From: Joel Neely <joel.neely@...>
Date: 2001-10-26T06:12:12+09:00
Subject: [ruby-talk:23352] Re: [OT] half below average

Well, to nitpick**3 ... ;-)

The Gaussian distribution is like the regression line, a model
whose virtue lies in its simplicity and tractability more than
its perfect accuracy.  The Gaussian distribution is often "good
enough" for practical purposes in the same sense that 22/7 (or
its bigger cousin 355/113) is usually "good enough" as pi.

Distributions that have a hard lower limit (e.g. zero) but no
a priori upper limit often are "close to" Gaussian but not
perfectly so...  My suspicion is that far more people have
zero skill in domain x than have twice the average skill level.

Studies have shown that excellent programmers are often 10 or
more times as effective as their lower-skilled counterparts,
hence my assumption that the distribution is actually quite
skewed.

-jn-

Avi Bryant wrote:
> 
> On Fri, 26 Oct 2001, Joel Neely wrote:
> 
> > But, unfortunately, not true.
> >
> > Half of all X are below the *median*, but not below the average
> > unless there's a totally symmetric distribution, which rarely
> > happens in nature.
> 
> To nitpick the nitpick - things in nature often have a gaussian (normal)
> distribution, which is indeed symmetric.  Human height, etc, show a normal
> distribution, and I don't see why programming skill wouldn't.
> 
> IANAS (I am not a statistican), however.
> 
> Cheers,
> avi
> 
> > For the collection of scores
> >
> >     50, 50, 50, 50, 50, 50, 50, 50, 50, 100
> >
> > 90% are below average.
> >
> > -jn-
> >
> > brucedickey wrote:
> > >
> > > Good one.
> > >
> > > > From: Ned Konz
> > > >
> > > > Oddly enough, half of all programmers are below average...
> >
> >

-- 
"This sentence contradicts itself -- no actually it doesn't."
-- Doug Hofstadter
                              joel<dot>neely<at>fedex<dot>com