P.S. Pincushion Priors, Part 2
It seems to me that a natural extension of the pincushion prior concept
is to use the head of the pin to represent the probability mass attributed to the spike. When comparing to spike-and-slab
models attributing different masses to the spikes(regardless of the respective slab models), the size of the pinhead can
be illustrated such that is proportional to the respective probability mass.
Thus, for two spike-and-slab models and
given as
the areas (A) of the pinheads for the illustrations of and
should be illustrated such that:
An example is shown below for with the spike probability masses unspecified. Note
that the y-axis scales for probability density will differ when the spike masses differ, so the labels are omitted here.