Seaborn Expert

Create publication-quality statistical visualizations with Seaborn, a Python library for informative, attractive plots from structured datasets. This skill covers categorical plots, distribution visualization, regression plots, matrix displays, multi-faceted figure grids, and custom styling for scientific publications.

When to Use This Skill

Choose Seaborn Expert when you need to:

• Create statistical visualizations that automatically compute and display aggregations
• Build multi-faceted plots (FacetGrid, PairGrid) for exploring relationships across subgroups
• Produce polished publication figures with minimal boilerplate code
• Visualize distributions, relationships, and categorical comparisons from pandas DataFrames

Consider alternatives when:

• You need full low-level control over every plot element (use matplotlib directly)
• You need interactive or web-based visualizations (use Plotly)
• You need geographic or spatial visualizations (use GeoPandas or Folium)

Quick Start

Copy
pip install seaborn pandas matplotlib numpy

Copy
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

# Set publication style
sns.set_theme(style="whitegrid", font_scale=1.2,
              rc={"figure.figsize": (8, 5), "savefig.dpi": 300})

# Load example dataset
tips = sns.load_dataset("tips")

# Faceted visualization
g = sns.FacetGrid(tips, col="time", row="smoker", hue="sex",
                  height=3.5, aspect=1.2, margin_titles=True)
g.map_dataframe(sns.scatterplot, x="total_bill", y="tip", alpha=0.7)
g.add_legend()
g.set_axis_labels("Total Bill ($)", "Tip ($)")
g.savefig("tips_facet.pdf")

# Distribution comparison
fig, axes = plt.subplots(1, 2, figsize=(10, 4))
sns.violinplot(data=tips, x="day", y="total_bill", hue="sex",
               split=True, inner="quart", ax=axes[0])
axes[0].set_title("Bill Distribution by Day")

sns.boxenplot(data=tips, x="day", y="tip", palette="Set2", ax=axes[1])
axes[1].set_title("Tip Distribution by Day")
fig.tight_layout()
fig.savefig("distributions.pdf")

Core Concepts

Plot Type Reference

Function	Category	Best For
sns.scatterplot()	Relational	Two continuous variables
sns.lineplot()	Relational	Trends over continuous axis
sns.histplot()	Distribution	Single variable distribution
sns.kdeplot()	Distribution	Smooth density estimate
sns.ecdfplot()	Distribution	Cumulative distribution
sns.boxplot()	Categorical	Quartile comparison
sns.violinplot()	Categorical	Distribution shape comparison
sns.barplot()	Categorical	Mean + confidence interval
sns.stripplot()	Categorical	Individual data points
sns.heatmap()	Matrix	Correlation or expression matrices
sns.clustermap()	Matrix	Hierarchically clustered heatmap
sns.lmplot()	Regression	Scatter + fitted regression line
sns.jointplot()	Joint	Bivariate + marginal distributions
sns.pairplot()	Multi	All pairwise relationships

Advanced Multi-Panel Figures

Copy
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib.gridspec as gridspec
import pandas as pd
import numpy as np

# Custom multi-panel figure
penguins = sns.load_dataset("penguins").dropna()

fig = plt.figure(figsize=(12, 8))
gs = gridspec.GridSpec(2, 3, hspace=0.35, wspace=0.35)

# Panel A: Scatter with regression
ax1 = fig.add_subplot(gs[0, :2])
sns.scatterplot(data=penguins, x="flipper_length_mm", y="body_mass_g",
                hue="species", style="sex", s=60, alpha=0.7, ax=ax1)
ax1.set_title("Flipper Length vs Body Mass", fontweight='bold')

# Panel B: Violin
ax2 = fig.add_subplot(gs[0, 2])
sns.violinplot(data=penguins, y="bill_depth_mm", x="species",
               palette="Set2", inner="box", ax=ax2)
ax2.set_title("Bill Depth", fontweight='bold')

# Panel C: KDE
ax3 = fig.add_subplot(gs[1, 0])
for species in penguins['species'].unique():
    subset = penguins[penguins['species'] == species]
    sns.kdeplot(subset['bill_length_mm'], label=species, ax=ax3, fill=True,
                alpha=0.3)
ax3.set_title("Bill Length Density", fontweight='bold')
ax3.legend(fontsize=8)

# Panel D: Heatmap correlation
ax4 = fig.add_subplot(gs[1, 1:])
numeric_cols = penguins.select_dtypes(include='number')
corr = numeric_cols.corr()
sns.heatmap(corr, annot=True, fmt='.2f', cmap='RdBu_r', center=0,
            square=True, linewidths=0.5, ax=ax4)
ax4.set_title("Feature Correlations", fontweight='bold')

# Panel labels
for ax, label in zip([ax1, ax2, ax3, ax4], ['A', 'B', 'C', 'D']):
    ax.text(-0.1, 1.05, label, transform=ax.transAxes,
            fontsize=14, fontweight='bold')

fig.savefig("penguin_analysis.pdf", bbox_inches='tight')

Configuration

Parameter	Description	Default
style	Plot background style (whitegrid, darkgrid, white, dark, ticks)	"whitegrid"
context	Scaling for different contexts (paper, notebook, talk, poster)	"notebook"
font_scale	Font size multiplier	1.0
palette	Color palette name or list	"deep"
rc	matplotlib rcParams overrides	{}
height	FacetGrid subplot height in inches	5
aspect	Width:height ratio for FacetGrid	1
ci	Confidence interval size for aggregation plots	95

Best Practices

1. Set the theme once at the top of your script — Use sns.set_theme(style="whitegrid", context="paper", font_scale=1.1) before any plotting. This applies consistent styling to all subsequent plots. Use context="talk" for presentations and context="poster" for large-format output.

2. Use figure-level functions for exploration, axes-level for publication — Figure-level functions (sns.lmplot, sns.catplot, FacetGrid) create their own figures and are great for quick exploration. For publication figures with custom GridSpec layouts, use axes-level functions (sns.scatterplot, sns.boxplot) that accept an ax parameter.

3. Always pass data as a DataFrame with named columns — Seaborn is designed for structured data. Pass data=df with x="column_name" rather than raw arrays. This enables automatic axis labels, legends, and semantic grouping with hue, style, and size parameters.

4. Choose the right plot for your data type — Continuous vs continuous: scatter/line. Continuous vs categorical: box/violin/strip. Distribution: histogram/KDE/ECDF. Matrix: heatmap/clustermap. Selecting the wrong plot type obscures patterns or misleads readers.

5. Combine strip/swarm plots with box/violin for complete pictures — Aggregation plots (boxplot, barplot) hide individual data points. Overlay sns.stripplot(alpha=0.3) on top of sns.boxplot() to show both the distribution summary and raw observations, especially for small sample sizes (n < 30).

Common Issues

Legend overlaps with plot data — Use plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left') to place the legend outside the plot area. For FacetGrid, use g.add_legend() which automatically positions it. If the legend is still problematic, move it to a separate call after all plotting.

Colors are inconsistent across subplots — When using hue across multiple separate plots, pass the same palette and hue_order to each plot call. Seaborn assigns colors based on the data present in each plot, so if a category is missing from one subplot, colors shift. Explicit hue_order prevents this.

Plot looks cramped with many categories on x-axis — Rotate tick labels with plt.xticks(rotation=45, ha='right') or switch to horizontal orientation using orient='h' in categorical plots. For more than 8-10 categories, consider using sns.catplot with col or row faceting instead of cramming everything into one panel.