EMF II

["Fritz A. Seiler" <faseiler@nmia.com>]

Another comment by Fritz Seiler and Joe Alvarez
on the EMF problem and attempts to keep it alive.

Let us first respond to Jim Dukelow's quotes from
the committee report:

On reading Jim's excerpts, one thing is immediately
apparent: Whenever the subject is on the potential
mechanisms of causation, the report abounds in the
‘might', ‘could', ‘would', and ‘should' forms of the
conjunctive.  Logically this indicates that there exists
another condition which has to be met also.  Well,
all these conditions start with the word "if"!  So, as
far as the Scientific Methods is concerned, there is
simply nothing of substance here, except some rank
speculations.  Even the case of carcinogen treated
mice which is stated in an active positive form, does
require INTENSE magnetic fields, and even then there
is that condition "if that also happens at the minute
fields of external origin at the cellular level, which
are drowned out by the thermal noise in the cell which
is about 1,000 times larger ..."!  Also that relation
could be what we like to call a "Coney Island Fallacy"
(On a hot day on Coney Island, more hamburgers are
sold and more people drown.  So ... ).

The epidemiological evidence is also unconvincing.  We
note with a certain amusement that in over 80 studies
about four or eight have to exceed the confidence limits
for 95 and 90 percent, respectively.  No attempt has
been made to discuss this or prevent such obviously
‘significant' effects by applying the Bonferroni or Sidak
inequalities in evaluations of the additive or multiplicative
formulations of familywise confidence limits.

Another serious deficiency of the epidemiological data is
the fact that significance is determined by using random
errors (size of sample) only.  This still is the usual procedure,
although it is well known (and we mentioned it in our last
post), that systematic errors can be sizeable and wipe out
what little significance there might be left in the data.  We do
not feel comfortable at all with the relative risk of 1.5
quoted for the relative incidence of childhood leukemia for
the highest wiring category, unless there is confirmation at
the same or a higher level of confidence, and to our
knowledge, there is none.

Now let us assume a 95% confidence level and ask: are the
scientists in charge really certain that their experimental value
has a systematic error which is decisively lower than 25%?
Say, 10%?  Then, and only then, can we declare an unknown
correlation to be the culprit.  In such a study, a claim of
ten percent systematic error or less would require a highly
confident experimenter, ... or, maybe, an overly confident
experimenter.  For that reason, many epidemiologists
have little or no confidence in relative risks that lie below
1.8 or 2.  And we think quite rightfully so.

As we said before, as long as there is no correlation and no
mechanism, the outcome is "innocent" and not "not guilty".
Also, there are two more aspects to be covered in this
context.  One with regard to terminology and the other
with regard to the dictum quoted by Jim Dukelow.

First, terminology: Most lay people and - unfortunately -
many scientists equate "possible" with "probable". Often,
they subscribe to the statement "If it is possible, you
must be able to calculate a probability, and if you do get
a number for it, that means it will happen sooner or later!"
As scientists, we know or should know that we never really
can demonstrate a certainty or an impossibility.  In either
case, the universe is not old enough to lend support to
that kind of statement.

So, let us consider a simple case in another field, canine
distemper, for instance.  If untreated, most dogs will die
from it, but it has not been found in man.  Is a transfer
impossible, then?  We do not think so.  But is it possible?
Yes, it is, but it is also improbable.  But since it is such a
devastating disease, do we need a research program, just
in case?  No!  There is neither a transfer mechanism, nor
any epidemiological evidence for it.  So is the proper
verdict "Not guilty", then?  No, until there is some
evidence to the contrary, it is clearly "innocent"!  Note
that a ‘not guilty' verdict implies a test with a particular
confidence level, and a test cannot be done in the
absence of data.  So, ‘innocent' is the proper term.

Second, there is the dictum that Jim Dukelow quoted:
"Absence of evidence is not evidence of absence."
Actually, it is often quoted in the more radical form
"Absence of proof does not imply proof of absence."
It has been used several times on this mailing list in
the past and we have been waiting for an occasion to
jump on it.  (So you are the victim chosen, Jim!)  We
are quite certain that most people have not really
thought about what is said in this statement, because
it is misleading to say the least.  Quite apart from the
ill-advised term "proof" in the second version, there
is, quite generally, the fact that there is no way in
which a measurement can demonstrate a value of
exactly zero for any quantity, i.e., the absence of the
corresponding term.  There is always a threshold of
detection which defines an upper limit or an upper and
a lower limit for the quantity.

But mainly, this dictum is nothing short of anti-scientific!
Just consider the equivalent statement: "An absence of
evidence of toxicity is not evidence for an absence of
toxicity".  Or to put it a bit more succinctly "Guilty until
proven innocent".  These statements have more the
sound of politically motivated demagoguery, not that of
sound science.

To state it quite clearly, the absence of a toxic effect
caused by an agent (and that includes radiation!) is
demonstrated by the inability of the data to support
the rejection of the null hypothesis (no effect) at a
given confidence level.  This level may range anywhere
from 50% to more than 95% and is an important choice
to be made by the analyst.  Now, if you cannot reject
the hypothesis, it remains in force.  And that means that
there is insufficient evidence for the existence of an
effect.  Or more succinctly and in the words of the null
hypothesis: There is NO effect!

So in the framework of the Scientific Method, this means
that, for the confidence limit chosen, there is truth in a
statement which runs directly counter the epidemiological
dictum cited by Jim Dukelow:

 Absence of Evidence IS Evidence of Absence!



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

Fritz A. Seiler, Ph.D.
Principal
Sigma Five Associates
P.O. Box 14006
Albuquerque, NM 87191-4006
Tel.     505-323-7848
Fax.    505-293-3911
e-mail: faseiler@nmia.com

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