Re: Mossman paper in Health Physics News - The Debate is Over

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

On Tue, 10 Jun 2003 epirad@mchsi.com wrote:



> Dr. Cohen, You wrote - Am I correct in assuming that is what you mean?

>

> No, you are not correct, I gave these examples below not the ones you stated

> (see below) -

>

> I have offered (since 1999) a plausible explanantion for your findings.

>

> In the previous email I pointed out that - your smoking data is co-correlated

> with socioeconomic status, poor health care, apartment living, mobility, lower

> education status, etc.



	-- Each and every one of these was included in my studies. I call

them confounding factors and they are treated in Item #7 on my web site.

Can you name one that I did not include?



  You have never assessed all these variables in a

> multivariate analyses.



	--I use a better method than multivariate analysis, as I explained

in my last message



 And in fact, this would be impossible to do adequately

> because of the non linear relationships between these factors within and

> between counties.



	--My method does not assume linear relationships

>

> If you do not feel it is plausible, which scientists do you recommend we have

> review whether or not this is a plausible explanation?

>

	--Any theoretical physicist you suggest



> You have not answered others questions regarding whether you would accept the

> NCRPs opinion on whether or not your poor smoking data in large part explains

> your finings.



	--Can you resend the message about my accepting the NCRP

decisions? I remember receiving it but cannot find it in my e-mail file.



  We previously pointed out the poor predictive ability of your

> smoking data to explain lung cancers.  Puskin has further shown your inverse

> association is also found for other smoking related cancers.  It is pretty

> clear that you have residual confounding from smoking and the confounding is

> enhanced by factors co-correlated to smoking.



	--I don't see why you can't give me a hypothetical specific

example of what can go wrong? If you want me to give hypothetical examples

of what can go wrong with any usual case-control study, I will give you

any number of things.



> I understand your "treatments", but disagree that you can validly use

> additional faulty summary data to "treat" your existing summary data.

>

	--Please say what you disagree with in the tightly reasoned

explanations given in item #7 on my web site.



> Bill Field

> > On Fri, 6 Jun 2003 epirad@mchsi.com wrote:

> >

> > > In the previous email I pointed out that - your smoking data is co-correlated

> > > with socioeconomic status, poor health care, apartment living, mobility, lower

> > > education status, etc.  You have never assessed all these variables in a

> > > multivariate analyses.  And in fact, this would be impossible to adequately

> > > because of the non linear relationships between these factors within and

> > > between counties.

> >

> > 	--The basic equation for my study is the following relationship

> > between lung cancer rates corrected for smoking prevalence, M, and radon

> > exposure, r, for U.S. counties:

> > 	M = A + B r

> > where A is close to 1.0 and according to LNT, B =+7.3 (in percent

> > per/pCi/L)> If you do not agree to this, that is another issue.

> > 	The issue you raise here is that there may be confounding factors

> > that correlate strongly, for unrelated reasons, with M and with r. As an

> > example, there may be pollutants X, Y, and Z that correlate with M because

>

> > they cause lung cancer or because they stimulate the desire to smoke. For

> > some unrecognized reason, they may scavenge radon out of the air in homes.

> > Thus, counties with high levels of these pollutants would tend to have

> > high lung cancer and low radon levels, and vice versa. Am I correct in

> > assuming that is what you mean?

> > 	This issue is addressed in Item #7 on my web site, in Section 3,

> > where it is shown that the existence of confounding factors with the

> > required correlations with both M and r is extremely implausible. If you

> > read and understand the treatments in that section and have reservations

> > about them, please let me know so we can discuss them. If you do not

> > understand them, please ask for explanations on specific point

>



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

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/