Stacking categorised seaborn bar plots

user_1208923 0
Viewed 27

Suppose we have a data frame with four columns, A, B, x, y as follows:

data = pd.DataFrame({
    'A' : [1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3],
    'B' : [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1],
    'x' : [4, 4, 5, 5, 6, 6, 4, 4, 5, 5, 6, 6, 4, 4, 5, 5, 6, 6],
    'y' : [166,171, 127, 150, 120, 185, 135, 152, 173,
           192, 174, 185, 101, 102, 134, 100, 110, 143]
})

For each (x, A) pair we have two values of y. I would like to create a bar plot of the total y versus x, with A column determining the categories, while designating what portion of total y is coming from B == 1. An incomplete solution

agg = data.groupby(['A', 'x'])['y'].sum().to_frame().reset_index()
seaborn.barplot(data=agg, x='x', y='y', hue='A', alpha=0.5)
seaborn.barplot(data=data[data.B==1], x='x', y='y', hue='A')

which yields:

enter image description here

The issue is the double legend. I am looking for a way to drop the legend after the first first barplot call while keeping the one from the second.

pandas seaborn python
1 Answers

You can use a "hack" and get your categories as strings starting with an underscore in the first plot. The labels will be hidden:

import seaborn
agg = data.groupby(['A', 'x'])['y'].sum().to_frame().reset_index()
seaborn.barplot(data=agg.assign(A='_'+agg['A'].astype(str).str.zfill(3)),
                x='x', y='y', hue='A', alpha=0.5)
seaborn.barplot(data=data[data.B==1], x='x', y='y', hue='A')

output:

hidden labels bar plot

"hacked" input (using zfill(3) to handle up to 999 sorted categories):

      A  x    y
0  _001  4  337
1  _001  5  277
2  _001  6  305
3  _002  4  287
4  _002  5  365
5  _002  6  359
6  _003  4  203
7  _003  5  234
8  _003  6  253

Other option, rework the legend afterwards:

import seaborn
agg = data.groupby(['A', 'x'])['y'].sum().to_frame().reset_index()
seaborn.barplot(data=agg, x='x', y='y', hue='A', alpha=0.5)
seaborn.barplot(data=data[data.B==1], x='x', y='y', hue='A')

h, l = plt.gca().get_legend_handles_labels()
plt.legend(h[3:], l[3:])
user_163434640
Related