Lesson 15

Colour Maps and Dot Sizes

Add a third and fourth measurement to a scatter plot using the c and s arguments, a colour map, and a colour bar that explains the scale.

Intermediate25 min

Colour Maps and Dot Sizes hero Module 25 Lesson 15

A scatter plot already carries two measurements, in x and y. The colour and the size of each dot can carry two more. It is a genuinely useful trick and an easy one to overdo: four measurements in one picture is close to the limit of what anyone can read.

A scatter plot whose dots vary in both colour and size

A colour per dot

c accepts a list of numbers, one per dot, and cmap names the colour map that turns those numbers into colours. Note the difference from color, which paints every dot the same: only c takes a list of values to be mapped.

python

import matplotlib.pyplot as plt

hours = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]
score = [41, 48, 44, 56, 61, 58, 69, 72, 68, 79, 84, 81]
lessons = [3, 4, 4, 6, 7, 7, 9, 11, 10, 13, 15, 14]

plt.scatter(hours, score, c=lessons, cmap="viridis")
plt.title("Colour shows lessons completed")
plt.xlabel("hours of practice")
plt.ylabel("score")
plt.show()

print("lessons range from", min(lessons), "to", max(lessons))

Output:

text

lessons range from 3 to 15

The colour bar is not optional

A colour scale that is never explained is decoration. plt.colorbar() draws the key beside the chart and label says what it measures. Without it, a reader can see that the dots differ and has no idea why.

python

import matplotlib.pyplot as plt

hours = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]
score = [41, 48, 44, 56, 61, 58, 69, 72, 68, 79, 84, 81]
lessons = [3, 4, 4, 6, 7, 7, 9, 11, 10, 13, 15, 14]

plt.scatter(hours, score, c=lessons, cmap="plasma")
plt.colorbar(label="lessons completed")
plt.title("Practice, score, and lessons")
plt.xlabel("hours of practice")
plt.ylabel("score")
plt.show()

print("the bar turns colour back into numbers")

Output:

text

the bar turns colour back into numbers

Choosing a colour map

Colour maps come in kinds, and the kind matters more than the prettiness. A sequential map such as viridis runs dark to light and suits a quantity. A diverging map such as coolwarm has a neutral middle and suits values above and below a baseline. Add _r to any name to reverse it.

python

import matplotlib.pyplot as plt

kinds = {
    "sequential": ["viridis", "plasma", "Blues", "magma"],
    "diverging": ["coolwarm", "RdBu", "PiYG"],
    "reversed": ["viridis_r", "Blues_r"],
}
for kind, names in kinds.items():
    print(f"{kind:<12} {', '.join(names)}")

change = [-8, -5, -3, -1, 0, 2, 4, 6, 9]
position = list(range(len(change)))
plt.scatter(position, change, c=change, cmap="coolwarm", s=140)
plt.axhline(0, color="#94A3B8", linewidth=1)
plt.colorbar(label="change from target")
plt.title("A diverging map around a meaningful zero")
plt.show()

Output:

text

sequential   viridis, plasma, Blues, magma
diverging    coolwarm, RdBu, PiYG
reversed     viridis_r, Blues_r

plt.axhline(0) draws the baseline the colours diverge around, so the neutral colour in the middle has something to point at.

Size as a fourth measurement

s takes a list of sizes, one per dot, in points squared. Because the eye compares areas poorly, use size for something rough, such as how big a group is, and never for the number you want read precisely.

python

import matplotlib.pyplot as plt

hours = [2, 4, 6, 8, 10, 12]
score = [44, 52, 61, 69, 78, 85]
cohort = [4, 9, 16, 12, 25, 20]
sizes = [size * 18 for size in cohort]

plt.scatter(hours, score, s=sizes, c=score, cmap="viridis", alpha=0.75)
plt.colorbar(label="score")
plt.title("Bigger dot means a bigger cohort")
plt.xlabel("hours of practice")
plt.ylabel("score")
plt.show()

print("largest cohort:", max(cohort), "learners")

Output:

text

largest cohort: 25 learners

Key takeaways

c takes one value per dot; color paints them all the same
cmap names the colour map that turns values into colours
Always add a colorbar, or the colours mean nothing
Sequential maps suit quantities, diverging maps suit a baseline
Next: Histograms
Previous: Scatter Plots

Loading playground…

Check your understanding

Colour Maps and Dot Sizes Quiz

Which argument maps a value to a colour per dot?

1 / 10

← Scatter Plots

Histograms →