Re: Fixing dinosaurian carnivour question

[<gejd@concentric.net>]

I have a general interest in the assumptions underlying statistical
models. Often we find that the Gaussian distribution is
assumed as the underlying distribution without any other
justification than that the math is easier. This is particularly
dangerous when a lognormal or other extreme value distribution
is a better approximation to truth.
Does anyone have a reasonably accessible reference to
the paper(s) cited below?:
Thanks,
Gus Derkits

Stanley Friesen wrote:

>
> And on top of that Peter Ashlock and his students have developed a
> mathematical method for determining an optimal choice of which clade to
> subtract, based on information theory.  [The method also helps *rank* the
> resulting taxa].