[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Ecological Fallacy
A somewhat lengthy (sorry!) comment by Fritz Seiler
on my positive evaluation of Bernie Cohen's Data set,
and on the Practice of Flaming by Private e-mail
Last Sunday, I received a communication marked as a private one
to me alone. Therefore, it has not been mass posted on RADSAFE
and RISKANAL. However, there are several points that finally need
to be made, and made in no uncertain terms. So, although the writer's
identity was given to me, I will in my discussion guard the writer's
privacy by using the term ANON(ymous). No pun intended! :-)
First off, flaming somebody in a private e-mail, is still flaming (A
rose by any other name....) and is, therefore, insulting. That I know
who wrote these personal insults, does not change the fact that they
are insults. I guess that all I can do here is to take them with a good
dose of some rather black humor.
Oho! The speller just told me that the spelling ‘humour' that I
used is incorrect! Poor British, they always want to be correct!! :-))
Now, let us see what we have here:
ANON: "I was sent one of your postings that appeared on the Radsafe
listserv the other day. I was glad to see that you could respond on your
own without Dr. Alvarez."
FAS: This is not worthy of comment. However, a comment on our
collaboration to the members of the mailing lists may be in order. Joe
and I have been collaborating for ten years, even though we never
worked in the same location. So we talk by phone, on the average at
least once every other day, discussing all kinds of items of common
interest. We have published together if the subject was a problem that
we have discussed before [sometimes almost endlessly! ;-)) ]. If the
subject did not involve substantial collaboration or an agreement to do
it together, we have published separately or with other co-authors. So
there is nothing unusual here. The same goes also for our letters to the
mailing lists. There is little that appears here that we have not discussed
already. So,..... .........But, ANON, rest assured that we can get along
quite well by ourselves, thank you. :-))
John Auxier, Joe's boss, once asked us how two people with such
oversize egos like ours could possibly get along. We looked at him, at
each other, and finally came up with: "Respect, I guess."
ANON: "I find it interesting that the only people who have commented
on Dr. Cohen's posting to date have been members of the band (RS&H).
I welcome scientific discussions. I disdain political mudslinging. It
appears you have a political cause you hope to further. Please at least
try to keep a scientific perspective."
FAS: Wow! You are quite hoity-toity, ANON! It seems that when
some people do not agree with you, you label their arguments as
‘political mudslinging'!? But, ANON, how can you be so blatantly
obvious?! Indeed, both Joe and I belong to the group ‘Radiation,
Safety and Health'. It is a group of people like Joe and myself who
are tired of hearing the same old 'Lame Excuses' of the 'Linearists'.
People who are dead set against the Bad Science and the immense
monetary damages that your group and its stand on low-dose health
effects have caused and are causing.
Despite ample evidence, your group is not willing to consider
any nonlinear dose-effect model. If you do consider a nonlinear
function, it is only for comparison purposes. But then, that
comparison does not mean anything, because your group went and
made some basic mistakes at the level of Linear Algebra 101 (see
BEIR III & V)! Now, do not get hot under the collar, ANON, that
is an old, old hat! We published this sorry fact 5 years ago, also in
the peer-reviewed literature, so it has been out there for a long, long
time.
And "members of the band"? Well, as for me, I do not play an
instrument......... :-))
But how would you like it if I were to start enumerating the many
characteristics that would justify me in calling your group the "Linear
Mafia"? Come on now, let us get off the flame wagon!
When I first read it, I thought: "Political Cause? Not one little bit!
That is a lot of pure @#$%^$#@...!" My second thought was: "Well,
if you don't have any valid scientific arguments left to make, you accuse
the other side of sinister political motives..." But then, it finally dawned
on me: While we do not have a political motive, we do have an agenda,
and it is a powerful one: Both as citizens and as scientists we are honor-
bound to expose "Bad Science" wherever we find it. And, if you check
it out, point by point, you find that THE LNT IS BAD SCIENCE!
The basic tenets of your group, their manner of, shall we say, data
"handling" (!?), violate just about every requirement of the Scientific
Method, as Joe and I have shown in previously published papers (Also
peer-reviewed!). And then there are the outright algebraic mistakes in
the BEIR III and BEIR V reports. All considered, it is no wonder that
Gunnar Walinder -- and he is not exactly a scientific lightweight -- has
called it "the greatest scientific scandal of the twentieth century!"
Our duty as scientists and citizens is particularly important, if we,
as a nation, spend billions of dollars more than necessary every year
on totally uncalled for remediations on the say-so of a group of "Linear
Enthusiasts"! If it makes you feel real good, deep down, when you
see a Superfund site, all torn up, and being cleaned way down to EPA
specs, then let me tell you what I see: I see the site with a sign over
it that says "Billions of Your Tax Dollars Wasted Here!" Again, on the
say-so of the "Linearists".
ANON: "But, I am amazed that you think the quality of Dr. Cohen's
data has not been called into question. Either you ignore our previous
peer reviewed published communications concerning Dr. Cohen's
work or you do not understand it. A third possibility is you hope to
misrepresent it. .......What have we said that you disagree with from
a scientific perspective? To date, I have not seen any specific
comment made that refutes what we said..... I would welcome your
comments concerning our publications that address specific statements
we have made."
FAS: Yes, ANON! You are right! We did indeed ignore your
published work! And with good reason. But then, you still do not
get it, do you? We have made a logical end run around you and all
of your careful arguments! Touchdown!! ;-))
That -- in a nutshell -- is what we have done, and I will now
demonstrate why this is so.
Let us start with Bernie Cohen's data. ANON, if you want to
question the validity of Bernie's data, you must show either that they do
not represent the measurable occurrence of lung cancer in the counties
used, or that the concentrations of radon are not representative for
those counties. Those are his data, no more! And I think they are
correct. The fact they are not in the form of your model output is not
Bernie's fault but yours.
The fact that his data do not support your model is not Bernie's
fault either. Remember the Scientific Method? Experimental Data are
the King of the Scientific Mountain! You use the data as a verification
for your model prediction, and not the other way around!
If confounding factors prevent a direct comparison, you either apply
the corrections your model prediction and show compliance with the data,
or then you can correct the data and show a favorable comparison with
your model output. There is no other way to do science, and - mind you,
my dear ANON -- ALL THIS WORK IS YOUR JOB, NOT OURS!
So much then, for your daily dose on the Scientific Method;.. ;-))
And please, let us stop this endless nitpicking on the data! All of the
deleterious confounding influences, which you guys mention and discuss
in such great detail, can be taken care of by simple iterative corrections.
This is a straightforward mathemaical procedure which is well known to
physicists, other scientists, and engineers. I can recommend it warmly to
the entire epidemiological community. So if you really know what you
are talking about, go and make the corrections. ;:-))
As I said before, Bernie has amply demonstrated that all confounded
factors (Pun intended!) are likely to be small, and that they are not nearly
of the size required to bridge a difference of twenty-odd sigmas. Also,
Joe and I are inclined to agree with Bernie that K(D), the "Ecological
Fallacy" factor, which is the aggregate of all confounding factors, is likely
to be small. Here, D is the dose or a suitable substitute for the dose, .
But actually, we will now show that we do not need to make these
corrections for K(D)!
Why? Well, now, let us face facts! The set of large confounding
factors, the so-called "Ecological Fallacy", is a "Linearist" invention!
It is an attempt to save their favorite model, the "Linear No-Threshold
Model", from being consigned to the oblivion of the History of Science.
(With some foornotes on "How not to..."..).
Some of the confounding factors are legitimate, others are also,
but are vastly overblown. However, this will not really matter in what
follows.
Using the aggregate "Ecological Fallacy" factor K(D), you can
correct Cohen's data by a multiplication with 1/K(D) and arrive at
whatever dose-effect function underlies the radon/lung cancer
measurements. You believe that this cleaned up function is linear.
So be it! But when you try to apply this linear dose-effect function
to a risk problem in the real world, in other words, when you need to
predict what is going to happen in a recently exposed population,
which again has all its idiosyncrasies and peculiarities, you have to
multiply your linear dose-effect function by the "Ecological Fallacy"
factor K(D), and all your "Ecological Fallacy" factors cancel!
This is not trivial, because it is here that the Equivalent Population
Assumption, which I discussed on Saturday, comes in. It makes the
two K(D)s for test population and prediction population identical, so
that we have indeed K(D)/K(D) = 1. So now we are right back
where we started: at Bernie's raw data!
As I wrote Saturday, the "Ecological Fallacy" factor K(D) is only
needed when you are interested in the rather academic pursuit of
obtaining a cleaned up, pure dose-effect function which you assume to
be linear and which you are trying to prove linear. This is where you,
ANON, focus your attention. But actually, this function is of no use
whatsoever, particularly if you are right and K(D) is large. Eventually,
you want to estimate the number of lung cancers in the ‘real world‘,
and then the factor K(D) and your long discussions of the finer points
of epidemiology are quite irrelevant, because K(D) CANCELS!
YOU DO NOT NEED TO KNOW the linear model or any other
model which underlies the data. All you need are Bernie's data.
And so, our ‘thought experiment' deals with your arguments in all
your papers in one fell swoop: In risk assessment, you do not need to
know K(D), the influence of the "Ecological Fallacy"! And, finally,
that means that the "Ecological Fallacy" is a fallacious concept.
Fritz Seiler
PS: Sorry, for flaming an anonymous person, but some words in ANON's
message, quoted above, really rubbed me the wrong way!
**************************
NOTE CHANGE OF ADDRESS
***************************
Fritz A. Seiler, Ph.D.
Principal
Sigma Five Associates
P.O. Box 1709
Los Lunas, NM 87031
Tel. 505-866-5193
Fax. 505-866-5197
e-mail: faseiler@nmia.com
***************************
************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html