Mosaic plots

Abstract

AbstractMosaic plots are a recursive generalization of barcharts to visualize multivariate categorical data. Mosaic plots were first introduced by Hartigan and Kleiner and are the de‐facto standard for advanced visualizations of categorical data.This review explains the general construction and the successful application of mosaic plots. Recent implementations of mosaic plots offer various generalizations which can effectively cope with almost any kind of categorical data. Besides plotting expected values of a proposed model, these variations include the doubledecker plot, the empty bin display and fluctuation diagrams. Within an interactive graphical analysis tool, subgroups reflecting a certain property can easily be linked into mosaic plots to effectively visualize the relationship between this property and the categorical variables in the plot. Mosaic plots are most powerful when investigating associations between categorical variables, conditioned by other categorical variables. This is illustrated along with an explanation of how Simpson's Paradox for categorical data can be explained using mosaic plots. WIREs Comput Stat 2012, 4:191–198. doi: 10.1002/wics.1192This article is categorized under: Statistical and Graphical Methods of Data Analysis > Statistical Graphics and Visualization