Seaborn: Statistical Data Visualization

TIP

Learning Objectives: After this lesson, you'll create beautiful statistical graphics with minimal code using seaborn—distributions, relationships, categorical comparisons, and styled presentations.

TIP

How to read this lesson. Seaborn isn't bundled in this in-browser Python sandbox, but matplotlib is — and seaborn is a thin, opinionated layer over matplotlib. So every runnable cell below renders a real chart with matplotlib, and each block names the one-line seaborn equivalent (marked ≡ sns.…). You learn what the chart is, and the exact seaborn call you'd reach for in your own environment.

Why Seaborn?

Before writing any code, get a feel for what a statistical chart actually communicates. The plotter below shows a frequency distribution—the same kind of shape sns.histplot() produces.

Try it: Switch the chart type and tweak the values, then watch the shape redraw. Notice how a tall center with tapering sides reads as a "normal" distribution—the exact pattern the histogram code below generates.

Seaborn is built on matplotlib but provides a high-level interface for creating attractive statistical graphics. What takes ten lines in matplotlib often takes one in seaborn.

Plotting one scatter per category in raw matplotlib is a loop:

COPY
import matplotlib.pyplot as plt
fig, ax = plt.subplots()
for group in data['category'].unique():
    subset = data[data['category'] == group]
    ax.scatter(subset['x'], subset['y'], label=group)
ax.legend()
ax.set_xlabel('X')
ax.set_ylabel('Y')
plt.show()

Seaborn collapses the whole loop — grouping, coloring, and the legend — into a single call:

COPY
sns.scatterplot(data=data, x='x', y='y', hue='category')

That economy is the whole point. Seaborn gives you beautiful defaults, built-in statistical aggregation (means, confidence intervals, KDEs), automatic legends and labels, native pandas integration, and a deep catalog of specialized statistical plots.

Setting Up Seaborn

Seaborn ships with sensible defaults you set once per session:

• Styles (sns.set_theme(style=...)): whitegrid and darkgrid add gridlines; white, dark, and ticks keep things minimal.
• Palettes (sns.set_palette(...)): deep is the default, muted/pastel are softer, husl/bright are colorful, and colorblind is accessibility-safe.

In a real project you import seaborn and pick a theme and palette once, up front:

COPY
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

sns.set_theme()                  # Apply seaborn's default look

# Themes — overall background + grid
sns.set_theme(style='whitegrid') # Clean, with a grid (great for stats)
sns.set_theme(style='darkgrid')  # Dark background with grid
sns.set_theme(style='white')     # Minimal, no grid
sns.set_theme(style='ticks')     # Axis ticks, no grid

# Palettes — the sequence of colors used for categories
sns.set_palette('deep')          # Default
sns.set_palette('muted')         # Softer
sns.set_palette('colorblind')    # Accessibility-safe

The single most useful habit: reach for colorblind or a perceptually-uniform palette like viridis so your charts stay readable for everyone.

Distribution Plots

Understanding data distributions is fundamental to data analysis.

Histograms and KDE

(The frequency-distribution plotter at the top of this lesson previews the shape this code produces.)

Box Plots and Violin Plots

Relationship Plots

Explore relationships between variables.

Scatter Plots with Regression

Pair Plots

Visualize relationships between all pairs of variables:

Heatmaps

Visualize matrices and correlations:

Categorical Plots

Compare categories effectively.

Bar Plots with Error Bars

Count Plots

Strip and Swarm Plots

Show individual data points:

Combining Plots

FacetGrid: Multiple Subplots

Create a grid of plots based on data subsets:

Complete Example: EDA Dashboard

Key Takeaways

✅ Distribution plots: histplot(), kdeplot(), boxplot(), violinplot()

✅ Relationship plots: scatterplot(), regplot(), pairplot(), heatmap()

✅ Categorical plots: barplot(), countplot(), stripplot(), swarmplot()

✅ FacetGrid: Create grids of plots by data subsets

✅ Styling: set_theme(), color palettes, context settings

✅ Integration: Works seamlessly with pandas DataFrames

Connections: Seaborn in Practice

🔗 Connection to Statistics

🔗 When to Use Which Plot

Practice Exercises

Exercise 1: Complete Visualization

Next Steps

Now you're ready for Exploratory Data Analysis (EDA)—a systematic approach to understanding data by combining all the techniques you've learned.

Ready to explore real datasets? EDA is next!

Further Reading

Visual Galleries

• Seaborn Example Gallery — every chart type with thumbnail + code. Bookmark.
• From Data to Viz — flowchart-driven guide: pick your data shape, get the right chart and the seaborn snippet.
• Python Graph Gallery — Seaborn — 200+ examples organized by chart type.

Official Docs

• Seaborn — Tutorial — official walkthrough.
• Seaborn — Figure-level vs Axes-level Functions — the most important conceptual distinction in modern seaborn.
• Seaborn — Color Palettes — colorblind-safe and perceptually-uniform palettes (viridis, cividis, colorblind).

Tutorials

• Real Python — Python Plotting With Seaborn — the canonical primer.
• seaborn.objects — declarative interface — the new pipe-friendly API.

Modern Declarative Alternatives

• Altair — Vega-Lite-based grammar of graphics. Most principled "declarative viz" library in Python.
• plotnine — ggplot2 ported to Python. If you came from R, start here.

Books

• Book: Storytelling with Data — Cole Nussbaumer Knaflic. The why of visualization.
• Book: Fundamentals of Data Visualization — Claus Wilke (free online). Library-agnostic; required reading for anyone making serious charts.