Re: ?Ecological Fallacy

["David Scherer" <DAVID@speedy.ehs.uiuc.edu>]

> You misrepresent the conclusion. The proof of the cause at the 
individual
> level "proves" the validity of ecological epi studies, with the caveat that
> _positive_ effects must show some causal connection. 
> 
> Of course, this does not exist in, e.g., the uranium miners who can NOT be
> shown to have contracted lung cancer from radon exposure. 
> 
> Whereas _negative_ or null effects in ecological studies CAN refute a proposed 
> causal connection, ie, IF radon actually caused lung cancer, then the German
> waterworks maintenance workers that work in 100,000s pCi/L concentrations MUST 
> have a lung cancer effect (they do not, so lung cancer is shown NOT to cause
> cancer at high doses). The same is true for workers and residents in radon
> spas and surrounding areas and other high-background areas (including some
> data with individual radiation monitoring and health effects tracking). 

You miss the point.  The ecologic study does not adequately control 
against confounding variables.  It turned out in the case of goiter 
that iodine was the cause.  But without the rigorous epi study, we 
would not know this.  It could have been that areas low in iodine are 
also low in some other element and that this other element was the 
cause.  In this case the suggestion from the ecologic study turned out 
to be valid, but this can only be known in retrospect.

I don't see why negative results in ecologic studies are any more 
valid than positive results.  For example, in the case of the radon 
study it could be that persons with low radon exposure have a higher 
incidence of other cancers that has metastasized to the lung.  This 
may be due to some other environmental agent or family history, or 
some other cause that has not yet been identified.  I don't know 
that this is true, but the lack of adequate controls prevents ruling 
it out.

As far as my other comments go, I was addressing the glib dismissal 
of the systematic errors reported by Dr. Strom et al.  It appears to 
me that the criticisms of the ecologic method is not an attempt to 
prove or disprove radiation risk coefficients.  It is a criticism of 
the method itself regardless of the health problem to which it is 
applied.  The responses on radsafe seem to be that the criticism are 
"potential" errors that "do not apply" in this study.  These do not 
seriously address the logic behind the criticisms.  Other responses 
have been impassioned defenses of the _results_ of the study, not 
analysis of the methods.  This is not science.

If the epidimological community has arrived at a consensus that 
ecologic studies are systematically flawed regardless of the agent 
under study, we health physicists should take heed.  If the 
epidimologists said these methods were valid for any agebt other than 
radiation, then I would be sceptical.  Then it might be reasonable to 
suspect their motives.  But it is my understanding that this approach 
is unreliable regardless of the suspected cause/effect linkage.

By the way, If a study has a "potential" error, the study is 
unreliable!  If I use an uncalibrated ion chamber to measure an 
exposure rate, the lack of calibration is only a "potential" error, 
but the results are not reliable.  It is not incumbent on the critic 
to check my meter to see if there are in fact calibration errors.  I 
doubt that Dr. Feynman meant that we should follow the data but 
ignore systematic errors.

Also, please note that I have no personal agenda that is advanced by 
radiophobia.  My life would be much better if the public had a better 
perspective on radiation risk.  (For example, I was called in at 3:30 
last night because a little scintillation fluid had leaked on a lab 
floor.)  Nor have I concluded that biological response _must_ follow a
linear, no threshold model.  But I am not willing to compromise my 
intellectual integrity to advance a view just because it would make 
my life easier.  This whole thread would be much better without 
impuning motives to those those with whom we disagree.

David Scherer
scherer@uiuc.edu