How to construct seaborn barplot when groups of bars do not have the same number of hues?

Question

I have a seaborn barplot. There are two groups (two x-axis ticks). The first group has 2 hues/sub-groups/conditions. The second group has 3 hues.

When constructing the barplot, seaborn assumes that there are 3 hues for each group. Thus, it leaves an empty gap where the "third bar" of the first group should be, even though it does not exist.

How do I change the barplot so that:

1. There is no "third bar" gap
2. The width of all bars is the same

By the way, in the left subplot, group2 should have a third bar but is left as "n/a" for unrelated reasons.

chatGPT gave me various manual plotting solutions, but ultimately they didn't work. I'm sure if I tried hard enough the manual plotting could work.

Answer 1

Here's one way to do it that requires manipulating the Rectangle patch positions made by the barplot command:

import seaborn as sns
import pandas as pd

from matplotlib.patches import Rectangle

# some data
data = {
    "group": [1, 1, 2, 2, 2],
    "measure": [0.7, 0.75, 0.78, 0.8, 1.3],
    "hue": [1, 2, 1, 2, 3]
}

df = pd.DataFrame(data)

fig = sns.barplot(data=df, x="group", y="measure", hue="hue")

# get xy positions of rectangles in bar
rects = {}
for p in fig.get_children():
    if isinstance(p, Rectangle):
        if p.get_height() in data["measure"]:
            # rectangle is one of the bars
            xy = p.get_xy()
            rects[xy[0]] = p

# number of values in each group
num_in_groups = df.groupby("group").count()["measure"].values

x0 = min(rects.keys())  # x-position of first bar
pad = 0.02  # some padding between groups

# sort bars on x-position
rects = sorted(rects.items())
width = rects[0][1].get_width()  # bar widths

tickpos = [
    x0 + (width * num_in_groups[0]) / 2,
    x0 + pad + (width * num_in_groups[0]) + (width * num_in_groups[1]) / 2
]

i = 0
for _, rect in rects:
    rect.set(x=x0)
    x0 += width
    i += 1
    if i == num_in_groups[0]:
        # add padding between groups
        x0 += pad

# reset the group labels
fig.set_xticks(tickpos, fig.get_xticklabels())

# reset the x-axis limits
edgepad = 0.05
fig.set_xlim([rects[0][0] - edgepad, rects[0][0] + width * sum(num_in_groups) + pad + edgepad])