[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

No Subject



Dear Radsafers,

Franz Schoenhofer took the words right out of my ..... keyboard!  I too 
am more than a little bit tired of epidemiological studies with minute
but 'significant' effetcs.  In the seventeen years that I have, as a risk 
assessor, been following such studies with a view of applying them 
to risk analyses, I have become extremely suspicious of any 'effect' 
with an increase in the relative risk which is less than 1.5.  In fact, 
even the relative risk of  1.83 derived from a number quoted in the 
letter does not impress me as something to write home about.  Not 
without confirmation anyway.

This attitude arises from two reasons: First, a small relative risk such 
as the value of 1.05 quoted, does not result in a risk worth worrying 
about, because that disappears in the variability of the noise of the 
background incidence (the error of the standard error, a quantity 
neglected by most statistical and epidemiological methods, except in 
the t-test, for instance).  Second, all these error estimates assume 
that the numbers of individuals on study and the numbers of cases 
are the only sources of uncertainty (random errors).  In the face of the 
many sources of systematic errors which are inherent in all types of 
epidemiological studies, this assumption is evidently not true.  

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

Fritz A. Seiler

Institute for Regulatory Science 	Sigma Five Associates
Western Office			4101 Lara Dr. NE
P.O. Box 14006			Albuquerque, NM 87111
Albuquerque, NM 87191-4006		USA
USA

Tel:              xx-505-323-7848		xx-505-292-0392
Fax:             xx-505-293-3911		xx-505-293-3911

e-mail: 	 seiler@nrsi.org		faseiler@nmia.com

All statements represent the author's opinions and not the opinions
or policies of any firm or institution.

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