I need to visualise joint and marginal frequencies of several low-cardinality categorical variables. Equivalently, I want to visualise sizes of groups and their intersections, where membership in some groups is mutually exclusive but "independent" (meaning compatible) with membership of all other groups. For example, the following features: smoker (Yes/No), sex (M/F), sexual orientation (Hetero/Homo/Bi), marital status (Single/Married/Divorced). An individual would choose exactly one option in each.
A related question has been asked about 5 Boolean variables, where the recommendation was to use a Venn diagram or an UpSet plot. I don't think these plots are well-suited for my case, as they are best for displaying "independent" binary membership information and do not take advantage of mutual exclusivity and grouping of sets (as in Single/Married/Divorced).
In contrast to nested categories plots, such as Dendograms, Treemaps, Sunburst, and also to Venn and UpSet diagrams, my case suggests an orthogonal arrangement, which is absent in all aforementioned plots. To be specific,
- Singe/Married/Divorced and Hetero/Homo/Bi make the 1st and 2nd axes, which produce a 3-by-3 table;
- M/F, can be stacked in the 3rd orthogonal dimension on top;
- the final categorical feature could be distinguished by other means (colour or texture).
Another option would be a 3-by-3 grid of 4-category pie-charts, where colours are chosen to suggest overlapping, but I still wonder if there are any standard solutions I overlooked.
What kind of visualisations, plots or diagrams would be most appropriate in the described situation? Are they implemented in any libraries or packages?
