Skip to main content

Seaborn (Part 3): Relational and categorical plots

With distributions covered in Part 2, relational plots answer joint behavior (“as X climbs, does Y drift?”), while categorical encodings highlight strata.

📚 Prerequisites​

  • Tidy DataFrames (one observation per row).

🎯 What you'll master​

  • Encode third dimensions with hue, style, size.
  • Prefer catplot when faceting categorical comparisons.

Relational grids​

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

tips = sns.load_dataset("tips")  # bundled sample data

sns.relplot(
    data=tips,
    x="total_bill",
    y="tip",
    hue="time",
    col="sex",
    kind="scatter",
    height=3.8,
)
plt.show()

relplot returns a facet grid—it is ideal when you iterate categories without manual subplot loops.

Categorical bar estimates​

sns.catplot(
    data=tips,
    x="day",
    y="total_bill",
    hue="smoker",
    kind="bar",
    errorbar=("ci", 95),
    height=4,
)
plt.show()

Bars summarize central tendency (default mean) plus confidence intervals—state both when publishing.

💡 Key takeaways​

  • relplot / catplot trade fine-grained Matplotlib tweaking for succinct faceting clauses.
  • Never stack unrelated metrics on identical axes scales without normalization—readers confuse units.

➡️ Next steps​

Step into dashboards with Plotly and Dash (introduction).