Re: NCRP bias?

[BERNARD L COHEN <blc+@PITT.EDU>]

On Fri, 2 May 2003, Otto G. Raabe wrote:

>

> A few years ago Bernie Cohen and I were invited to attend a committee

> hearing on LNT at the NCRP headquarters in Bethesda. At that meeting, we

> watched as Jay Lubin derived on the board the mathematical relationships

> associated with possible slope observations in an ecological study such as

> Bernie's and showed convincingly that unrecognized and possibly

> undetectable cross-interactions between the variables could result in a

> completely meaningless "observed" regression slope irrespective of the

> underlying true relationship. Hence, it is a mathematical truth (not a

> religious conviction) that the "observed" slope in Bernie's study of the

> possible relationship for radon-induced lung cancer could be totally

> spurious. On the other hand, this is not a proof that it is necessarily wrong.



     --As a former physicist who has published many papers with

mathematical proofs, I can testify that mathematical proofs are not a

substitute for rational consideration. Very few things about nature have

been discovered by mathematical proofs; they are used to "dress up" the

prsentations after the discoveries have been made.



     --The substance of the Lubin mathematical proof was recognized even

in my original paper, in which I gave an example of such a confounding

relationship that would invalidate my conclusions. However, I showed

there that the requirements on that example were completely implausible.

As I have pointed out in publications, Lubin's mathematical proof has a

corrolary requiring plausibility, and the only way to satisfy the

plausibility requirement is by proposing an example.

	The paper posted as Item #7 on my web site gives, in Section 3.1,

a general treatment of the plausibility requirements on a confounding

factor that would invalidate my conclusions. It finds that the existence

of such a confounding factor is extremely implausible. If anyone finds

that treatment less than completely convincing, please let me know.



	--It is also easy to show that an ignored confounding factor can

invalidate any epidemiological study, and even the best epidemiological

studies consider only a handful of potential confounding factors, seldom

more than 5 or 10. These potential confounding factors are selected by the

authors' impressions of what is plausible. On the other hand, my studies

have investigated over 500 potential confounding factors, and uses an

argument on "plausibility of correlations" to consider the effects of

unrecognized confounding factors. Nothing approaching such an elaborate

treatment is offered in any other epidemiology study that I know of.



************************************************************************

You are currently subscribed to the Radsafe mailing list. To unsubscribe,

send an e-mail to Majordomo@list.vanderbilt.edu  Put the text "unsubscribe

radsafe" (no quote marks) in the body of the e-mail, with no subject line.

You can view the Radsafe archives at http://www.vanderbilt.edu/radsafe/