Visualize density with 2D KDE contours

Last updated 2020-04-08 Read time 1 min

seaborn.kdeplot can draw the joint density of two variables as contours or filled areas. It is especially helpful when a scatter plot becomes overcrowded.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
import seaborn as sns
import matplotlib.pyplot as plt

penguins = sns.load_dataset("penguins").dropna(subset=["bill_length_mm", "bill_depth_mm"])

fig, ax = plt.subplots(figsize=(5.5, 4.5))
sns.kdeplot(
    data=penguins,
    x="bill_length_mm",
    y="bill_depth_mm",
    hue="species",
    fill=True,
    thresh=0.05,
    levels=6,
    alpha=0.6,
    ax=ax,
)

ax.set_xlabel("Bill length (mm)")
ax.set_ylabel("Bill depth (mm)")
ax.set_title("2D KDE by penguin species")
ax.grid(alpha=0.2)

fig.tight_layout()

plt.show()

When scatter plots are crowded, a 2D KDE highlights the dense areas.

Reading tips #

Tighter contours indicate denser regions, and the color intensity provides an intuitive sense of frequency.
Tune thresh to drop very low-density rings and keep the figure clean.
For very large data sets, KDE can be expensive, so consider sampling or adjusting bw_adjust to widen the bandwidth.

Density Plot — Visualize distribution with a smooth curve
Hexbin Plot — Aggregate scatter density with hexagonal bins

Density Plot — understanding this concept first will make learning smoother