Re: Linear Model

["Wesley M. Dunn" <WDUNN@brc1.tdh.state.tx.us>]

I'd like to clarify one item which should be obvious:

"Peer-Review" does not make any published paper true or false.  It
means that it was subjected to a higher standard to attempt to ensure
that the research was performed in a scientific fashion and that the
published results accurately reflect the findings.

Other Stuff:

Yes, there are problems with ecological studies.  That is because
there are problems with ALL epidemiological studies.  The controls
can NEVER be identical to the cases (even identical twins aren't).
However, one can design a study to minimize the problems and maximize
the value.

The question I ask is: what does Cohen's latest paper (presented
at last month's HPS meeting) mean?  You can argue the fine details of
his methodology all you want.  However, if Radon is as deadly as the
EPA claims, there SHOULD have been a strong positive dose/effect
correlation for lung cancer, not a slight negative correlation.

(For those not there, randomly selected households were sent radon
detectors and questionaires.  He plotted known "fatal lung cancers/all
fatal cancers" for all residents of the household against measured
radon levels.  Average residency was about 19 years with total
knowledge of prior occupants out to about 23 years.  No matter how
he massaged the data, the best fit came out flat or negative (e.g.,
higher dose, lower lung cancer ratio).  3 Sigma would sometimes get
close to BEIR calculated levels.  If you want more, wait for his
paper to be published....)


Wes