Rolf R. Engel" wrote on Mon, Jun 14, 1999, 5:08 pm
>
>The first part is an oversimplification that does not take into
>consideration to what purpose a study has been done. Why not report
>relative risks to state causality, and absolute risks to justify
>treatments?
>
>R. Engel
neither relative risk nor absolute risk (and its inverse, nnt) prove
causality - they demonstrate differences between groups within a given
study. whether we choose to believe the difference is caused by an
intervention will depend on either 1. the chances of the difference arising
bering less than a specified fig (traditionally p<0.05) OR 2. whether the
confidence intervals around the difference are sufficiently narrow - the
latter option has gained preference over the past 15 years - see previous
discussions.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|