Bubble Chart

A bubble chart encodes a third continuous variable as the size of each scatter point. This allows three-dimensional relationships to be displayed in a 2D figure. Its most common error is encoding size as radius rather than area —which makes large values appear disproportionately large (area scales with radius squared, so a 2xradius bubble looks 4xas large).

STEP 1 —When to Use

Use a bubble chart when: you have three continuous variables where the third variable modulates the meaning of the x-y relationship (e.g., effect size on x-y correlation); sample size n is encoded as bubble size to show that larger studies have more weight; a third variable has a natural size interpretation (population, volume, mass).

Avoid bubble charts when: the third variable range is small (bubbles will be nearly identical in size —use color instead); you have more than ~30 data points (overplotting of bubbles is unresolvable without transparency); the size encoding is difficult to read at the figure's final print size.

[Image placeholder: Two bubble charts of the same data. Left: bubble radius proportional to the third variable —large bubbles appear >4xas large as they should. Right: bubble area proportional —correct visual encoding. Same data, different perceived ratios. Size scale legends shown in both.]

Radius encoding (left) distorts size perception. Area encoding (right) is correct —a bubble with twice the value has twice the area.

STEP 2 —Journal Requirement

Size scale legend is mandatory: example bubbles at specific values must be shown.
The legend must state whether size represents area or radius (area is correct).
The third variable name and units must be in the legend.
Overplotting: if bubbles overlap, transparency must be applied and noted in the legend.
Correlation coefficients for any two-variable pair must be noted if relevant.

STEP 3 —Licklider Implementation

Input

Three continuous variables: x, y, and size variable
Optional: categorical grouping variable (bubble outline color)
Size scaling: area proportional (default and required), or user-override with warning

Output

Scatter + bubbles with area proportional to the size variable
Size scale legend with 3— reference bubbles at representative values
Transparency for overlapping bubbles (auto-applied when overlap detected)
Optional: regression line through the x-y relationship (with causation disclosure)
Radius-encoding guard: if user selects radius encoding, a warning is shown and area encoding is set as default

[Image placeholder: Licklider bubble chart with 22 data points. X=effect size, Y=study duration, size=sample size (area proportional). Size scale legend in bottom right showing bubbles at n=20, 50, 100, 200. Semi-transparent bubbles with 0.5 alpha. Pearson r annotation for x-y correlation.]

Licklider bubble chart: area-encoded size, scale legend with reference bubbles, transparency for overlap, correlation annotated.

STEP 4 —Draft Output (Draft / Needs review)

Bubble chart displaying the relationship between [X variable] (x-axis),
[Y variable] (y-axis), and [Z variable] (bubble size). Bubble area is
proportional to [Z variable] (size scale shown in legend). n = 22 data
points (studies). Transparency (alpha = 0.5) applied to overlapping bubbles.
Pearson correlation between X and Y: r = 0.38 (95% CI [0.05, 0.63], p = 0.037).