A split violin plot displaying two distributions side-by-side within each violin, with each half representing a different group. Unlike standard violin plots that mirror the same distribution, split violins use the left and right halves to compare two conditions (such as before/after, male/female, or control/treatment) at each category level. This enables direct visual comparison of distribution shapes between paired groups.
""" anyplot.ai
violin-split: Split Violin Plot
Library: pygal 3.1.0 | Python 3.13.13
Quality: 92/100 | Updated: 2026-05-08
"""
import os
import numpy as np
import pygal
from pygal.style import Style
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_MUTED = "#6B6A63" if THEME == "light" else "#A8A79F"
# Okabe-Ito palette: position 1 (brand green) and position 2 (vermillion)
COLOR_1 = "#009E73" # Before
COLOR_2 = "#C475FD" # After
# Data - Patient recovery scores before/after treatment across clinics
np.random.seed(42)
categories = ["Clinic A", "Clinic B", "Clinic C", "Clinic D"]
split_groups = ["Before", "After"]
# Generate realistic before/after data with different improvements per clinic
data = {}
for cat in categories:
data[cat] = {}
if cat == "Clinic A":
data[cat]["Before"] = np.random.normal(45, 12, 80)
data[cat]["After"] = np.random.normal(72, 10, 80)
elif cat == "Clinic B":
data[cat]["Before"] = np.random.normal(50, 15, 80)
data[cat]["After"] = np.random.normal(68, 12, 80)
elif cat == "Clinic C":
data[cat]["Before"] = np.random.normal(42, 10, 80)
data[cat]["After"] = np.random.normal(78, 8, 80)
else: # Clinic D
data[cat]["Before"] = np.random.normal(55, 18, 80)
data[cat]["After"] = np.random.normal(65, 14, 80)
# Clip to realistic 0-100 range
for cat in categories:
for group in split_groups:
data[cat][group] = np.clip(data[cat][group], 10, 95)
# Custom style for 4800x2700 px canvas
custom_style = Style(
background=PAGE_BG,
plot_background=PAGE_BG,
foreground=INK,
foreground_strong=INK,
foreground_subtle=INK_MUTED,
guide_stroke_color=INK_MUTED,
colors=(COLOR_1, COLOR_2),
title_font_size=28,
label_font_size=22,
major_label_font_size=18,
legend_font_size=16,
value_font_size=14,
opacity=0.7,
opacity_hover=0.9,
)
# Create XY chart for split violin plot
chart = pygal.XY(
width=4800,
height=2700,
style=custom_style,
title="violin-split · pygal · anyplot.ai",
x_title="Clinic",
y_title="Recovery Score (0-100)",
show_legend=True,
legend_at_bottom=True,
legend_at_bottom_columns=2,
stroke=True,
fill=True,
dots_size=0,
show_x_guides=False,
show_y_guides=True,
range=(0, 100),
xrange=(0, 5.5),
margin=60,
)
# Parameters for violin shapes
violin_width = 0.38
n_points = 80
marker_width = 0.04
# Pre-compute all violin shapes and markers
before_violins = []
after_violins = []
before_markers = []
after_markers = []
for i, category in enumerate(categories):
center_x = i + 1.25
for group in split_groups:
values = data[category][group]
# Create range of y values for density
y_min, y_max = values.min(), values.max()
padding = (y_max - y_min) * 0.15
y_range = np.linspace(y_min - padding, y_max + padding, n_points)
# Compute Gaussian KDE using Silverman's rule (inlined)
n = len(values)
std = np.std(values)
iqr = np.percentile(values, 75) - np.percentile(values, 25)
bandwidth = 0.9 * min(std, iqr / 1.34) * n ** (-0.2)
density = np.zeros_like(y_range)
for v in values:
density += np.exp(-0.5 * ((y_range - v) / bandwidth) ** 2)
density /= n * bandwidth * np.sqrt(2 * np.pi)
# Normalize density to desired width
density = density / density.max() * violin_width
# Compute quartile statistics
median = float(np.median(values))
q1 = float(np.percentile(values, 25))
q3 = float(np.percentile(values, 75))
# Create half-violin shape (split violin - each group on one side)
if group == "Before":
# Left half - density extends to the left
half_points = [(center_x - d, y) for y, d in zip(y_range, density, strict=True)]
half_points = [(center_x, y_range[0])] + half_points + [(center_x, y_range[-1]), (center_x, y_range[0])]
before_violins.append(half_points)
before_markers.append((center_x, median, q1, q3, -0.08))
else:
# Right half - density extends to the right
half_points = [(center_x + d, y) for y, d in zip(y_range, density, strict=True)]
half_points = [(center_x, y_range[0])] + half_points + [(center_x, y_range[-1]), (center_x, y_range[0])]
after_violins.append(half_points)
after_markers.append((center_x, median, q1, q3, 0.08))
# Add all "Before" violins first (green, with legend entry for first one only)
for i, violin in enumerate(before_violins):
label = "Before" if i == 0 else None
chart.add(label, violin, show_dots=False)
# Add all "After" violins (orange, with legend entry for first one only)
for i, violin in enumerate(after_violins):
label = "After" if i == 0 else None
chart.add(label, violin, show_dots=False)
# Add quartile markers for "Before" group (no legend entries)
for center_x, median, q1, q3, offset in before_markers:
# IQR line (thin vertical line)
iqr_line = [(center_x + offset, q1), (center_x + offset, q3)]
chart.add(None, iqr_line, stroke=True, fill=False, show_dots=False, stroke_style={"width": 8})
# Median marker (small horizontal line)
median_line = [(center_x + offset - marker_width, median), (center_x + offset + marker_width, median)]
chart.add(None, median_line, stroke=True, fill=False, show_dots=False, stroke_style={"width": 12})
# Add quartile markers for "After" group (no legend entries)
for center_x, median, q1, q3, offset in after_markers:
# IQR line
iqr_line = [(center_x + offset, q1), (center_x + offset, q3)]
chart.add(None, iqr_line, stroke=True, fill=False, show_dots=False, stroke_style={"width": 8})
# Median marker
median_line = [(center_x + offset - marker_width, median), (center_x + offset + marker_width, median)]
chart.add(None, median_line, stroke=True, fill=False, show_dots=False, stroke_style={"width": 12})
# X-axis labels for categories
chart.x_labels = [
{"value": 0, "label": ""},
{"value": 1.25, "label": "Clinic A"},
{"value": 2.25, "label": "Clinic B"},
{"value": 3.25, "label": "Clinic C"},
{"value": 4.25, "label": "Clinic D"},
{"value": 5.5, "label": ""},
]
# Save outputs
chart.render_to_file(f"plot-{THEME}.html")
chart.render_to_png(f"plot-{THEME}.png")
Part of Split Violin Plot on anyplot.ai.