AW: [ RadSafe ] Excess relative risk
Bernard L. Cohen
blc+ at pitt.edu
Mon Feb 18 11:25:23 CST 2008
I am not sure what the question is, but if I interpret it correctly, you
are asking how I use my method of stratification to treat confounding
factors. As an example, consider data on cancer mortality risk, M, vs
radiation exposure, r, where the confounder is economic status as
represented by annual family dollar income, D. I divide the data into
10 separate data files as (1) cases where D < $5000, (2) cases where D
= $5000-10,000, (3) cases where D = $10,000-$20,000, .......(9) cases
where D= $100,000 - $150,000, (10) cases where D > $150,000. Note that
for each of these ten data files, confounding by D is of greatly reduced
importance because all cases in the file have essentially the same value
of D. For each of these ten data files, I study M vs r. If the result
for all ten give reasonably consistent results for M vs r, an average of
these gives the result I am seeking, M vs r, free of confounding by D.
John Jacobus wrote:
> Dr. Cohen,
> That being sais, how do you treat your data sets. Do you obtain your
> results linear or multiple regression, or what?
>
> */"Bernard L. Cohen" <blc+ at pitt.edu>/* wrote:
>
> THOMAS POTTER wrote:
> > I'm no epidemiologist or even a statistician, for that matter,
> though I've studied statistics some. But I think Dr. Cohen's
> position re epidemiology is way too radical. Epidemiologists
> develop hypotheses. They design experiments to test them, taking
> care not to introduce serious biases into the design. They
> evaluate data using mathematically rigorous methods that analyze
> the entire data set, not just a few points that might look
> interesting based on potentially misleading eyeball inspection.
> What's not scientific about this approach? Dr. Cohen's response to
> the Iowa grad student should have been, "How so?"
> >
> > Tom Potter
> >
> >
> All of that is staying inside the box, refusing to think outside the
> box. Epidemiology 101 is a treatment of the inside of the box. Of
> course
> thinking outside the box is fraught with danger and must be done
> carefully, but that doesn't mean it should not be done. It has led to
> great progress in other areas of science. For me, staying inside
> the box
> makes epidemiology a technology rather than a science..
>
> >
> > Date: Tue, 12 Feb 2008 15:47:57 -0500
> > From: "Bernard L. Cohen"
> > Subject: Re: AW: [ RadSafe ] Excess relative risk
> > To: John Jacobus
> > Cc: Rainer.Facius at dlr.de, radsafe at radlab.nl
> > Message-ID: <47B205FD.7020501 at pitt.edu>
> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> >
> > I hope this does not sound too radical: In my experience,
> > epidemiologists are not *scientists* in the broad sense of that
> term. They practice a *technology* which has been "boxed" and they
> are not willing to think outside that box. For any new ideas or
> approaches, it is "their way or the highway"; they reject them
> with only the shallowest reasoning, or more commonly ignore them.
> That is not the way scientists have traditionally operated. If
> physicists operated that way, we would never have had quantum
> theory, relativity, or any of the many great advances of 20th
> century physics.
> > Their box has been successful in many applications, but it is
> prone to failure where statistics are marginal or where there can
> be multiple confounding factors. They customarily treat the latter
> with multiple regression analysis which is fraught with dangers --
> they call this "adjusting for confounding factors", factors they
> select with something less than air-tight reasoning.
> > If 2 or 3 epidemiologists take a position, a "consensus" forms
> without those who constitute the consensus spending any time or
> thought on the subject. They just don't like to think outside
> their box. They seem to have no experience in doing that.
> >
> > Date: Wed, 13 Feb 2008 11:44:37 +1100
> > From: "Brent Rogers"
> > Subject: RE: AW: [ RadSafe ] Excess relative risk
> > To: "'Bernard L. Cohen'" , "'John Jacobus'"
> >
> > Cc: Rainer.Facius at dlr.de, radsafe at radlab.nl
> > Message-ID: <000901c86dd9$a59cb2e0$7400a8c0 at brent28417bbe6>
> > Content-Type: text/plain; charset="iso-8859-1"
> >
> > Great reply, Professor Cohen. I'm pleasantly reminded of the
> best rejoinder
> > I ever saw on RadSafe. One that you provided about 10 years ago,
> and I
> > paraphrase now:
> >
> > University of Iowa epidemiology grad student: "Your method violates
> > Epidemiology 101."
> >
> > Prof Cohen: "I am bound by the Scientific Method, not
> Epidemiology 101."
> >
> > I'm still waiting for the opportunity to use a similar line myself.
> >
> > Brent Rogers
> >
>
>
> --
> Bernard L. Cohen
> Physics Dept., University of Pittsburgh
> Pittsburgh, PA 15260
> Tel: (412)624-9245 Fax: (412)624-9163
> e-mail: blc at pitt.edu web site: http://www.phyast.pitt.edu/~blc
>
>
> _______________________________________________
> You are currently subscribed to the RadSafe mailing list
>
> Before posting a message to RadSafe be sure to have read and
> understood the RadSafe rules. These can be found at:
> http://radlab.nl/radsafe/radsaferules.html
>
> For information on how to subscribe or unsubscribe and other
> settings visit: http://radlab.nl/radsafe/
>
>
>
>
> +++++++++++++++++++
> "If history teaches any lesson it is that no nation has an inherent
> right to greatness. Greatness has to be earned and continually re-earned."
> - Norman Augustine, Chairman of the National Academies Committee
>
> -- John
> John Jacobus, MS
> Certified Health Physicist
> e-mail: crispy_bird at yahoo.com
>
> ------------------------------------------------------------------------
> Never miss a thing. Make Yahoo your homepage.
> <http://us.rd.yahoo.com/evt=51438/*http://www.yahoo.com/r/hs>
--
Bernard L. Cohen
Physics Dept., University of Pittsburgh
Pittsburgh, PA 15260
Tel: (412)624-9245 Fax: (412)624-9163
e-mail: blc at pitt.edu web site: http://www.phyast.pitt.edu/~blc
More information about the RadSafe
mailing list