Scatter plot encodes values as position of points on two quantitative axes. John Herschel plotted orbits; Galton’s 1880s regressions made the form famous.
In reports, ads and assignments it fits correlation, outliers and clusters in two measures. Skip it when one axis is not quantitative, or time is the real story.
Parts
- Mark — The core mark of Scatter plot is position of points on two quantitative axes.
- Scale — The scale must show zero (or a declared break) and units.
- Axis — Axis titles name the variable, not a mood.
- Legend — A legend is weaker than a direct label. Write beside the mark when you can.
- Label — Numbers are optional. If comparison is the job, label the ends only.
- Grid — Keep the grid faint. Data should own the ink.
- Annotation — One annotation sentence. Date the source.
- Frame — A thin frame. Never a 3-D box.
Variants
- XY plot — Variant 1 of Scatter plot. Same encoding, different layout.
- Correlation chart — Variant 2 of Scatter plot. Same encoding, different layout.
- Point cloud — Variant 3 of Scatter plot. Same encoding, different layout.
- Scatter plot 4 — Variant 4 of Scatter plot. Same encoding, different layout.
Labels
- Axis — On Scatter plot, make Mark readable.
- Legend — On Scatter plot, make Scale readable.
- Label — On Scatter plot, make Axis readable.
- Grid — On Scatter plot, make Legend readable.
- Annotation — On Scatter plot, make Label readable.
Scatter plot: honour position of points on two quantitative axes, and switch when one axis is not quantitative, or time is the real story.