Subject:  (Reuss) (Maxey) Probability of an association..... (fwd)
Date:     Thu, 13 May 1999 084108 -0500 (CDT)
From:     "Roy L. Beavers" <rbeavers@llion.org>
To:       emfguru <rbeavers@llion.org>
--------------------------------------------------


---------- Forwarded message ----------
Date: Thu, 13 May 1999 15:10:52 +0200
From: Christoph Reuss 
To: "Roy L. Beavers" 
Subject: Re: (Maxey) Probability of an association.....

Roy,
your readers may be interested in my reply to Ed's query too...
Cheerio,
Chris



> Date: Thu, 29 Apr 1999 17:32:28 +0200
> To: Edward S Maxey 
> From: creuss@bluewin.ch (Christoph Reuss)
> Subject: Re: Ozarks
> Cc: rcmahq@nwark.com, rbeavers@llion.org
>
>
> Dear Ed,
>
> thank you very much for quoting me at the Ozarks meeting.
>
> > If it is convenient, please run your computer program with input
> > values of  624 and 215 to confirm the 295,971,724,905,491result
> > obtained here.
>
> The 295,971,724,905,491 result is correct for the input of 215.
> For the median of 0.5 mG, you could even use a lower input (getting an
> even higher result), because the distribution is Gaussian instead of linear.
> However, assuming the "general" median for the Linet data is a bit 'bold';
> for the above calculation to be correct, we would need the actual median
> of the Linet measurements.  (After all, it could be that the Linet study
> considered unrepresentatively high-EMF or low-EMF homes.)
>
> Also, the statement we have from the above probability is only that it is
> very unlikely that the Linet results are 'random'.  We could be much more
> specific if we would have  not only the number of cancer cases per interval,
> but also the number of measurements per interval, so we could calculate the
> _relative_ incidence of cancer.  Did Linet provide the the number of
> measurements per interval too?  If so, we could run a correlation between
> strength of EMF and relative incidence of cancer.  (Preferably with smaller
> intervals, say 0.1 mG wide)
>
> Finally, let's remember the difference between correlation and causality.
> Although the probability you quote is extremely small, it is no scientific
> proof that EMFs _cause_ leukemia.  Because there could be another factor
> (or several) that correlates both with cancer and EMFs -- e.g. diesel or
> lead pollution.  It could be that on average, the homes with high EMFs
> happen to be the homes with high pollution (e.g. inner-city areas), and
> that leukemia is caused by this pollution.  In that case, there could be
> a high correlation between EMF and cancer _without_ any causality between
> the two.  A scientific study should control for as many other possible
> factors (confounders) as possible, to rule out non-causal correlations.
> If the study didn't do this, we cannot scientifically deduct causality
> from correlation, or from the above probability.
>
> Cordially,
> Chris






Archive provided courtesy of WaveGuide, http://www.wave-guide.org
Reprinted with permission of Roy Beavers, http://www.feb.se/EMF-L/EMF-L.html