Pat Brennan wrote:
This distribution is a good way to describe measures on some of the traits
of the selfings population, but it often does not fit. If a trait is caused
by a single gene and the mother has Xx, the selfing trait distribution will
be a two or three humped distribution. Add more genes to a trait and who
knows what you come up with. I do not think it even has to be continuous.

Actually, when you have several random processes, their sum looks more
and more like a normal distribution (ie, bell shaped) as you add new
processes. This is know as the Central Limit Theorem in statistics. As
little as the sum of six uniform deviates is a pretty good approximation
to the Gaussian distribution.

--Matt