I finally made one of these mosiac plots in R with an interface to OrTSDA. This plot is of a contingency table of collision types by functional class for state highways in Oregon, 2003 to 2007. The colors across columns match the collision type on the y-axis. The width of the column is proportional to contribution of that facility type; the height is proportional to contribution to that crash type for that facility type. It should be considered draft. Of no surprise to anyone with experience with the Oregon data, fixed-object crashes have the largest "footprint" - and are mostly a rural problem (the yellow squares). You can also see other trends if you study closely.
