stef, Jan 03 '21 at 12:04
It would be better if you made your code written rather than a picture. — M. Small, Jan 04 '21 at 19:13

score 3 · Accepted Answer · answered Feb 28 '21 at 23:12

I suggest that you solve this issue by displaying an enlarged view of the violin plot half that is hard to perceive. This can be done by using an inset axes. Here is an example based on the one you have shared from the seaborn docs:

import seaborn as sns  # v 0.11.0

# Import sample dataset and generate seaborn violinplot
tips = sns.load_dataset('tips')
ax = sns.violinplot(x="day", y="total_bill", hue="sex", data=tips, palette="Set2",
                    split=True, scale="count", inner="stick", scale_hue=False, bw=.2)
ax.figure.set_size_inches(10,4)
ax.legend(loc='upper left', frameon=False)

# Create inset axes to zoom in on violinplot showing 'day'='Fri' & 'sex'='Female'
tips_fri_fem = tips[(tips['day']=='Fri') & (tips['sex']=='Female')]
axins = ax.inset_axes([0.44, 0.1, 0.06, 0.8])
sns.violinplot(data=tips_fri_fem, x='sex', y="total_bill", hue="sex",
               palette="Set2", split=True, scale="count", inner="stick",
               scale_hue=False, bw=.2, ax=axins)
axins.set_xlim(1, 1.5)
axins.set_ylim(0, 30)
axins.set_xlabel('')
axins.set_ylabel('')
axins.set_xticklabels('')
axins.legend().remove()

# Add rectangle without the connector lines to indicate area that is enlarged
rect, lines = ax.indicate_inset_zoom(axins, alpha=0.8)
rect.set_bounds(*rect.get_xy(), 0.1, 30)
for line in lines:
    line.remove()

How to manually scale count of Seaborn Violinplot

1 Answers1