[FoRK] Re: [physnews@aip.org: Physics News Update 765]

Robert Harley < robert.harley at gmail.com > on > Fri Mar 10 16:47:54 PST 2006

*Dave Long:
*>[...] Sure, gaussian normal distributions are common, but there's no
reason why lack of one implies any underhandedness[0].
>[0] "Everybody firmly believes in [the central limit theorem] because the
mathematicians imagine it is a fact of observation, and observers that it is
a theory of mathematics" -- Poincare

Gaussian distributions are the best way to model generic movements 99% of
the time but everyone agrees that heavy-tailed (Student-T, Chi^2...)
distributions and jumpy processes (generalized hyperbolic, normal inverse
gaussian...) are the way to go for the 1% of the time that 3+ sigma events
happen that "shouldn't".


More information about the FoRK mailing list