Hello EBH group (sorry, i deleted Ted Harding's original message)
Just a quick response to the message, which raised some very interesting points:
a) if there is a concern about using p-values with an arbitrary threshold value of 0.05 (or less than
before something is statistcally significant) as the basis on which to introduce a new intervention
and to replace an existing one, then why not change the p-value to 0.001 (and use 99% confidence intervals)
before you suggest that there should be a change in practice.
b) i don't think clinical practice is changed solely on the basis of statistics - there is a whole array of other factors
that are taken into consideration such as clinical significance (i.e. how a diagnostic technology X may change the
management of a patient in comparison with technology Y), acceptability of a new technology in comparison with an
existing one to the relevant patients and professionals, medico-legal considerations etc.
c) perhaps there are occassions when it is appropriate to set the p-value at 0.05.
Although statistics may be inherent in the process of designing the study and in various facets of assesing a health
technology, it must be a rather sweeping statement to discredit all health research on the basis of the sometimes
arbitrary use of p-values (probably better to state what the explicit p-value)?
[I think Matthews (the journalist?) must have recently read "Zen and the Art of Motorcycling Maintenance"
by Robert Pirzig].
Stephen Brealey
York University
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|