[ RadSafe ] LNT

[JPreisig at aol.com]

Dear Radsafe,
 
    Hey All.  Interesting comments on LNT --- Linear,  No Threshold on 
Radsafe lately.
    Sometimes, LNT may be someone's initials.
 
 
    I understand the desire to extrapolate a fitted  line/curve (in a 
dose/effect plot) into the low dose
region.  However, rigorously, one should only plot such a fitted  
line/curve through the region of
actual data points.  A good attorney/Health Physicist (there seems to  be 
at least one of these persons
on Radsafe) could probably argue in a courtroom (Federal???) that the  
extrapolation should not occur.
Any laws based on such an extrapolation should be reviewed and perhaps  
modified, depending on
any legal ruling.   Time to take the EPA to court???  Bring  along a 
statistician???
 
   Other comments follow.  The low dose (or low dose rate)  region of a 
dose/effect curve is
inherently noisy.  Hormesis may be occuring there.  Cancer  causation may 
be occuring???
Competing effects???  Should the fits be linear with an offset and  trend 
(i.e. rate)???  Should a 
quadratic term be introduced???  One should use a Chi-square test (or  
whatever statistical test is
relevant) to test the goodness of fit.  Should the data fitting be  
Gaussian or Poisson???  Poisson 
fitting may be more proper in radiation data fitting.
 
    There are many good books on statistics.  Bevington  is one for Physics 
students.  Another book out of
Carnegie-Mellon is also good but I have forgotten the names of the  
authors.  IMSL data routines
(sometimes provided with Fortran Compilers) are useful for  
statistics/fitting.  The Book Numerical
Recipes by Press, Teukolsky, Flannery et al. is also useful.
 
    One book on dose-effect curves is the book by Eric  Hall.
 
    Take Care...  Joe Preisig