Demo of Half-Violin Plots for Showing distributions

accuracy_by_window_size.csv

demo_half_violin_plot.py

from matplotlib import pyplot as plt
import seaborn as sns
import numpy as np
import pandas as pd

sns.set_style('white')

y = 'window_size'
x = 'accuracy'

x_df = pd.read_csv('accuracy_by_window_size.csv')
mean_df = pd.read_csv('mean_by_window_size.csv')
#x_df = x_df.query("sample_id < 100").copy()

x_df[y] = x_df[y].astype("category")

palette = sns.cubehelix_palette(start=.5, rot=-.5, dark=0.3, light=0.7)
ax = sns.violinplot(
	y=y, x=x, data=x_df,
    palette=palette,
    scale='width',
    inner=None)
xlim = ax.get_xlim()
ylim = ax.get_ylim()
for violin in ax.collections:
    bbox = violin.get_paths()[0].get_extents()
    x0, y0, width, height = bbox.bounds
    violin.set_clip_path(plt.Rectangle((x0, y0), width, height / 2, transform=ax.transData))

# Add box plot showing 25th, 50th, and 75th percentiles
sns.boxplot(
 	y=y, x=x, data=x_df,
 	saturation=1, showfliers=False,
	whis=[10,90], # whiskers at 10th and 90th percentiles
    width=0.3, boxprops={'zorder': 3, 'facecolor': 'none'}, ax=ax)
old_len_collections = len(ax.collections)

# Add a strip plot of semi-transparent dots indicating individual samples
sns.stripplot(
 	y=y, x=x, data=x_df,
 	color='dodgerblue',
 	alpha=0.1,
 	ax=ax)
for dots in ax.collections[old_len_collections:]:
     dots.set_offsets(dots.get_offsets() + np.array([0, 0.12]))


# Add a line indicating the mean
x_W2 = mean_df[x].values[:,np.newaxis] * np.ones((1, 2))


yticks = ax.get_yticks()
ydelta = float(yticks[2] - yticks[1])
y_W2 = yticks[:,np.newaxis] + np.asarray([[0.0, -0.3 * ydelta]])


print(y_W2)
plt.plot(x_W2.T, y_W2.T, '-', color='k', linewidth=1)

ax.set_xlim(xlim)
ax.set_ylim(ylim)
plt.show()

mean_by_window_size.csv

Instructions:

1. Have a modern python installed with Pandas, NumPy, Scipy, matplotlib and seaborn.

Something like the COMP 136 environment would do fine.

2. Run the script with no arguments:

python demo_half_violin_plot.py

3. Expected result: This plot