Our Correspondence in April 2004 issue of Health Physics News

["Bernard L. Cohen" <blc+@PITT.EDU>]

Dear Jerry:



            Your response to my letter, both published in the April 2004 

issue of Health Physics News, is extremely difficult for me to 

understand. Hopefully, you can  help me understand it.



            You state "The observed negative correlation between various 

smoking related types of cancers and radon strongly indicates that the 

negative correlation between lung cancer and radon is due to confounding 

by smoking". But the smoking variable, S, is included in the analysis 

for the express purpose of removing this confounding by smoking, so the 

only way I can interpret your position is to assume that you believe my 

S-values, which you use, are erroneous.



            But, as pointed out in my letter (referring to my paper in 

Health Physics 86:203-204;2004), I mathematically demonstrated that no 

possible set of S-values, irregardless of their size or the shape of 

their distribution, or of how they are assigned to the 1600 U.S. 

counties, changes your observation of similar dependence on radon 

exposure, r, for lung cancer and other smoking-related cancers. Doesn't 

this prove that erroneous S-values are not the problem? Surely you 

cannot question a mathematical demonstration unless you can suggest 

something that might be wrong with it? You, or anyone else, can easily 

repeat my mathematical demonstration, and I am certain you will find the 

same results.



            You say, without explanation, that you do not agree with my 

"recent treatment of confounding by smoking". Does this refer to my 

procedure for treating it? But the procedure I use is your procedure, 

double regression on r and S. I never used that procedure until you 

introduced it to make your point.



            Please explain specifically what you do not agree with when 

you say you do not agree with my analysis.