W0UN wrote:
>
>NSF funded a study to look at the increased incidence
>of cancer around nuclear power plants and saw a small but
>significant increase. To better benchmark their results they
>then studied cancer around PROPOSED nuclear power plants
>and also found a significant increase. It would appear that
>just THINKING about installing a nuclear power plant will
>increase the incidence of cancer in the area under consideration.
>Obviously that is not correct--but rather the methodology was
>flawed.
>
The methodology is to stick a pin in the map and draw a circle round it.
Then you count the number of cancers inside that circle and compare with
the national average.
If you draw a circle big enough to enclose the whole nation, then of
course the number of cancers will *be* the national average. But if you
draw a much smaller circle, your results will be subject to a lot of
statistical "noise", ultimately depending whether particular individuals
inside the circle get cancer or not. You're only looking for a very
small excess of cancers above average, so even in a population of
thousands, the fate of one or two individuals can make a big difference
either way.
If you increase the size of the circle, step by step, the statistical
noise goes down - the more people you count, the more accurate your
average. But in order to converge onto the national figure when the
circle is big enough, some sizes of circle must give results above the
average, while others must give results below.
So you can get very different results, depending on the size of circle
you decide to draw. Most studies require a circle drawn around an
existing community... and then it's literally a matter of luck whether
your results are high or low.
That's the fundamental weakness of epidemiology - the numbers of deaths
are factual enough, but they tell you *absolutely nothing* about why
those individual people actually died.
--
73 from Ian G3SEK Editor, 'The VHF/UHF DX Book'
'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.com/g3sek
--
FAQ on WWW: http://www.contesting.com/FAQ/amps
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
|