Mosaic Plot

A mosaic plot displays the joint distribution of two categorical characteristics so that every area is proportional to the corresponding cell frequency of a contingency table. Unlike the stacked bar chart, whose columns are all equally wide, it encodes both margins geometrically.

• Column width — proportional to the marginal frequency of the X category: wide means frequent, narrow means rare
• Segment height — proportional to the conditional probability $P(G\mid X)$ within that column
• Area — width times height, hence proportional to the cell frequency $n_{ij}$

Hence its real strength: if the segment heights are equal across all columns, then $P(G\mid X)=P(G)$ — the characteristics are independent. The mosaic plot is thus a visual precursor to the chi-square test, and the column widths simultaneously show which category even has enough cases to support a claim.