[R] [OT] "normal" (as in "Guassian")

Johannes Hüsing johannes at huesing.name
Sun Mar 2 20:58:52 CET 2008


Am 02.03.2008 um 17:44 schrieb Gabor Csardi:

> I'm not a statistician, but do i remember well that among all
> distributions with a given mean and variance, the normal distribution
> has the highest entropy? This is good enough for me to call it
> "normal"....

There's more. Among all rotation-symmetric distributions,
the standard bivariate normal is the only one where x and
y are independent.

Also, the formula for the standard normal distribution is
the only one that is its own Fourier transform. So, if we
assume the same distribution for a momentum and
a location of a physical object, according to Heisenberg's
Law it has to be the normal.

Whereas we ought to be wary about assumption of normality
for the distribution of phenomena in nature, the normal and
its henchmen play a defendable role when describing summaries
of phenomena, like arithmetic means. I'd even go as far as
buy into Youden's hype described in that Kruskal and Stigler
essay.



More information about the R-help mailing list