Matplotlib: Stacked and Grouped Bar Plot

matplotlib_barplot.md

Stacked and Grouped Bar Plot

Oddly enough ggplot2 has no support for a stacked and grouped (position="dodge") bar plot. The seaborn python package, although excellent, also does not provide an alternative. However, I knew it was surely possible to make such a plot in regular matplotlib. Matplotlib, although sometimes clunky, gives you enough flexibility to precisely place plotting elements which is needed for a stacked and grouped bar plot.

Below is a working example of making a stacked and grouped bar plot.

import seaborn as sns
import matplotlib.pyplot as plt
import numpy as np

# make up some fake data
pos_mut_pcts = np.array([20, 10, 5, 7.5, 30, 50])
pos_cna_pcts = np.array([10, 0, 0, 7.5, 10, 0])
pos_both_pcts = np.array([10, 0, 0, 0, 0, 0])
neg_mut_pcts = np.array([10, 30, 5, 0, 10, 25])
neg_cna_pcts = np.array([5, 0, 7.5, 0, 0, 10])
neg_both_pcts = np.array([0, 0, 0, 0, 0, 10])
genes = ['PIK3CA', 'PTEN', 'CDKN2A', 'FBXW7', 'KRAS', 'TP53']

with sns.axes_style("white"):
    sns.set_style("ticks")
    sns.set_context("talk")
    
    # plot details
    bar_width = 0.35
    epsilon = .015
    line_width = 1
    opacity = 0.7
    pos_bar_positions = np.arange(len(pos_mut_pcts))
    neg_bar_positions = pos_bar_positions + bar_width

    # make bar plots
    hpv_pos_mut_bar = plt.bar(pos_bar_positions, pos_mut_pcts, bar_width,
                              color='#ED0020',
                              label='HPV+ Mutations')
    hpv_pos_cna_bar = plt.bar(pos_bar_positions, pos_cna_pcts, bar_width-epsilon,
                              bottom=pos_mut_pcts,
                              alpha=opacity,
                              color='white',
                              edgecolor='#ED0020',
                              linewidth=line_width,
                              hatch='//',
                              label='HPV+ CNA')
    hpv_pos_both_bar = plt.bar(pos_bar_positions, pos_both_pcts, bar_width-epsilon,
                               bottom=pos_cna_pcts+pos_mut_pcts,
                               alpha=opacity,
                               color='white',
                               edgecolor='#ED0020',
                               linewidth=line_width,
                               hatch='0',
                               label='HPV+ Both')
    hpv_neg_mut_bar = plt.bar(neg_bar_positions, neg_mut_pcts, bar_width,
                              color='#0000DD',
                              label='HPV- Mutations')
    hpv_neg_cna_bar = plt.bar(neg_bar_positions, neg_cna_pcts, bar_width-epsilon,
                              bottom=neg_mut_pcts,
                              color="white",
                              hatch='//',
                              edgecolor='#0000DD',
                              ecolor="#0000DD",
                              linewidth=line_width,
                              label='HPV- CNA')
    hpv_neg_both_bar = plt.bar(neg_bar_positions, neg_both_pcts, bar_width-epsilon,
                               bottom=neg_cna_pcts+neg_mut_pcts,
                               color="white",
                               hatch='0',
                               edgecolor='#0000DD',
                               ecolor="#0000DD",
                               linewidth=line_width,
                               label='HPV- Both')
    plt.xticks(neg_bar_positions, genes, rotation=45)
    plt.ylabel('Percentage of Samples')
    plt.legend(loc='best')
    sns.despine()

FYI, the actual bar plot image is missing. :)

I had to modify your code slightly so that the legend wasn't on top of the chart over the bars. I replaced the last two lines with the following three lines:

plt.legend(bbox_to_anchor=(1.1, 1.05))  
sns.despine()  
plt.show()  

Hey, thanks for sharing this example! I'm trying to switch the axis and show it as a horizontal bar char, but replacing all "bar" with "barh" leads nowhere... Any ideas?

-- Solved it by removing the "bottom" argument completely.

Appreciate your shared code!
Thanks!

I modified your code slightly so that xticks are centered between the positive and negative bars.

I changed neg_bar_positions to (neg_bar_positions+pos_bar_positions)/2

plt.xticks((neg_bar_positions+pos_bar_postions)/2, genes, rotation=45)

plt.legend(bbox_to_anchor=(1.1, 1.05))
sns.despine()
plt.show()

Thanks for this helpful code! By the way, if you add edgecolor and linewidth arguments to the solid filled bars, then that makes the width of all the stacked bars perfectly consistent without needing to hack with the epsilon value.

@mchun96 Good point, that will make the plotting more robust to the data