Re: Radon and Lung Cancer

[Al Tschaeche <antatnsu@pacbell.net>]

Otto:  Thanks for clarifying what, for me, was beginning to be an
impenetrable thicket of verbosity.  However, given what you say is
true, I have a great problem believing that any conceivable
or inconceivable complex interactions could result in Cohen's curve that
is some 20 standard deviations away from any LNTH curve, whichever one
you pick.  That would just be too much of a coincidence, don't you
think?  So my vote remains with Bernie.  However, if someone
could develop such improbably complex (but not unreasonable)  interactions
and fit the data to Bernie's curve, perhaps we could discuss the result? 
If no one can do that, I believe that argument against Bernie's curve is
hypothetical and, if not demonstrated to be true, can and should be ignored. 
Al Tschaeche antatnsu@pacbell.net

"Otto G. Raabe" wrote:

March 12, 2000
Davis, CA

As I understand Jay Lubin's point about Cohen's data, in an ecological
process it is mathematically possible to consistently obtain an overall
inverse correlation between radon in countries and lung cancer rates
(Cohen's results) even if the dose response relationship for radon is LNT
if there are regular complex interactions (cross correlations) among the
various factors that contribute to or hinder lung cancer rates. This
mathematical possibility exits even if we don't understand the nature of
these interactions. Hence, I believe Lubin is saying that we cannot be sure
that Cohen's results show anything but the results of these complex
interactions between all of these factors. The factors themselves may look
straightforward, such as age and associates radon exposure history, smoking
history, environmental factors, available medical services, economic and
employment factors, etc. Even though we might understand these various
factors by themselves and separate them out (stratify the data), the
complex cross interactions of these factor may me difficult to interpret or
predict and may continue to affect the results. In a case-control study it
is theoretically possible to control for know factors.

Anyway, this is my understanding of Lubin's position. There are, of course,
problems with case-control studies, also.