Pair Plots

Why pair plots?

Pair plots are a fast way to inspect:

• Pairwise relationships between variables
• Distributions on the diagonal
• Group separation using hue

They are great early in EDA.

diagram How pairplot fills a grid mermaid

  flowchart TD
  A["n numeric columns"] --> B["Build an n x n grid"]
  B --> C["Off-diagonal cells:
scatter of column i vs j"]
  B --> D["Diagonal cells:
histogram or KDE of column i"]

pairplot builds an n×n grid: scatter plots off the diagonal, a distribution plot on the diagonal.

From the book

McKinney calls this a “pairs plot or scatter plot matrix” and notes that “making such a plot from scratch is a bit of work, so seaborn has a convenient pairplot function” — it even supports diag_kind=\"kde\" for smooth diagonal curves and a plot_kws dict to pass options (like {\"alpha\": 0.2}) down to every individual scatter plot.

Basic pairplot

pairplot

import seaborn as sns
 
penguins = sns.load_dataset("penguins").dropna()
 
sns.pairplot(penguins[[
    "bill_length_mm",
    "bill_depth_mm",
    "flipper_length_mm",
    "body_mass_g",
]])

Pairplot with hue

pairplot with hue

import seaborn as sns
 
penguins = sns.load_dataset("penguins").dropna()
 
sns.pairplot(
    penguins,
    vars=["bill_length_mm", "bill_depth_mm", "flipper_length_mm", "body_mass_g"],
    hue="species",
)

Tips

• Pair plots can be slow for large datasets. Sample rows if needed.
• Use hue to check if groups separate naturally.

Next

Continue to Facet Grids to split any plot type into small multiples by category.

🧪 Try It Yourself

Exercise 1 – A Basic Pair Plot

pch.exerciseTag pch.exerciseTitle

Exercise 2 – Color by Species

pch.exerciseTag pch.exerciseTitle

Exercise 3 – Use KDE on the Diagonal

pch.exerciseTag pch.exerciseTitle