Relationship

Mosaic plot (contingency table)

Represents contingency-table frequencies through recursively partitioned rectangular areas.

Plot
Relationship
Composition

What it answers

How are categorical variables associated?

Shape: Proportionally partitioned tiles
What it shows: Unequal combinations and association patterns.

Example

The question

Does the share of repeat buyers differ by sign-up channel?

Customers by sign-up channel and repeat purchase (mosaic plot)

Tile widths show each channel’s share of customers; heights split repeat and one-off buyers. Search: 45% of customers, 62% bought again; Social: 30% of customers, 38% bought again; Referral: 25% of customers, 74% bought again.

Illustrative data. It belongs to no company.

When to use it

Use it when

Exploring contingency tables and category association.

Avoid it when

Sparse high-dimensional categories or unclear partition order.

How to read it

Read tile area as count/share and inspect departures from a reference model if shown.

Note

Related to Marimekko charts but often used for statistical association.

The data you need

Layout: Long table
Fields: CategoryA: text; CategoryB: text; Count: nonnegative number
One row is: One category combination and its frequency. What is the grain?
Rules: Aggregate duplicate combinations. Preserve zero cells and use a positive total.
Optional: Raw alternative: one row per observation with both category columns. More factors are possible.
Preparation: Build a complete contingency table; calculate joint frequencies and optionally expected counts for the chosen association display.
Units: Tile area represents contingency count or grand-total share. Residual shading, if present, is a different statistic.

A long table: one row per observation, few columns. New data adds rows.

Worked example

Counts A-Yes 30, A-No 10, B-Yes 20, B-No 40 total 100. Tiles occupy 30%,10%,20%,40%; group composition differs.

CategoryA	CategoryB	Count
A	Yes	30
A	No	10
B	Yes	20
B	No	40

All charts in the library