[R] Fitting this data with a gaussian would be great

Bert Gunter gunter.berton at gene.com
Sat Feb 23 22:03:51 CET 2013


FWIW (It may not be much):

1. Data **are never** Gaussian. Failure to reject the hypothesis that
the data are sampled from a Gaussian does not mean that the data can
be assumed to be drawn from a Gaussian. That depends on the
statistical methodology and the application context.

2. Given a large enough sample, one will always reject the hypothesis
that the data are drawn from a Gaussian. That does not mean that
nevertheless making that assumption will result in any problems.  That
depends on the statistical methodology and the application context.

3. Mostly 1) and 2) are of little import anyway, despite what the
statistical texts say. Much more important in practice -- and the
source of much grief and many irreproducible results -- is the "i.i."
of iid. ... aka unknown exogenous systematic effects, measurement
biases, clustering ...

4. I know this is OT, and I apologize. I also know that this is just
my 2 cents opinion -- and probably not really worth even that much --
so feel free to dismiss. Also, if you care to reply or argue, please
do so off list. I will not defend anything I've said on list. I've
already got an apartment reserved for me in one of Pat Burns's "R
Inferno" levels, and I don't want to descend even further.

Best,
Bert



More information about the R-help mailing list