Radon and Lung Cancer

[Holloway3@aol.com]

The large size of the data set used by Dr. Cohen gives me considerable 
confidence in his results.  That large size will eliminate many problems that 
might be present with a smaller data set. Consider that to predict the 
results of a national election, a sample size of  only a little over a 
thousand voters is required to give a result accurate to within about 3%.  If 
the polling service increased the sample size to 100 million voters, then the 
error would be reduced  considerably and assuming no one changed their mind 
and the sample was randomly selected, the result would be just about fool 
proof in terms of any significant statistical problem or confounding factor.  
Or course taking a sample that large would be prohibitively expensive but 
that is beside the point in this context.  Cohen's data resembles the sample 
size of 100 million voters more than it does the sample of 1,000 voters.  The 
only thing remarkable in this situation is how resistant some people are to 
the obvious conclusion.  The obvious conclusion is that contrary to 
expectation, lung cancer has a strong negative correlation with radon 
concentrations, suggesting the possibility that radon may helpful in 
preventing lung cancer deaths rather than promoting lung cancer.  But is this 
really such a shocking conclusion?  After all, radiation has been used for 
many years to treat cancer.  There is a slight possibility that his results 
are an illusion but the probability of that is small, given the amount of 
data that produces the negative correlation.  Sometimes it pays to be 
convinced by the obvious, rather than to get hung up with details that have 
little impact on the overall picture.  
************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html