Here is an attempt at describing how circular displays, like those proposed by Circos, work and how they can be used to summarize large cross-classification Tables.
A bit of context I am particularly interested in displaying large association Tables in a graphical manner so as to make possible the visual comparison of different results (e.g., in different clinical subgroups, or following different clustering algorithms) and alleviate hard-to-read numerical displays. The basic idea is to be able to display (1) the relative prevalence of each symptom/response and (2) the magnitude of their links.
When browsing Tweeter feeds yesterday, I just noticed a post by J.D. Long (alias @CMastication) referring to a nice way of illustrating SQL joins statements with Venn diagrams by Jeff Atwood. So I wonder how it could be reproduced in R.
I initially thought of hacking the venneuler package. However, it happens that I really need a few things, so that I just wrote a wrapper function that takes care of drawing two spheres and shading the appropriate areas.
Before talking about social networks or graph visualization, let’s look at the article written by Matthew Bloch and Jonathan Corum for the New York Times (May 5, 2008): Mapping the Human Diseasome.
Surprisingly, this gives a very clear picture of the links between various disorders, mostly of genetic origin. This original scientific work, entitled The human disease network, has been published in PNAS and the abstract is reproduced below: Goh, K.