[geeks] Ok I am ready to kill this Fu**ing linux box

[Joshua D Boyd]

On Fri, May 24, 2002 at 11:05:40AM -0400, Kurt Huhn wrote:
> > But more to the point, you are skewing the results by automatically chucking
> > out the messages on the list from people who said it works fine.  If you
> > went back through the archive, I think you would find a sample size closer
> > to 10, which is closer to usefull, but 30 people would be nicer.
> 
> I think I agree.  But my eyes usually glaze over the instant statics are
> mentioned - as they can be seriously skewed an anumber of different
> directions.  Statisticly, statics seem to be innacurate representations
> of reality...

Simple statistics (mean, stdev, etc) can be easily skewed.  Confidence 
intervals and hypothesis tests are methods for dealing with that skewing.
True, they too can be skewed, but my understanding is that the odds of
it happening are pretty astronomical, unless someone is intentionally 
trying to  screw them up.

-- 
Joshua D. Boyd