Re: We are killing nuclear workers!

["Joan Stovall" <joans@pcez.com>]

"The key point is, that for any set of epidemiological data, when you break
 it into smaller and smaller sub-sets (by age, type of cancer, etc.) you
will
 get some sub-sets that show a positive effect, and some that show a
negative."

How true.

When one looks at many statistical samples, about 5% of the samples will
fall outside of the 95% confidence level limits for the distribution.
Hence, if an unscrupulous researcher focuses on those cases to draw his
conclusions, it is easy to prove anything with statistics.


************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html